Table Of Contents

Search

Enter search terms or a module, class or function name.

What’s New

These are new features and improvements of note in each release.

v0.23.2

This is a minor bug-fix release in the 0.23.x series and includes some small regression fixes and bug fixes. We recommend that all users upgrade to this version.

Note

Pandas 0.23.2 is first pandas release that’s compatible with Python 3.7 (GH20552)

Logical Reductions over Entire DataFrame

DataFrame.all() and DataFrame.any() now accept axis=None to reduce over all axes to a scalar (GH19976)

In [1]: df = pd.DataFrame({"A": [1, 2], "B": [True, False]})

In [2]: df.all(axis=None)
Out[2]: False

This also provides compatibility with NumPy 1.15, which now dispatches to DataFrame.all. With NumPy 1.15 and pandas 0.23.1 or earlier, numpy.all() will no longer reduce over every axis:

>>> # NumPy 1.15, pandas 0.23.1
>>> np.any(pd.DataFrame({"A": [False], "B": [False]}))
A    False
B    False
dtype: bool

With pandas 0.23.2, that will correctly return False, as it did with NumPy < 1.15.

In [3]: np.any(pd.DataFrame({"A": [False], "B": [False]}))
Out[3]: False

Fixed Regressions

  • Fixed regression in to_csv() when handling file-like object incorrectly (GH21471)
  • Re-allowed duplicate level names of a MultiIndex. Accessing a level that has a duplicate name by name still raises an error (GH19029).
  • Bug in both DataFrame.first_valid_index() and Series.first_valid_index() raised for a row index having duplicate values (GH21441)
  • Fixed printing of DataFrames with hierarchical columns with long names (GH21180)
  • Fixed regression in reindex() and groupby() with a MultiIndex or multiple keys that contains categorical datetime-like values (GH21390).
  • Fixed regression in unary negative operations with object dtype (GH21380)
  • Bug in Timestamp.ceil() and Timestamp.floor() when timestamp is a multiple of the rounding frequency (GH21262)
  • Fixed regression in to_clipboard() that defaulted to copying dataframes with space delimited instead of tab delimited (GH21104)

Build Changes

  • The source and binary distributions no longer include test data files, resulting in smaller download sizes. Tests relying on these data files will be skipped when using pandas.test(). (GH19320)

Bug Fixes

Conversion

Indexing

I/O

  • Bug in read_csv() that caused it to incorrectly raise an error when nrows=0, low_memory=True, and index_col was not None (GH21141)
  • Bug in json_normalize() when formatting the record_prefix with integer columns (GH21536)

Categorical

  • Bug in rendering Series with Categorical dtype in rare conditions under Python 2.7 (GH21002)

Timezones

  • Bug in Timestamp and DatetimeIndex where passing a Timestamp localized after a DST transition would return a datetime before the DST transition (GH20854)
  • Bug in comparing DataFrame`s with tz-aware :class:`DatetimeIndex columns with a DST transition that raised a KeyError (GH19970)

Timedelta

  • Bug in Timedelta where non-zero timedeltas shorter than 1 microsecond were considered False (GH21484)

v0.23.1

This is a minor bug-fix release in the 0.23.x series and includes some small regression fixes and bug fixes. We recommend that all users upgrade to this version.

Fixed Regressions

Comparing Series with datetime.date

We’ve reverted a 0.23.0 change to comparing a Series holding datetimes and a datetime.date object (GH21152). In pandas 0.22 and earlier, comparing a Series holding datetimes and datetime.date objects would coerce the datetime.date to a datetime before comapring. This was inconsistent with Python, NumPy, and DatetimeIndex, which never consider a datetime and datetime.date equal.

In 0.23.0, we unified operations between DatetimeIndex and Series, and in the process changed comparisons between a Series of datetimes and datetime.date without warning.

We’ve temporarily restored the 0.22.0 behavior, so datetimes and dates may again compare equal, but restore the 0.23.0 behavior in a future release.

To summarize, here’s the behavior in 0.22.0, 0.23.0, 0.23.1:

# 0.22.0... Silently coerce the datetime.date
>>> Series(pd.date_range('2017', periods=2)) == datetime.date(2017, 1, 1)
0     True
1    False
dtype: bool

# 0.23.0... Do not coerce the datetime.date
>>> Series(pd.date_range('2017', periods=2)) == datetime.date(2017, 1, 1)
0    False
1    False
dtype: bool

# 0.23.1... Coerce the datetime.date with a warning
>>> Series(pd.date_range('2017', periods=2)) == datetime.date(2017, 1, 1)
/bin/python:1: FutureWarning: Comparing Series of datetimes with 'datetime.date'.  Currently, the
'datetime.date' is coerced to a datetime. In the future pandas will
not coerce, and the values not compare equal to the 'datetime.date'.
To retain the current behavior, convert the 'datetime.date' to a
datetime with 'pd.Timestamp'.
  #!/bin/python3
0     True
1    False
dtype: bool

In addition, ordering comparisons will raise a TypeError in the future.

Other Fixes

  • Reverted the ability of to_sql() to perform multivalue inserts as this caused regression in certain cases (GH21103). In the future this will be made configurable.
  • Fixed regression in the DatetimeIndex.date and DatetimeIndex.time attributes in case of timezone-aware data: DatetimeIndex.time returned a tz-aware time instead of tz-naive (GH21267) and DatetimeIndex.date returned incorrect date when the input date has a non-UTC timezone (GH21230).
  • Fixed regression in pandas.io.json.json_normalize() when called with None values in nested levels in JSON, and to not drop keys with value as None (GH21158, GH21356).
  • Bug in to_csv() causes encoding error when compression and encoding are specified (GH21241, GH21118)
  • Bug preventing pandas from being importable with -OO optimization (GH21071)
  • Bug in Categorical.fillna() incorrectly raising a TypeError when value the individual categories are iterable and value is an iterable (GH21097, GH19788)
  • Fixed regression in constructors coercing NA values like None to strings when passing dtype=str (GH21083)
  • Regression in pivot_table() where an ordered Categorical with missing values for the pivot’s index would give a mis-aligned result (GH21133)
  • Fixed regression in merging on boolean index/columns (GH21119).

Performance Improvements

  • Improved performance of CategoricalIndex.is_monotonic_increasing(), CategoricalIndex.is_monotonic_decreasing() and CategoricalIndex.is_monotonic() (GH21025)
  • Improved performance of CategoricalIndex.is_unique() (GH21107)

Bug Fixes

Groupby/Resample/Rolling

Data-type specific

Sparse

  • Bug in SparseArray.shape which previously only returned the shape SparseArray.sp_values (GH21126)

Indexing

  • Bug in Series.reset_index() where appropriate error was not raised with an invalid level name (GH20925)
  • Bug in interval_range() when start/periods or end/periods are specified with float start or end (GH21161)
  • Bug in MultiIndex.set_names() where error raised for a MultiIndex with nlevels == 1 (GH21149)
  • Bug in IntervalIndex constructors where creating an IntervalIndex from categorical data was not fully supported (GH21243, GH21253)
  • Bug in MultiIndex.sort_index() which was not guaranteed to sort correctly with level=1; this was also causing data misalignment in particular DataFrame.stack() operations (GH20994, GH20945, GH21052)

Plotting

  • New keywords (sharex, sharey) to turn on/off sharing of x/y-axis by subplots generated with pandas.DataFrame().groupby().boxplot() (GH20968)

I/O

  • Bug in IO methods specifying compression='zip' which produced uncompressed zip archives (GH17778, GH21144)
  • Bug in DataFrame.to_stata() which prevented exporting DataFrames to buffers and most file-like objects (GH21041)
  • Bug in read_stata() and StataReader which did not correctly decode utf-8 strings on Python 3 from Stata 14 files (dta version 118) (GH21244)
  • Bug in IO JSON read_json() reading empty JSON schema with orient='table' back to DataFrame caused an error (GH21287)

Reshaping

  • Bug in concat() where error was raised in concatenating Series with numpy scalar and tuple names (GH21015)
  • Bug in concat() warning message providing the wrong guidance for future behavior (GH21101)

Other

  • Tab completion on Index in IPython no longer outputs deprecation warnings (GH21125)
  • Bug preventing pandas being used on Windows without C++ redistributable installed (GH21106)

v0.23.0 (May 15, 2018)

This is a major release from 0.22.0 and includes a number of API changes, deprecations, new features, enhancements, and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this version.

Highlights include:

Check the API Changes and deprecations before updating.

Warning

Starting January 1, 2019, pandas feature releases will support Python 3 only. See Plan for dropping Python 2.7 for more.

What’s new in v0.23.0

New features

JSON read/write round-trippable with orient='table'

A DataFrame can now be written to and subsequently read back via JSON while preserving metadata through usage of the orient='table' argument (see GH18912 and GH9146). Previously, none of the available orient values guaranteed the preservation of dtypes and index names, amongst other metadata.

In [1]: df = pd.DataFrame({'foo': [1, 2, 3, 4],
   ...:                    'bar': ['a', 'b', 'c', 'd'],
   ...:                    'baz': pd.date_range('2018-01-01', freq='d', periods=4),
   ...:                    'qux': pd.Categorical(['a', 'b', 'c', 'c'])
   ...:                    }, index=pd.Index(range(4), name='idx'))
   ...: 

In [2]: df
Out[2]: 
     foo bar        baz qux
idx                        
0      1   a 2018-01-01   a
1      2   b 2018-01-02   b
2      3   c 2018-01-03   c
3      4   d 2018-01-04   c

In [3]: df.dtypes
Out[3]: 
foo             int64
bar            object
baz    datetime64[ns]
qux          category
dtype: object

In [4]: df.to_json('test.json', orient='table')

In [5]: new_df = pd.read_json('test.json', orient='table')

In [6]: new_df
Out[6]: 
     foo bar        baz qux
idx                        
0      1   a 2018-01-01   a
1      2   b 2018-01-02   b
2      3   c 2018-01-03   c
3      4   d 2018-01-04   c

In [7]: new_df.dtypes
Out[7]: 
foo             int64
bar            object
baz    datetime64[ns]
qux          category
dtype: object

Please note that the string index is not supported with the round trip format, as it is used by default in write_json to indicate a missing index name.

In [8]: df.index.name = 'index'

In [9]: df.to_json('test.json', orient='table')

In [10]: new_df = pd.read_json('test.json', orient='table')

In [11]: new_df
Out[11]: 
   foo bar        baz qux
0    1   a 2018-01-01   a
1    2   b 2018-01-02   b
2    3   c 2018-01-03   c
3    4   d 2018-01-04   c

In [12]: new_df.dtypes
Out[12]: 
foo             int64
bar            object
baz    datetime64[ns]
qux          category
dtype: object

.assign() accepts dependent arguments

The DataFrame.assign() now accepts dependent keyword arguments for python version later than 3.6 (see also PEP 468). Later keyword arguments may now refer to earlier ones if the argument is a callable. See the documentation here (GH14207)

In [13]: df = pd.DataFrame({'A': [1, 2, 3]})

In [14]: df
Out[14]: 
   A
0  1
1  2
2  3

In [15]: df.assign(B=df.A, C=lambda x:x['A']+ x['B'])
Out[15]: 
   A  B  C
0  1  1  2
1  2  2  4
2  3  3  6

Warning

This may subtly change the behavior of your code when you’re using .assign() to update an existing column. Previously, callables referring to other variables being updated would get the “old” values

Previous Behavior:

In [2]: df = pd.DataFrame({"A": [1, 2, 3]})

In [3]: df.assign(A=lambda df: df.A + 1, C=lambda df: df.A * -1)
Out[3]:
   A  C
0  2 -1
1  3 -2
2  4 -3

New Behavior:

In [16]: df.assign(A=df.A+1, C= lambda df: df.A* -1)
Out[16]: 
   A  C
0  2 -2
1  3 -3
2  4 -4

Merging on a combination of columns and index levels

Strings passed to DataFrame.merge() as the on, left_on, and right_on parameters may now refer to either column names or index level names. This enables merging DataFrame instances on a combination of index levels and columns without resetting indexes. See the Merge on columns and levels documentation section. (GH14355)

In [17]: left_index = pd.Index(['K0', 'K0', 'K1', 'K2'], name='key1')

In [18]: left = pd.DataFrame({'A': ['A0', 'A1', 'A2', 'A3'],
   ....:                      'B': ['B0', 'B1', 'B2', 'B3'],
   ....:                      'key2': ['K0', 'K1', 'K0', 'K1']},
   ....:                     index=left_index)
   ....: 

In [19]: right_index = pd.Index(['K0', 'K1', 'K2', 'K2'], name='key1')

In [20]: right = pd.DataFrame({'C': ['C0', 'C1', 'C2', 'C3'],
   ....:                       'D': ['D0', 'D1', 'D2', 'D3'],
   ....:                       'key2': ['K0', 'K0', 'K0', 'K1']},
   ....:                      index=right_index)
   ....: 

In [21]: left.merge(right, on=['key1', 'key2'])
Out[21]: 
       A   B key2   C   D
key1                     
K0    A0  B0   K0  C0  D0
K1    A2  B2   K0  C1  D1
K2    A3  B3   K1  C3  D3

Sorting by a combination of columns and index levels

Strings passed to DataFrame.sort_values() as the by parameter may now refer to either column names or index level names. This enables sorting DataFrame instances by a combination of index levels and columns without resetting indexes. See the Sorting by Indexes and Values documentation section. (GH14353)

# Build MultiIndex
In [22]: idx = pd.MultiIndex.from_tuples([('a', 1), ('a', 2), ('a', 2),
   ....:                                  ('b', 2), ('b', 1), ('b', 1)])
   ....: 

In [23]: idx.names = ['first', 'second']

# Build DataFrame
In [24]: df_multi = pd.DataFrame({'A': np.arange(6, 0, -1)},
   ....:                         index=idx)
   ....: 

In [25]: df_multi
Out[25]: 
              A
first second   
a     1       6
      2       5
      2       4
b     2       3
      1       2
      1       1

# Sort by 'second' (index) and 'A' (column)
In [26]: df_multi.sort_values(by=['second', 'A'])
Out[26]: 
              A
first second   
b     1       1
      1       2
a     1       6
b     2       3
a     2       4
      2       5

Extending Pandas with Custom Types (Experimental)

Pandas now supports storing array-like objects that aren’t necessarily 1-D NumPy arrays as columns in a DataFrame or values in a Series. This allows third-party libraries to implement extensions to NumPy’s types, similar to how pandas implemented categoricals, datetimes with timezones, periods, and intervals.

As a demonstration, we’ll use cyberpandas, which provides an IPArray type for storing ip addresses.

In [1]: from cyberpandas import IPArray

In [2]: values = IPArray([
   ...:     0,
   ...:     3232235777,
   ...:     42540766452641154071740215577757643572
   ...: ])
   ...:
   ...:

IPArray isn’t a normal 1-D NumPy array, but because it’s a pandas ~pandas.api.extension.ExtensionArray, it can be stored properly inside pandas’ containers.

In [3]: ser = pd.Series(values)

In [4]: ser
Out[4]:
0                         0.0.0.0
1                     192.168.1.1
2    2001:db8:85a3::8a2e:370:7334
dtype: ip

Notice that the dtype is ip. The missing value semantics of the underlying array are respected:

In [5]: ser.isna()
Out[5]:
0     True
1    False
2    False
dtype: bool

For more, see the extension types documentation. If you build an extension array, publicize it on our ecosystem page.

New observed keyword for excluding unobserved categories in groupby

Grouping by a categorical includes the unobserved categories in the output. When grouping by multiple categorical columns, this means you get the cartesian product of all the categories, including combinations where there are no observations, which can result in a large number of groups. We have added a keyword observed to control this behavior, it defaults to observed=False for backward-compatiblity. (GH14942, GH8138, GH15217, GH17594, GH8669, GH20583, GH20902)

In [27]: cat1 = pd.Categorical(["a", "a", "b", "b"],
   ....:                       categories=["a", "b", "z"], ordered=True)
   ....: 

In [28]: cat2 = pd.Categorical(["c", "d", "c", "d"],
   ....:                       categories=["c", "d", "y"], ordered=True)
   ....: 

In [29]: df = pd.DataFrame({"A": cat1, "B": cat2, "values": [1, 2, 3, 4]})

In [30]: df['C'] = ['foo', 'bar'] * 2

In [31]: df
Out[31]: 
   A  B  values    C
0  a  c       1  foo
1  a  d       2  bar
2  b  c       3  foo
3  b  d       4  bar

To show all values, the previous behavior:

In [32]: df.groupby(['A', 'B', 'C'], observed=False).count()
Out[32]: 
         values
A B C          
a c bar     NaN
    foo     1.0
  d bar     1.0
    foo     NaN
  y bar     NaN
    foo     NaN
b c bar     NaN
...         ...
  y foo     NaN
z c bar     NaN
    foo     NaN
  d bar     NaN
    foo     NaN
  y bar     NaN
    foo     NaN

[18 rows x 1 columns]

To show only observed values:

In [33]: df.groupby(['A', 'B', 'C'], observed=True).count()
Out[33]: 
         values
A B C          
a c foo       1
  d bar       1
b c foo       1
  d bar       1

For pivotting operations, this behavior is already controlled by the dropna keyword:

In [34]: cat1 = pd.Categorical(["a", "a", "b", "b"],
   ....:                       categories=["a", "b", "z"], ordered=True)
   ....: 

In [35]: cat2 = pd.Categorical(["c", "d", "c", "d"],
   ....:                       categories=["c", "d", "y"], ordered=True)
   ....: 

In [36]: df = DataFrame({"A": cat1, "B": cat2, "values": [1, 2, 3, 4]})

In [37]: df
Out[37]: 
   A  B  values
0  a  c       1
1  a  d       2
2  b  c       3
3  b  d       4
In [38]: pd.pivot_table(df, values='values', index=['A', 'B'],
   ....:                dropna=True)
   ....: 
Out[38]: 
     values
A B        
a c       1
  d       2
b c       3
  d       4

In [39]: pd.pivot_table(df, values='values', index=['A', 'B'],
   ....:                dropna=False)
   ....: 
Out[39]: 
     values
A B        
a c     1.0
  d     2.0
  y     NaN
b c     3.0
  d     4.0
  y     NaN
z c     NaN
  d     NaN
  y     NaN

Rolling/Expanding.apply() accepts raw=False to pass a Series to the function

Series.rolling().apply(), DataFrame.rolling().apply(), Series.expanding().apply(), and DataFrame.expanding().apply() have gained a raw=None parameter. This is similar to DataFame.apply(). This parameter, if True allows one to send a np.ndarray to the applied function. If False a Series will be passed. The default is None, which preserves backward compatibility, so this will default to True, sending an np.ndarray. In a future version the default will be changed to False, sending a Series. (GH5071, GH20584)

In [40]: s = pd.Series(np.arange(5), np.arange(5) + 1)

In [41]: s
Out[41]: 
1    0
2    1
3    2
4    3
5    4
dtype: int64

Pass a Series:

In [42]: s.rolling(2, min_periods=1).apply(lambda x: x.iloc[-1], raw=False)
Out[42]: 
1    0.0
2    1.0
3    2.0
4    3.0
5    4.0
dtype: float64

Mimic the original behavior of passing a ndarray:

In [43]: s.rolling(2, min_periods=1).apply(lambda x: x[-1], raw=True)
Out[43]: 
1    0.0
2    1.0
3    2.0
4    3.0
5    4.0
dtype: float64

DataFrame.interpolate has gained the limit_area kwarg

DataFrame.interpolate() has gained a limit_area parameter to allow further control of which NaN s are replaced. Use limit_area='inside' to fill only NaNs surrounded by valid values or use limit_area='outside' to fill only NaN s outside the existing valid values while preserving those inside. (GH16284) See the full documentation here.

In [44]: ser = pd.Series([np.nan, np.nan, 5, np.nan, np.nan, np.nan, 13, np.nan, np.nan])

In [45]: ser
Out[45]: 
0     NaN
1     NaN
2     5.0
3     NaN
4     NaN
5     NaN
6    13.0
7     NaN
8     NaN
dtype: float64

Fill one consecutive inside value in both directions

In [46]: ser.interpolate(limit_direction='both', limit_area='inside', limit=1)
Out[46]: 
0     NaN
1     NaN
2     5.0
3     7.0
4     NaN
5    11.0
6    13.0
7     NaN
8     NaN
dtype: float64

Fill all consecutive outside values backward

In [47]: ser.interpolate(limit_direction='backward', limit_area='outside')
Out[47]: 
0     5.0
1     5.0
2     5.0
3     NaN
4     NaN
5     NaN
6    13.0
7     NaN
8     NaN
dtype: float64

Fill all consecutive outside values in both directions

In [48]: ser.interpolate(limit_direction='both', limit_area='outside')
Out[48]: 
0     5.0
1     5.0
2     5.0
3     NaN
4     NaN
5     NaN
6    13.0
7    13.0
8    13.0
dtype: float64

get_dummies now supports dtype argument

The get_dummies() now accepts a dtype argument, which specifies a dtype for the new columns. The default remains uint8. (GH18330)

In [49]: df = pd.DataFrame({'a': [1, 2], 'b': [3, 4], 'c': [5, 6]})

In [50]: pd.get_dummies(df, columns=['c']).dtypes
Out[50]: 
a      int64
b      int64
c_5    uint8
c_6    uint8
dtype: object

In [51]: pd.get_dummies(df, columns=['c'], dtype=bool).dtypes
Out[51]: 
a      int64
b      int64
c_5     bool
c_6     bool
dtype: object

Timedelta mod method

mod (%) and divmod operations are now defined on Timedelta objects when operating with either timedelta-like or with numeric arguments. See the documentation here. (GH19365)

In [52]: td = pd.Timedelta(hours=37)

In [53]: td % pd.Timedelta(minutes=45)
Out[53]: Timedelta('0 days 00:15:00')

.rank() handles inf values when NaN are present

In previous versions, .rank() would assign inf elements NaN as their ranks. Now ranks are calculated properly. (GH6945)

In [54]: s = pd.Series([-np.inf, 0, 1, np.nan, np.inf])

In [55]: s
Out[55]: 
0        -inf
1    0.000000
2    1.000000
3         NaN
4         inf
dtype: float64

Previous Behavior:

In [11]: s.rank()
Out[11]:
0    1.0
1    2.0
2    3.0
3    NaN
4    NaN
dtype: float64

Current Behavior:

In [56]: s.rank()
Out[56]: 
0    1.0
1    2.0
2    3.0
3    NaN
4    4.0
dtype: float64

Furthermore, previously if you rank inf or -inf values together with NaN values, the calculation won’t distinguish NaN from infinity when using ‘top’ or ‘bottom’ argument.

In [57]: s = pd.Series([np.nan, np.nan, -np.inf, -np.inf])

In [58]: s
Out[58]: 
0    NaN
1    NaN
2   -inf
3   -inf
dtype: float64

Previous Behavior:

In [15]: s.rank(na_option='top')
Out[15]:
0    2.5
1    2.5
2    2.5
3    2.5
dtype: float64

Current Behavior:

In [59]: s.rank(na_option='top')
Out[59]: 
0    1.5
1    1.5
2    3.5
3    3.5
dtype: float64

These bugs were squashed:

  • Bug in DataFrame.rank() and Series.rank() when method='dense' and pct=True in which percentile ranks were not being used with the number of distinct observations (GH15630)
  • Bug in Series.rank() and DataFrame.rank() when ascending='False' failed to return correct ranks for infinity if NaN were present (GH19538)
  • Bug in DataFrameGroupBy.rank() where ranks were incorrect when both infinity and NaN were present (GH20561)

Series.str.cat has gained the join kwarg

Previously, Series.str.cat() did not – in contrast to most of pandas – align Series on their index before concatenation (see GH18657). The method has now gained a keyword join to control the manner of alignment, see examples below and here.

In v.0.23 join will default to None (meaning no alignment), but this default will change to 'left' in a future version of pandas.

In [60]: s = pd.Series(['a', 'b', 'c', 'd'])

In [61]: t = pd.Series(['b', 'd', 'e', 'c'], index=[1, 3, 4, 2])

In [62]: s.str.cat(t)
Out[62]: 
0    ab
1    bd
2    ce
3    dc
dtype: object

In [63]: s.str.cat(t, join='left', na_rep='-')
Out[63]: 
0    a-
1    bb
2    cc
3    dd
dtype: object

Furthermore, Series.str.cat() now works for CategoricalIndex as well (previously raised a ValueError; see GH20842).

DataFrame.astype performs column-wise conversion to Categorical

DataFrame.astype() can now perform column-wise conversion to Categorical by supplying the string 'category' or a CategoricalDtype. Previously, attempting this would raise a NotImplementedError. See the Object Creation section of the documentation for more details and examples. (GH12860, GH18099)

Supplying the string 'category' performs column-wise conversion, with only labels appearing in a given column set as categories:

In [64]: df = pd.DataFrame({'A': list('abca'), 'B': list('bccd')})

In [65]: df = df.astype('category')

In [66]: df['A'].dtype
Out[66]: CategoricalDtype(categories=['a', 'b', 'c'], ordered=False)

In [67]: df['B'].dtype
Out[67]: CategoricalDtype(categories=['b', 'c', 'd'], ordered=False)

Supplying a CategoricalDtype will make the categories in each column consistent with the supplied dtype:

In [68]: from pandas.api.types import CategoricalDtype

In [69]: df = pd.DataFrame({'A': list('abca'), 'B': list('bccd')})

In [70]: cdt = CategoricalDtype(categories=list('abcd'), ordered=True)

In [71]: df = df.astype(cdt)

In [72]: df['A'].dtype
Out[72]: CategoricalDtype(categories=['a', 'b', 'c', 'd'], ordered=True)

In [73]: df['B'].dtype
Out[73]: CategoricalDtype(categories=['a', 'b', 'c', 'd'], ordered=True)

Other Enhancements

Backwards incompatible API changes

Dependencies have increased minimum versions

We have updated our minimum supported versions of dependencies (GH15184). If installed, we now require:

Package Minimum Version Required Issue
python-dateutil 2.5.0 X GH15184
openpyxl 2.4.0   GH15184
beautifulsoup4 4.2.1   GH20082
setuptools 24.2.0   GH20698

Instantiation from dicts preserves dict insertion order for python 3.6+

Until Python 3.6, dicts in Python had no formally defined ordering. For Python version 3.6 and later, dicts are ordered by insertion order, see PEP 468. Pandas will use the dict’s insertion order, when creating a Series or DataFrame from a dict and you’re using Python version 3.6 or higher. (GH19884)

Previous Behavior (and current behavior if on Python < 3.6):

pd.Series({'Income': 2000,
           'Expenses': -1500,
           'Taxes': -200,
           'Net result': 300})
Expenses     -1500
Income        2000
Net result     300
Taxes         -200
dtype: int64

Note the Series above is ordered alphabetically by the index values.

New Behavior (for Python >= 3.6):

In [74]: pd.Series({'Income': 2000,
   ....:            'Expenses': -1500,
   ....:            'Taxes': -200,
   ....:            'Net result': 300})
   ....: 
Out[74]: 
Income        2000
Expenses     -1500
Taxes         -200
Net result     300
dtype: int64

Notice that the Series is now ordered by insertion order. This new behavior is used for all relevant pandas types (Series, DataFrame, SparseSeries and SparseDataFrame).

If you wish to retain the old behavior while using Python >= 3.6, you can use .sort_index():

In [75]: pd.Series({'Income': 2000,
   ....:            'Expenses': -1500,
   ....:            'Taxes': -200,
   ....:            'Net result': 300}).sort_index()
   ....: 
Out[75]: 
Expenses     -1500
Income        2000
Net result     300
Taxes         -200
dtype: int64

Deprecate Panel

Panel was deprecated in the 0.20.x release, showing as a DeprecationWarning. Using Panel will now show a FutureWarning. The recommended way to represent 3-D data are with a MultiIndex on a DataFrame via the to_frame() or with the xarray package. Pandas provides a to_xarray() method to automate this conversion. For more details see Deprecate Panel documentation. (GH13563, GH18324).

In [76]: p = tm.makePanel()

In [77]: p
Out[77]: 
<class 'pandas.core.panel.Panel'>
Dimensions: 3 (items) x 3 (major_axis) x 4 (minor_axis)
Items axis: ItemA to ItemC
Major_axis axis: 2000-01-03 00:00:00 to 2000-01-05 00:00:00
Minor_axis axis: A to D

Convert to a MultiIndex DataFrame

In [78]: p.to_frame()
Out[78]: 
                     ItemA     ItemB     ItemC
major      minor                              
2000-01-03 A      1.474071 -0.964980 -1.197071
           B      0.781836  1.846883 -0.858447
           C      2.353925 -1.717693  0.384316
           D     -0.744471  0.901805  0.476720
2000-01-04 A     -0.064034 -0.845696 -1.066969
           B     -1.071357 -1.328865  0.306996
           C      0.583787  0.888782  1.574159
           D      0.758527  1.171216  0.473424
2000-01-05 A     -1.282782 -1.340896 -0.303421
           B      0.441153  1.682706 -0.028665
           C      0.221471  0.228440  1.588931
           D      1.729689  0.520260 -0.242861

Convert to an xarray DataArray

In [79]: p.to_xarray()
Out[79]: 
<xarray.DataArray (items: 3, major_axis: 3, minor_axis: 4)>
array([[[ 1.474071,  0.781836,  2.353925, -0.744471],
        [-0.064034, -1.071357,  0.583787,  0.758527],
        [-1.282782,  0.441153,  0.221471,  1.729689]],

       [[-0.96498 ,  1.846883, -1.717693,  0.901805],
        [-0.845696, -1.328865,  0.888782,  1.171216],
        [-1.340896,  1.682706,  0.22844 ,  0.52026 ]],

       [[-1.197071, -0.858447,  0.384316,  0.47672 ],
        [-1.066969,  0.306996,  1.574159,  0.473424],
        [-0.303421, -0.028665,  1.588931, -0.242861]]])
Coordinates:
  * items       (items) object 'ItemA' 'ItemB' 'ItemC'
  * major_axis  (major_axis) datetime64[ns] 2000-01-03 2000-01-04 2000-01-05
  * minor_axis  (minor_axis) object 'A' 'B' 'C' 'D'

pandas.core.common removals

The following error & warning messages are removed from pandas.core.common (GH13634, GH19769):

  • PerformanceWarning
  • UnsupportedFunctionCall
  • UnsortedIndexError
  • AbstractMethodError

These are available from import from pandas.errors (since 0.19.0).

Changes to make output of DataFrame.apply consistent

DataFrame.apply() was inconsistent when applying an arbitrary user-defined-function that returned a list-like with axis=1. Several bugs and inconsistencies are resolved. If the applied function returns a Series, then pandas will return a DataFrame; otherwise a Series will be returned, this includes the case where a list-like (e.g. tuple or list is returned) (GH16353, GH17437, GH17970, GH17348, GH17892, GH18573, GH17602, GH18775, GH18901, GH18919).

In [80]: df = pd.DataFrame(np.tile(np.arange(3), 6).reshape(6, -1) + 1, columns=['A', 'B', 'C'])

In [81]: df
Out[81]: 
   A  B  C
0  1  2  3
1  1  2  3
2  1  2  3
3  1  2  3
4  1  2  3
5  1  2  3

Previous Behavior: if the returned shape happened to match the length of original columns, this would return a DataFrame. If the return shape did not match, a Series with lists was returned.

In [3]: df.apply(lambda x: [1, 2, 3], axis=1)
Out[3]:
   A  B  C
0  1  2  3
1  1  2  3
2  1  2  3
3  1  2  3
4  1  2  3
5  1  2  3

In [4]: df.apply(lambda x: [1, 2], axis=1)
Out[4]:
0    [1, 2]
1    [1, 2]
2    [1, 2]
3    [1, 2]
4    [1, 2]
5    [1, 2]
dtype: object

New Behavior: When the applied function returns a list-like, this will now always return a Series.

In [82]: df.apply(lambda x: [1, 2, 3], axis=1)
Out[82]: 
0    [1, 2, 3]
1    [1, 2, 3]
2    [1, 2, 3]
3    [1, 2, 3]
4    [1, 2, 3]
5    [1, 2, 3]
dtype: object

In [83]: df.apply(lambda x: [1, 2], axis=1)
Out[83]: 
0    [1, 2]
1    [1, 2]
2    [1, 2]
3    [1, 2]
4    [1, 2]
5    [1, 2]
dtype: object

To have expanded columns, you can use result_type='expand'

In [84]: df.apply(lambda x: [1, 2, 3], axis=1, result_type='expand')
Out[84]: 
   0  1  2
0  1  2  3
1  1  2  3
2  1  2  3
3  1  2  3
4  1  2  3
5  1  2  3

To broadcast the result across the original columns (the old behaviour for list-likes of the correct length), you can use result_type='broadcast'. The shape must match the original columns.

In [85]: df.apply(lambda x: [1, 2, 3], axis=1, result_type='broadcast')
Out[85]: 
   A  B  C
0  1  2  3
1  1  2  3
2  1  2  3
3  1  2  3
4  1  2  3
5  1  2  3

Returning a Series allows one to control the exact return structure and column names:

In [86]: df.apply(lambda x: Series([1, 2, 3], index=['D', 'E', 'F']), axis=1)
Out[86]: 
   D  E  F
0  1  2  3
1  1  2  3
2  1  2  3
3  1  2  3
4  1  2  3
5  1  2  3

Concatenation will no longer sort

In a future version of pandas pandas.concat() will no longer sort the non-concatenation axis when it is not already aligned. The current behavior is the same as the previous (sorting), but now a warning is issued when sort is not specified and the non-concatenation axis is not aligned (GH4588).

In [87]: df1 = pd.DataFrame({"a": [1, 2], "b": [1, 2]}, columns=['b', 'a'])

In [88]: df2 = pd.DataFrame({"a": [4, 5]})

In [89]: pd.concat([df1, df2])
Out[89]: 
   a    b
0  1  1.0
1  2  2.0
0  4  NaN
1  5  NaN

To keep the previous behavior (sorting) and silence the warning, pass sort=True

In [90]: pd.concat([df1, df2], sort=True)
Out[90]: 
   a    b
0  1  1.0
1  2  2.0
0  4  NaN
1  5  NaN

To accept the future behavior (no sorting), pass sort=False

Note that this change also applies to DataFrame.append(), which has also received a sort keyword for controlling this behavior.

Build Changes

  • Building pandas for development now requires cython >= 0.24 (GH18613)
  • Building from source now explicitly requires setuptools in setup.py (GH18113)
  • Updated conda recipe to be in compliance with conda-build 3.0+ (GH18002)

Index Division By Zero Fills Correctly

Division operations on Index and subclasses will now fill division of positive numbers by zero with np.inf, division of negative numbers by zero with -np.inf and 0 / 0 with np.nan. This matches existing Series behavior. (GH19322, GH19347)

Previous Behavior:

In [6]: index = pd.Int64Index([-1, 0, 1])

In [7]: index / 0
Out[7]: Int64Index([0, 0, 0], dtype='int64')

# Previous behavior yielded different results depending on the type of zero in the divisor
In [8]: index / 0.0
Out[8]: Float64Index([-inf, nan, inf], dtype='float64')

In [9]: index = pd.UInt64Index([0, 1])

In [10]: index / np.array([0, 0], dtype=np.uint64)
Out[10]: UInt64Index([0, 0], dtype='uint64')

In [11]: pd.RangeIndex(1, 5) / 0
ZeroDivisionError: integer division or modulo by zero

Current Behavior:

In [91]: index = pd.Int64Index([-1, 0, 1])

# division by zero gives -infinity where negative, +infinity where positive, and NaN for 0 / 0
In [92]: index / 0
Out[92]: Float64Index([-inf, nan, inf], dtype='float64')

# The result of division by zero should not depend on whether the zero is int or float
In [93]: index / 0.0
Out[93]: Float64Index([-inf, nan, inf], dtype='float64')

In [94]: index = pd.UInt64Index([0, 1])

In [95]: index / np.array([0, 0], dtype=np.uint64)
Out[95]: Float64Index([nan, inf], dtype='float64')

In [96]: pd.RangeIndex(1, 5) / 0
Out[96]: Float64Index([inf, inf, inf, inf], dtype='float64')

Extraction of matching patterns from strings

By default, extracting matching patterns from strings with str.extract() used to return a Series if a single group was being extracted (a DataFrame if more than one group was extracted). As of Pandas 0.23.0 str.extract() always returns a DataFrame, unless expand is set to False. Finallay, None was an accepted value for the expand parameter (which was equivalent to False), but now raises a ValueError. (GH11386)

Previous Behavior:

In [1]: s = pd.Series(['number 10', '12 eggs'])

In [2]: extracted = s.str.extract('.*(\d\d).*')

In [3]: extracted
Out [3]:
0    10
1    12
dtype: object

In [4]: type(extracted)
Out [4]:
pandas.core.series.Series

New Behavior:

In [97]: s = pd.Series(['number 10', '12 eggs'])

In [98]: extracted = s.str.extract('.*(\d\d).*')

In [99]: extracted
Out[99]: 
    0
0  10
1  12

In [100]: type(extracted)
Out[100]: pandas.core.frame.DataFrame

To restore previous behavior, simply set expand to False:

In [101]: s = pd.Series(['number 10', '12 eggs'])

In [102]: extracted = s.str.extract('.*(\d\d).*', expand=False)

In [103]: extracted
Out[103]: 
0    10
1    12
dtype: object

In [104]: type(extracted)
Out[104]: pandas.core.series.Series

Default value for the ordered parameter of CategoricalDtype

The default value of the ordered parameter for CategoricalDtype has changed from False to None to allow updating of categories without impacting ordered. Behavior should remain consistent for downstream objects, such as Categorical (GH18790)

In previous versions, the default value for the ordered parameter was False. This could potentially lead to the ordered parameter unintentionally being changed from True to False when users attempt to update categories if ordered is not explicitly specified, as it would silently default to False. The new behavior for ordered=None is to retain the existing value of ordered.

New Behavior:

In [105]: from pandas.api.types import CategoricalDtype

In [106]: cat = pd.Categorical(list('abcaba'), ordered=True, categories=list('cba'))

In [107]: cat
Out[107]: 
[a, b, c, a, b, a]
Categories (3, object): [c < b < a]

In [108]: cdt = CategoricalDtype(categories=list('cbad'))

In [109]: cat.astype(cdt)
Out[109]: 
[a, b, c, a, b, a]
Categories (4, object): [c < b < a < d]

Notice in the example above that the converted Categorical has retained ordered=True. Had the default value for ordered remained as False, the converted Categorical would have become unordered, despite ordered=False never being explicitly specified. To change the value of ordered, explicitly pass it to the new dtype, e.g. CategoricalDtype(categories=list('cbad'), ordered=False).

Note that the unintenional conversion of ordered discussed above did not arise in previous versions due to separate bugs that prevented astype from doing any type of category to category conversion (GH10696, GH18593). These bugs have been fixed in this release, and motivated changing the default value of ordered.

Better pretty-printing of DataFrames in a terminal

Previously, the default value for the maximum number of columns was pd.options.display.max_columns=20. This meant that relatively wide data frames would not fit within the terminal width, and pandas would introduce line breaks to display these 20 columns. This resulted in an output that was relatively difficult to read:

_images/print_df_old.png

If Python runs in a terminal, the maximum number of columns is now determined automatically so that the printed data frame fits within the current terminal width (pd.options.display.max_columns=0) (GH17023). If Python runs as a Jupyter kernel (such as the Jupyter QtConsole or a Jupyter notebook, as well as in many IDEs), this value cannot be inferred automatically and is thus set to 20 as in previous versions. In a terminal, this results in a much nicer output:

_images/print_df_new.png

Note that if you don’t like the new default, you can always set this option yourself. To revert to the old setting, you can run this line:

pd.options.display.max_columns = 20

Datetimelike API Changes

  • The default Timedelta constructor now accepts an ISO 8601 Duration string as an argument (GH19040)
  • Subtracting NaT from a Series with dtype='datetime64[ns]' returns a Series with dtype='timedelta64[ns]' instead of dtype='datetime64[ns]' (GH18808)
  • Addition or subtraction of NaT from TimedeltaIndex will return TimedeltaIndex instead of DatetimeIndex (GH19124)
  • DatetimeIndex.shift() and TimedeltaIndex.shift() will now raise NullFrequencyError (which subclasses ValueError, which was raised in older versions) when the index object frequency is None (GH19147)
  • Addition and subtraction of NaN from a Series with dtype='timedelta64[ns]' will raise a TypeError instead of treating the NaN as NaT (GH19274)
  • NaT division with datetime.timedelta will now return NaN instead of raising (GH17876)
  • Operations between a Series with dtype dtype='datetime64[ns]' and a PeriodIndex will correctly raises TypeError (GH18850)
  • Subtraction of Series with timezone-aware dtype='datetime64[ns]' with mis-matched timezones will raise TypeError instead of ValueError (GH18817)
  • Timestamp will no longer silently ignore unused or invalid tz or tzinfo keyword arguments (GH17690)
  • Timestamp will no longer silently ignore invalid freq arguments (GH5168)
  • CacheableOffset and WeekDay are no longer available in the pandas.tseries.offsets module (GH17830)
  • pandas.tseries.frequencies.get_freq_group() and pandas.tseries.frequencies.DAYS are removed from the public API (GH18034)
  • Series.truncate() and DataFrame.truncate() will raise a ValueError if the index is not sorted instead of an unhelpful KeyError (GH17935)
  • Series.first and DataFrame.first will now raise a TypeError rather than NotImplementedError when index is not a DatetimeIndex (GH20725).
  • Series.last and DataFrame.last will now raise a TypeError rather than NotImplementedError when index is not a DatetimeIndex (GH20725).
  • Restricted DateOffset keyword arguments. Previously, DateOffset subclasses allowed arbitrary keyword arguments which could lead to unexpected behavior. Now, only valid arguments will be accepted. (GH17176, GH18226).
  • pandas.merge() provides a more informative error message when trying to merge on timezone-aware and timezone-naive columns (GH15800)
  • For DatetimeIndex and TimedeltaIndex with freq=None, addition or subtraction of integer-dtyped array or Index will raise NullFrequencyError instead of TypeError (GH19895)
  • Timestamp constructor now accepts a nanosecond keyword or positional argument (GH18898)
  • DatetimeIndex will now raise an AttributeError when the tz attribute is set after instantiation (GH3746)
  • DatetimeIndex with a pytz timezone will now return a consistent pytz timezone (GH18595)

Other API Changes

  • Series.astype() and Index.astype() with an incompatible dtype will now raise a TypeError rather than a ValueError (GH18231)
  • Series construction with an object dtyped tz-aware datetime and dtype=object specified, will now return an object dtyped Series, previously this would infer the datetime dtype (GH18231)
  • A Series of dtype=category constructed from an empty dict will now have categories of dtype=object rather than dtype=float64, consistently with the case in which an empty list is passed (GH18515)
  • All-NaN levels in a MultiIndex are now assigned float rather than object dtype, promoting consistency with Index (GH17929).
  • Levels names of a MultiIndex (when not None) are now required to be unique: trying to create a MultiIndex with repeated names will raise a ValueError (GH18872)
  • Both construction and renaming of Index/MultiIndex with non-hashable name/names will now raise TypeError (GH20527)
  • Index.map() can now accept Series and dictionary input objects (GH12756, GH18482, GH18509).
  • DataFrame.unstack() will now default to filling with np.nan for object columns. (GH12815)
  • IntervalIndex constructor will raise if the closed parameter conflicts with how the input data is inferred to be closed (GH18421)
  • Inserting missing values into indexes will work for all types of indexes and automatically insert the correct type of missing value (NaN, NaT, etc.) regardless of the type passed in (GH18295)
  • When created with duplicate labels, MultiIndex now raises a ValueError. (GH17464)
  • Series.fillna() now raises a TypeError instead of a ValueError when passed a list, tuple or DataFrame as a value (GH18293)
  • pandas.DataFrame.merge() no longer casts a float column to object when merging on int and float columns (GH16572)
  • pandas.merge() now raises a ValueError when trying to merge on incompatible data types (GH9780)
  • The default NA value for UInt64Index has changed from 0 to NaN, which impacts methods that mask with NA, such as UInt64Index.where() (GH18398)
  • Refactored setup.py to use find_packages instead of explicitly listing out all subpackages (GH18535)
  • Rearranged the order of keyword arguments in read_excel() to align with read_csv() (GH16672)
  • wide_to_long() previously kept numeric-like suffixes as object dtype. Now they are cast to numeric if possible (GH17627)
  • In read_excel(), the comment argument is now exposed as a named parameter (GH18735)
  • Rearranged the order of keyword arguments in read_excel() to align with read_csv() (GH16672)
  • The options html.border and mode.use_inf_as_null were deprecated in prior versions, these will now show FutureWarning rather than a DeprecationWarning (GH19003)
  • IntervalIndex and IntervalDtype no longer support categorical, object, and string subtypes (GH19016)
  • IntervalDtype now returns True when compared against 'interval' regardless of subtype, and IntervalDtype.name now returns 'interval' regardless of subtype (GH18980)
  • KeyError now raises instead of ValueError in drop(), drop(), drop(), drop() when dropping a non-existent element in an axis with duplicates (GH19186)
  • Series.to_csv() now accepts a compression argument that works in the same way as the compression argument in DataFrame.to_csv() (GH18958)
  • Set operations (union, difference…) on IntervalIndex with incompatible index types will now raise a TypeError rather than a ValueError (GH19329)
  • DateOffset objects render more simply, e.g. <DateOffset: days=1> instead of <DateOffset: kwds={'days': 1}> (GH19403)
  • Categorical.fillna now validates its value and method keyword arguments. It now raises when both or none are specified, matching the behavior of Series.fillna() (GH19682)
  • pd.to_datetime('today') now returns a datetime, consistent with pd.Timestamp('today'); previously pd.to_datetime('today') returned a .normalized() datetime (GH19935)
  • Series.str.replace() now takes an optional regex keyword which, when set to False, uses literal string replacement rather than regex replacement (GH16808)
  • DatetimeIndex.strftime() and PeriodIndex.strftime() now return an Index instead of a numpy array to be consistent with similar accessors (GH20127)
  • Constructing a Series from a list of length 1 no longer broadcasts this list when a longer index is specified (GH19714, GH20391).
  • DataFrame.to_dict() with orient='index' no longer casts int columns to float for a DataFrame with only int and float columns (GH18580)
  • A user-defined-function that is passed to Series.rolling().aggregate(), DataFrame.rolling().aggregate(), or its expanding cousins, will now always be passed a Series, rather than a np.array; .apply() only has the raw keyword, see here. This is consistent with the signatures of .aggregate() across pandas (GH20584)
  • Rolling and Expanding types raise NotImplementedError upon iteration (GH11704).

Deprecations

  • Series.from_array and SparseSeries.from_array are deprecated. Use the normal constructor Series(..) and SparseSeries(..) instead (GH18213).
  • DataFrame.as_matrix is deprecated. Use DataFrame.values instead (GH18458).
  • Series.asobject, DatetimeIndex.asobject, PeriodIndex.asobject and TimeDeltaIndex.asobject have been deprecated. Use .astype(object) instead (GH18572)
  • Grouping by a tuple of keys now emits a FutureWarning and is deprecated. In the future, a tuple passed to 'by' will always refer to a single key that is the actual tuple, instead of treating the tuple as multiple keys. To retain the previous behavior, use a list instead of a tuple (GH18314)
  • Series.valid is deprecated. Use Series.dropna() instead (GH18800).
  • read_excel() has deprecated the skip_footer parameter. Use skipfooter instead (GH18836)
  • ExcelFile.parse() has deprecated sheetname in favor of sheet_name for consistency with read_excel() (GH20920).
  • The is_copy attribute is deprecated and will be removed in a future version (GH18801).
  • IntervalIndex.from_intervals is deprecated in favor of the IntervalIndex constructor (GH19263)
  • DataFrame.from_items is deprecated. Use DataFrame.from_dict() instead, or DataFrame.from_dict(OrderedDict()) if you wish to preserve the key order (GH17320, GH17312)
  • Indexing a MultiIndex or a FloatIndex with a list containing some missing keys will now show a FutureWarning, which is consistent with other types of indexes (GH17758).
  • The broadcast parameter of .apply() is deprecated in favor of result_type='broadcast' (GH18577)
  • The reduce parameter of .apply() is deprecated in favor of result_type='reduce' (GH18577)
  • The order parameter of factorize() is deprecated and will be removed in a future release (GH19727)
  • Timestamp.weekday_name, DatetimeIndex.weekday_name, and Series.dt.weekday_name are deprecated in favor of Timestamp.day_name(), DatetimeIndex.day_name(), and Series.dt.day_name() (GH12806)
  • pandas.tseries.plotting.tsplot is deprecated. Use Series.plot() instead (GH18627)
  • Index.summary() is deprecated and will be removed in a future version (GH18217)
  • NDFrame.get_ftype_counts() is deprecated and will be removed in a future version (GH18243)
  • The convert_datetime64 parameter in DataFrame.to_records() has been deprecated and will be removed in a future version. The NumPy bug motivating this parameter has been resolved. The default value for this parameter has also changed from True to None (GH18160).
  • Series.rolling().apply(), DataFrame.rolling().apply(), Series.expanding().apply(), and DataFrame.expanding().apply() have deprecated passing an np.array by default. One will need to pass the new raw parameter to be explicit about what is passed (GH20584)
  • The data, base, strides, flags and itemsize properties of the Series and Index classes have been deprecated and will be removed in a future version (GH20419).
  • DatetimeIndex.offset is deprecated. Use DatetimeIndex.freq instead (GH20716)
  • Floor division between an integer ndarray and a Timedelta is deprecated. Divide by Timedelta.value instead (GH19761)
  • Setting PeriodIndex.freq (which was not guaranteed to work correctly) is deprecated. Use PeriodIndex.asfreq() instead (GH20678)
  • Index.get_duplicates() is deprecated and will be removed in a future version (GH20239)
  • The previous default behavior of negative indices in Categorical.take is deprecated. In a future version it will change from meaning missing values to meaning positional indices from the right. The future behavior is consistent with Series.take() (GH20664).
  • Passing multiple axes to the axis parameter in DataFrame.dropna() has been deprecated and will be removed in a future version (GH20987)

Removal of prior version deprecations/changes

  • Warnings against the obsolete usage Categorical(codes, categories), which were emitted for instance when the first two arguments to Categorical() had different dtypes, and recommended the use of Categorical.from_codes, have now been removed (GH8074)
  • The levels and labels attributes of a MultiIndex can no longer be set directly (GH4039).
  • pd.tseries.util.pivot_annual has been removed (deprecated since v0.19). Use pivot_table instead (GH18370)
  • pd.tseries.util.isleapyear has been removed (deprecated since v0.19). Use .is_leap_year property in Datetime-likes instead (GH18370)
  • pd.ordered_merge has been removed (deprecated since v0.19). Use pd.merge_ordered instead (GH18459)
  • The SparseList class has been removed (GH14007)
  • The pandas.io.wb and pandas.io.data stub modules have been removed (GH13735)
  • Categorical.from_array has been removed (GH13854)
  • The freq and how parameters have been removed from the rolling/expanding/ewm methods of DataFrame and Series (deprecated since v0.18). Instead, resample before calling the methods. (GH18601 & GH18668)
  • DatetimeIndex.to_datetime, Timestamp.to_datetime, PeriodIndex.to_datetime, and Index.to_datetime have been removed (GH8254, GH14096, GH14113)
  • read_csv() has dropped the skip_footer parameter (GH13386)
  • read_csv() has dropped the as_recarray parameter (GH13373)
  • read_csv() has dropped the buffer_lines parameter (GH13360)
  • read_csv() has dropped the compact_ints and use_unsigned parameters (GH13323)
  • The Timestamp class has dropped the offset attribute in favor of freq (GH13593)
  • The Series, Categorical, and Index classes have dropped the reshape method (GH13012)
  • pandas.tseries.frequencies.get_standard_freq has been removed in favor of pandas.tseries.frequencies.to_offset(freq).rule_code (GH13874)
  • The freqstr keyword has been removed from pandas.tseries.frequencies.to_offset in favor of freq (GH13874)
  • The Panel4D and PanelND classes have been removed (GH13776)
  • The Panel class has dropped the to_long and toLong methods (GH19077)
  • The options display.line_with and display.height are removed in favor of display.width and display.max_rows respectively (GH4391, GH19107)
  • The labels attribute of the Categorical class has been removed in favor of Categorical.codes (GH7768)
  • The flavor parameter have been removed from func:to_sql method (GH13611)
  • The modules pandas.tools.hashing and pandas.util.hashing have been removed (GH16223)
  • The top-level functions pd.rolling_*, pd.expanding_* and pd.ewm* have been removed (Deprecated since v0.18). Instead, use the DataFrame/Series methods rolling, expanding and ewm (GH18723)
  • Imports from pandas.core.common for functions such as is_datetime64_dtype are now removed. These are located in pandas.api.types. (GH13634, GH19769)
  • The infer_dst keyword in Series.tz_localize(), DatetimeIndex.tz_localize() and DatetimeIndex have been removed. infer_dst=True is equivalent to ambiguous='infer', and infer_dst=False to ambiguous='raise' (GH7963).
  • When .resample() was changed from an eager to a lazy operation, like .groupby() in v0.18.0, we put in place compatibility (with a FutureWarning), so operations would continue to work. This is now fully removed, so a Resampler will no longer forward compat operations (GH20554)
  • Remove long deprecated axis=None parameter from .replace() (GH20271)

Performance Improvements

Documentation Changes

Thanks to all of the contributors who participated in the Pandas Documentation Sprint, which took place on March 10th. We had about 500 participants from over 30 locations across the world. You should notice that many of the API docstrings have greatly improved.

There were too many simultaneous contributions to include a release note for each improvement, but this GitHub search should give you an idea of how many docstrings were improved.

Special thanks to Marc Garcia for organizing the sprint. For more information, read the NumFOCUS blogpost recapping the sprint.

  • Changed spelling of “numpy” to “NumPy”, and “python” to “Python”. (GH19017)
  • Consistency when introducing code samples, using either colon or period. Rewrote some sentences for greater clarity, added more dynamic references to functions, methods and classes. (GH18941, GH18948, GH18973, GH19017)
  • Added a reference to DataFrame.assign() in the concatenate section of the merging documentation (GH18665)

Bug Fixes

Categorical

Warning

A class of bugs were introduced in pandas 0.21 with CategoricalDtype that affects the correctness of operations like merge, concat, and indexing when comparing multiple unordered Categorical arrays that have the same categories, but in a different order. We highly recommend upgrading or manually aligning your categories before doing these operations.

  • Bug in Categorical.equals returning the wrong result when comparing two unordered Categorical arrays with the same categories, but in a different order (GH16603)
  • Bug in pandas.api.types.union_categoricals() returning the wrong result when for unordered categoricals with the categories in a different order. This affected pandas.concat() with Categorical data (GH19096).
  • Bug in pandas.merge() returning the wrong result when joining on an unordered Categorical that had the same categories but in a different order (GH19551)
  • Bug in CategoricalIndex.get_indexer() returning the wrong result when target was an unordered Categorical that had the same categories as self but in a different order (GH19551)
  • Bug in Index.astype() with a categorical dtype where the resultant index is not converted to a CategoricalIndex for all types of index (GH18630)
  • Bug in Series.astype() and Categorical.astype() where an existing categorical data does not get updated (GH10696, GH18593)
  • Bug in Series.str.split() with expand=True incorrectly raising an IndexError on empty strings (GH20002).
  • Bug in Index constructor with dtype=CategoricalDtype(...) where categories and ordered are not maintained (GH19032)
  • Bug in Series constructor with scalar and dtype=CategoricalDtype(...) where categories and ordered are not maintained (GH19565)
  • Bug in Categorical.__iter__ not converting to Python types (GH19909)
  • Bug in pandas.factorize() returning the unique codes for the uniques. This now returns a Categorical with the same dtype as the input (GH19721)
  • Bug in pandas.factorize() including an item for missing values in the uniques return value (GH19721)
  • Bug in Series.take() with categorical data interpreting -1 in indices as missing value markers, rather than the last element of the Series (GH20664)

Datetimelike

  • Bug in Series.__sub__() subtracting a non-nanosecond np.datetime64 object from a Series gave incorrect results (GH7996)
  • Bug in DatetimeIndex, TimedeltaIndex addition and subtraction of zero-dimensional integer arrays gave incorrect results (GH19012)
  • Bug in DatetimeIndex and TimedeltaIndex where adding or subtracting an array-like of DateOffset objects either raised (np.array, pd.Index) or broadcast incorrectly (pd.Series) (GH18849)
  • Bug in Series.__add__() adding Series with dtype timedelta64[ns] to a timezone-aware DatetimeIndex incorrectly dropped timezone information (GH13905)
  • Adding a Period object to a datetime or Timestamp object will now correctly raise a TypeError (GH17983)
  • Bug in Timestamp where comparison with an array of Timestamp objects would result in a RecursionError (GH15183)
  • Bug in Series floor-division where operating on a scalar timedelta raises an exception (GH18846)
  • Bug in DatetimeIndex where the repr was not showing high-precision time values at the end of a day (e.g., 23:59:59.999999999) (GH19030)
  • Bug in .astype() to non-ns timedelta units would hold the incorrect dtype (GH19176, GH19223, GH12425)
  • Bug in subtracting Series from NaT incorrectly returning NaT (GH19158)
  • Bug in Series.truncate() which raises TypeError with a monotonic PeriodIndex (GH17717)
  • Bug in pct_change() using periods and freq returned different length outputs (GH7292)
  • Bug in comparison of DatetimeIndex against None or datetime.date objects raising TypeError for == and != comparisons instead of all-False and all-True, respectively (GH19301)
  • Bug in Timestamp and to_datetime() where a string representing a barely out-of-bounds timestamp would be incorrectly rounded down instead of raising OutOfBoundsDatetime (GH19382)
  • Bug in Timestamp.floor() DatetimeIndex.floor() where time stamps far in the future and past were not rounded correctly (GH19206)
  • Bug in to_datetime() where passing an out-of-bounds datetime with errors='coerce' and utc=True would raise OutOfBoundsDatetime instead of parsing to NaT (GH19612)
  • Bug in DatetimeIndex and TimedeltaIndex addition and subtraction where name of the returned object was not always set consistently. (GH19744)
  • Bug in DatetimeIndex and TimedeltaIndex addition and subtraction where operations with numpy arrays raised TypeError (GH19847)
  • Bug in DatetimeIndex and TimedeltaIndex where setting the freq attribute was not fully supported (GH20678)

Timedelta

  • Bug in Timedelta.__mul__() where multiplying by NaT returned NaT instead of raising a TypeError (GH19819)
  • Bug in Series with dtype='timedelta64[ns]' where addition or subtraction of TimedeltaIndex had results cast to dtype='int64' (GH17250)
  • Bug in Series with dtype='timedelta64[ns]' where addition or subtraction of TimedeltaIndex could return a Series with an incorrect name (GH19043)
  • Bug in Timedelta.__floordiv__() and Timedelta.__rfloordiv__() dividing by many incompatible numpy objects was incorrectly allowed (GH18846)
  • Bug where dividing a scalar timedelta-like object with TimedeltaIndex performed the reciprocal operation (GH19125)
  • Bug in TimedeltaIndex where division by a Series would return a TimedeltaIndex instead of a Series (GH19042)
  • Bug in Timedelta.__add__(), Timedelta.__sub__() where adding or subtracting a np.timedelta64 object would return another np.timedelta64 instead of a Timedelta (GH19738)
  • Bug in Timedelta.__floordiv__(), Timedelta.__rfloordiv__() where operating with a Tick object would raise a TypeError instead of returning a numeric value (GH19738)
  • Bug in Period.asfreq() where periods near datetime(1, 1, 1) could be converted incorrectly (GH19643, GH19834)
  • Bug in Timedelta.total_seconds() causing precision errors, for example Timedelta('30S').total_seconds()==30.000000000000004 (GH19458)
  • Bug in Timedelta.__rmod__() where operating with a numpy.timedelta64 returned a timedelta64 object instead of a Timedelta (GH19820)
  • Multiplication of TimedeltaIndex by TimedeltaIndex will now raise TypeError instead of raising ValueError in cases of length mis-match (GH19333)
  • Bug in indexing a TimedeltaIndex with a np.timedelta64 object which was raising a TypeError (GH20393)

Timezones

  • Bug in creating a Series from an array that contains both tz-naive and tz-aware values will result in a Series whose dtype is tz-aware instead of object (GH16406)
  • Bug in comparison of timezone-aware DatetimeIndex against NaT incorrectly raising TypeError (GH19276)
  • Bug in DatetimeIndex.astype() when converting between timezone aware dtypes, and converting from timezone aware to naive (GH18951)
  • Bug in comparing DatetimeIndex, which failed to raise TypeError when attempting to compare timezone-aware and timezone-naive datetimelike objects (GH18162)
  • Bug in localization of a naive, datetime string in a Series constructor with a datetime64[ns, tz] dtype (GH174151)
  • Timestamp.replace() will now handle Daylight Savings transitions gracefully (GH18319)
  • Bug in tz-aware DatetimeIndex where addition/subtraction with a TimedeltaIndex or array with dtype='timedelta64[ns]' was incorrect (GH17558)
  • Bug in DatetimeIndex.insert() where inserting NaT into a timezone-aware index incorrectly raised (GH16357)
  • Bug in DataFrame constructor, where tz-aware Datetimeindex and a given column name will result in an empty DataFrame (GH19157)
  • Bug in Timestamp.tz_localize() where localizing a timestamp near the minimum or maximum valid values could overflow and return a timestamp with an incorrect nanosecond value (GH12677)
  • Bug when iterating over DatetimeIndex that was localized with fixed timezone offset that rounded nanosecond precision to microseconds (GH19603)
  • Bug in DataFrame.diff() that raised an IndexError with tz-aware values (GH18578)
  • Bug in melt() that converted tz-aware dtypes to tz-naive (GH15785)
  • Bug in Dataframe.count() that raised an ValueError, if Dataframe.dropna() was called for a single column with timezone-aware values. (GH13407)

Offsets

  • Bug in WeekOfMonth and Week where addition and subtraction did not roll correctly (GH18510, GH18672, GH18864)
  • Bug in WeekOfMonth and LastWeekOfMonth where default keyword arguments for constructor raised ValueError (GH19142)
  • Bug in FY5253Quarter, LastWeekOfMonth where rollback and rollforward behavior was inconsistent with addition and subtraction behavior (GH18854)
  • Bug in FY5253 where datetime addition and subtraction incremented incorrectly for dates on the year-end but not normalized to midnight (GH18854)
  • Bug in FY5253 where date offsets could incorrectly raise an AssertionError in arithmetic operatons (GH14774)

Numeric

  • Bug in Series constructor with an int or float list where specifying dtype=str, dtype='str' or dtype='U' failed to convert the data elements to strings (GH16605)
  • Bug in Index multiplication and division methods where operating with a Series would return an Index object instead of a Series object (GH19042)
  • Bug in the DataFrame constructor in which data containing very large positive or very large negative numbers was causing OverflowError (GH18584)
  • Bug in Index constructor with dtype='uint64' where int-like floats were not coerced to UInt64Index (GH18400)
  • Bug in DataFrame flex arithmetic (e.g. df.add(other, fill_value=foo)) with a fill_value other than None failed to raise NotImplementedError in corner cases where either the frame or other has length zero (GH19522)
  • Multiplication and division of numeric-dtyped Index objects with timedelta-like scalars returns TimedeltaIndex instead of raising TypeError (GH19333)
  • Bug where NaN was returned instead of 0 by Series.pct_change() and DataFrame.pct_change() when fill_method is not None (GH19873)

Strings

  • Bug in Series.str.get() with a dictionary in the values and the index not in the keys, raising KeyError (GH20671)

Indexing

  • Bug in Index construction from list of mixed type tuples (GH18505)
  • Bug in Index.drop() when passing a list of both tuples and non-tuples (GH18304)
  • Bug in DataFrame.drop(), Panel.drop(), Series.drop(), Index.drop() where no KeyError is raised when dropping a non-existent element from an axis that contains duplicates (GH19186)
  • Bug in indexing a datetimelike Index that raised ValueError instead of IndexError (GH18386).
  • Index.to_series() now accepts index and name kwargs (GH18699)
  • DatetimeIndex.to_series() now accepts index and name kwargs (GH18699)
  • Bug in indexing non-scalar value from Series having non-unique Index will return value flattened (GH17610)
  • Bug in indexing with iterator containing only missing keys, which raised no error (GH20748)
  • Fixed inconsistency in .ix between list and scalar keys when the index has integer dtype and does not include the desired keys (GH20753)
  • Bug in __setitem__ when indexing a DataFrame with a 2-d boolean ndarray (GH18582)
  • Bug in str.extractall when there were no matches empty Index was returned instead of appropriate MultiIndex (GH19034)
  • Bug in IntervalIndex where empty and purely NA data was constructed inconsistently depending on the construction method (GH18421)
  • Bug in IntervalIndex.symmetric_difference() where the symmetric difference with a non-IntervalIndex did not raise (GH18475)
  • Bug in IntervalIndex where set operations that returned an empty IntervalIndex had the wrong dtype (GH19101)
  • Bug in DataFrame.drop_duplicates() where no KeyError is raised when passing in columns that don’t exist on the DataFrame (GH19726)
  • Bug in Index subclasses constructors that ignore unexpected keyword arguments (GH19348)
  • Bug in Index.difference() when taking difference of an Index with itself (GH20040)
  • Bug in DataFrame.first_valid_index() and DataFrame.last_valid_index() in presence of entire rows of NaNs in the middle of values (GH20499).
  • Bug in IntervalIndex where some indexing operations were not supported for overlapping or non-monotonic uint64 data (GH20636)
  • Bug in Series.is_unique where extraneous output in stderr is shown if Series contains objects with __ne__ defined (GH20661)
  • Bug in .loc assignment with a single-element list-like incorrectly assigns as a list (GH19474)
  • Bug in partial string indexing on a Series/DataFrame with a monotonic decreasing DatetimeIndex (GH19362)
  • Bug in performing in-place operations on a DataFrame with a duplicate Index (GH17105)
  • Bug in IntervalIndex.get_loc() and IntervalIndex.get_indexer() when used with an IntervalIndex containing a single interval (GH17284, GH20921)
  • Bug in .loc with a uint64 indexer (GH20722)

MultiIndex

I/O

Plotting

  • Better error message when attempting to plot but matplotlib is not installed (GH19810).
  • DataFrame.plot() now raises a ValueError when the x or y argument is improperly formed (GH18671)
  • Bug in DataFrame.plot() when x and y arguments given as positions caused incorrect referenced columns for line, bar and area plots (GH20056)
  • Bug in formatting tick labels with datetime.time() and fractional seconds (GH18478).
  • Series.plot.kde() has exposed the args ind and bw_method in the docstring (GH18461). The argument ind may now also be an integer (number of sample points).
  • DataFrame.plot() now supports multiple columns to the y argument (GH19699)

Groupby/Resample/Rolling

Sparse

  • Bug in which creating a SparseDataFrame from a dense Series or an unsupported type raised an uncontrolled exception (GH19374)
  • Bug in SparseDataFrame.to_csv causing exception (GH19384)
  • Bug in SparseSeries.memory_usage which caused segfault by accessing non sparse elements (GH19368)
  • Bug in constructing a SparseArray: if data is a scalar and index is defined it will coerce to float64 regardless of scalar’s dtype. (GH19163)

Reshaping

  • Bug in DataFrame.merge() where referencing a CategoricalIndex by name, where the by kwarg would KeyError (GH20777)
  • Bug in DataFrame.stack() which fails trying to sort mixed type levels under Python 3 (GH18310)
  • Bug in DataFrame.unstack() which casts int to float if columns is a MultiIndex with unused levels (GH17845)
  • Bug in DataFrame.unstack() which raises an error if index is a MultiIndex with unused labels on the unstacked level (GH18562)
  • Fixed construction of a Series from a dict containing NaN as key (GH18480)
  • Fixed construction of a DataFrame from a dict containing NaN as key (GH18455)
  • Disabled construction of a Series where len(index) > len(data) = 1, which previously would broadcast the data item, and now raises a ValueError (GH18819)
  • Suppressed error in the construction of a DataFrame from a dict containing scalar values when the corresponding keys are not included in the passed index (GH18600)
  • Fixed (changed from object to float64) dtype of DataFrame initialized with axes, no data, and dtype=int (GH19646)
  • Bug in Series.rank() where Series containing NaT modifies the Series inplace (GH18521)
  • Bug in cut() which fails when using readonly arrays (GH18773)
  • Bug in DataFrame.pivot_table() which fails when the aggfunc arg is of type string. The behavior is now consistent with other methods like agg and apply (GH18713)
  • Bug in DataFrame.merge() in which merging using Index objects as vectors raised an Exception (GH19038)
  • Bug in DataFrame.stack(), DataFrame.unstack(), Series.unstack() which were not returning subclasses (GH15563)
  • Bug in timezone comparisons, manifesting as a conversion of the index to UTC in .concat() (GH18523)
  • Bug in concat() when concatting sparse and dense series it returns only a SparseDataFrame. Should be a DataFrame. (GH18914, GH18686, and GH16874)
  • Improved error message for DataFrame.merge() when there is no common merge key (GH19427)
  • Bug in DataFrame.join() which does an outer instead of a left join when being called with multiple DataFrames and some have non-unique indices (GH19624)
  • Series.rename() now accepts axis as a kwarg (GH18589)
  • Bug in rename() where an Index of same-length tuples was converted to a MultiIndex (GH19497)
  • Comparisons between Series and Index would return a Series with an incorrect name, ignoring the Index’s name attribute (GH19582)
  • Bug in qcut() where datetime and timedelta data with NaT present raised a ValueError (GH19768)
  • Bug in DataFrame.iterrows(), which would infers strings not compliant to ISO8601 to datetimes (GH19671)
  • Bug in Series constructor with Categorical where a ValueError is not raised when an index of different length is given (GH19342)
  • Bug in DataFrame.astype() where column metadata is lost when converting to categorical or a dictionary of dtypes (GH19920)
  • Bug in cut() and qcut() where timezone information was dropped (GH19872)
  • Bug in Series constructor with a dtype=str, previously raised in some cases (GH19853)
  • Bug in get_dummies(), and select_dtypes(), where duplicate column names caused incorrect behavior (GH20848)
  • Bug in isna(), which cannot handle ambiguous typed lists (GH20675)
  • Bug in concat() which raises an error when concatenating TZ-aware dataframes and all-NaT dataframes (GH12396)
  • Bug in concat() which raises an error when concatenating empty TZ-aware series (GH18447)

Other

v0.22.0 (December 29, 2017)

This is a major release from 0.21.1 and includes a single, API-breaking change. We recommend that all users upgrade to this version after carefully reading the release note (singular!).

Backwards incompatible API changes

Pandas 0.22.0 changes the handling of empty and all-NA sums and products. The summary is that

  • The sum of an empty or all-NA Series is now 0
  • The product of an empty or all-NA Series is now 1
  • We’ve added a min_count parameter to .sum() and .prod() controlling the minimum number of valid values for the result to be valid. If fewer than min_count non-NA values are present, the result is NA. The default is 0. To return NaN, the 0.21 behavior, use min_count=1.

Some background: In pandas 0.21, we fixed a long-standing inconsistency in the return value of all-NA series depending on whether or not bottleneck was installed. See Sum/Prod of all-NaN or empty Series/DataFrames is now consistently NaN. At the same time, we changed the sum and prod of an empty Series to also be NaN.

Based on feedback, we’ve partially reverted those changes.

Arithmetic Operations

The default sum for empty or all-NA Series is now 0.

pandas 0.21.x

In [1]: pd.Series([]).sum()
Out[1]: nan

In [2]: pd.Series([np.nan]).sum()
Out[2]: nan

pandas 0.22.0

In [1]: pd.Series([]).sum()
Out[1]: 0.0

In [2]: pd.Series([np.nan]).sum()
Out[2]: 0.0

The default behavior is the same as pandas 0.20.3 with bottleneck installed. It also matches the behavior of NumPy’s np.nansum on empty and all-NA arrays.

To have the sum of an empty series return NaN (the default behavior of pandas 0.20.3 without bottleneck, or pandas 0.21.x), use the min_count keyword.

In [3]: pd.Series([]).sum(min_count=1)
Out[3]: nan

Thanks to the skipna parameter, the .sum on an all-NA series is conceptually the same as the .sum of an empty one with skipna=True (the default).

In [4]: pd.Series([np.nan]).sum(min_count=1)  # skipna=True by default
Out[4]: nan

The min_count parameter refers to the minimum number of non-null values required for a non-NA sum or product.

Series.prod() has been updated to behave the same as Series.sum(), returning 1 instead.

In [5]: pd.Series([]).prod()
Out[5]: 1.0

In [6]: pd.Series([np.nan]).prod()
Out[6]: 1.0

In [7]: pd.Series([]).prod(min_count=1)
Out[7]: nan

These changes affect DataFrame.sum() and DataFrame.prod() as well. Finally, a few less obvious places in pandas are affected by this change.

Grouping by a Categorical

Grouping by a Categorical and summing now returns 0 instead of NaN for categories with no observations. The product now returns 1 instead of NaN.

pandas 0.21.x

In [8]: grouper = pd.Categorical(['a', 'a'], categories=['a', 'b'])

In [9]: pd.Series([1, 2]).groupby(grouper).sum()
Out[9]:
a    3.0
b    NaN
dtype: float64

pandas 0.22

In [8]: grouper = pd.Categorical(['a', 'a'], categories=['a', 'b'])

In [9]: pd.Series([1, 2]).groupby(grouper).sum()
Out[9]: 
a    3
b    0
dtype: int64

To restore the 0.21 behavior of returning NaN for unobserved groups, use min_count>=1.

In [10]: pd.Series([1, 2]).groupby(grouper).sum(min_count=1)
Out[10]: 
a    3.0
b    NaN
dtype: float64

Resample

The sum and product of all-NA bins has changed from NaN to 0 for sum and 1 for product.

pandas 0.21.x

In [11]: s = pd.Series([1, 1, np.nan, np.nan],
   ...:                index=pd.date_range('2017', periods=4))
   ...:  s
Out[11]:
2017-01-01    1.0
2017-01-02    1.0
2017-01-03    NaN
2017-01-04    NaN
Freq: D, dtype: float64

In [12]: s.resample('2d').sum()
Out[12]:
2017-01-01    2.0
2017-01-03    NaN
Freq: 2D, dtype: float64

pandas 0.22.0

In [11]: s = pd.Series([1, 1, np.nan, np.nan],
   ....:               index=pd.date_range('2017', periods=4))
   ....: 

In [12]: s.resample('2d').sum()
Out[12]: 
2017-01-01    2.0
2017-01-03    0.0
dtype: float64

To restore the 0.21 behavior of returning NaN, use min_count>=1.

In [13]: s.resample('2d').sum(min_count=1)
Out[13]: 
2017-01-01    2.0
2017-01-03    NaN
dtype: float64

In particular, upsampling and taking the sum or product is affected, as upsampling introduces missing values even if the original series was entirely valid.

pandas 0.21.x

In [14]: idx = pd.DatetimeIndex(['2017-01-01', '2017-01-02'])

In [15]: pd.Series([1, 2], index=idx).resample('12H').sum()
Out[15]:
2017-01-01 00:00:00    1.0
2017-01-01 12:00:00    NaN
2017-01-02 00:00:00    2.0
Freq: 12H, dtype: float64

pandas 0.22.0

In [14]: idx = pd.DatetimeIndex(['2017-01-01', '2017-01-02'])

In [15]: pd.Series([1, 2], index=idx).resample("12H").sum()
Out[15]: 
2017-01-01 00:00:00    1
2017-01-01 12:00:00    0
2017-01-02 00:00:00    2
Freq: 12H, dtype: int64

Once again, the min_count keyword is available to restore the 0.21 behavior.

In [16]: pd.Series([1, 2], index=idx).resample("12H").sum(min_count=1)
Out[16]: 
2017-01-01 00:00:00    1.0
2017-01-01 12:00:00    NaN
2017-01-02 00:00:00    2.0
Freq: 12H, dtype: float64

Rolling and Expanding

Rolling and expanding already have a min_periods keyword that behaves similar to min_count. The only case that changes is when doing a rolling or expanding sum with min_periods=0. Previously this returned NaN, when fewer than min_periods non-NA values were in the window. Now it returns 0.

pandas 0.21.1

In [17]: s = pd.Series([np.nan, np.nan])

In [18]: s.rolling(2, min_periods=0).sum()
Out[18]:
0   NaN
1   NaN
dtype: float64

pandas 0.22.0

In [17]: s = pd.Series([np.nan, np.nan])

In [18]: s.rolling(2, min_periods=0).sum()
Out[18]: 
0    0.0
1    0.0
dtype: float64

The default behavior of min_periods=None, implying that min_periods equals the window size, is unchanged.

Compatibility

If you maintain a library that should work across pandas versions, it may be easiest to exclude pandas 0.21 from your requirements. Otherwise, all your sum() calls would need to check if the Series is empty before summing.

With setuptools, in your setup.py use:

install_requires=['pandas!=0.21.*', ...]

With conda, use

requirements:
  run:
    - pandas !=0.21.0,!=0.21.1

Note that the inconsistency in the return value for all-NA series is still there for pandas 0.20.3 and earlier. Avoiding pandas 0.21 will only help with the empty case.

v0.21.1 (December 12, 2017)

This is a minor bug-fix release in the 0.21.x series and includes some small regression fixes, bug fixes and performance improvements. We recommend that all users upgrade to this version.

Highlights include:

  • Temporarily restore matplotlib datetime plotting functionality. This should resolve issues for users who implicitly relied on pandas to plot datetimes with matplotlib. See here.
  • Improvements to the Parquet IO functions introduced in 0.21.0. See here.

Restore Matplotlib datetime Converter Registration

Pandas implements some matplotlib converters for nicely formatting the axis labels on plots with datetime or Period values. Prior to pandas 0.21.0, these were implicitly registered with matplotlib, as a side effect of import pandas.

In pandas 0.21.0, we required users to explicitly register the converter. This caused problems for some users who relied on those converters being present for regular matplotlib.pyplot plotting methods, so we’re temporarily reverting that change; pandas 0.21.1 again registers the converters on import, just like before 0.21.0.

We’ve added a new option to control the converters: pd.options.plotting.matplotlib.register_converters. By default, they are registered. Toggling this to False removes pandas’ formatters and restore any converters we overwrote when registering them (GH18301).

We’re working with the matplotlib developers to make this easier. We’re trying to balance user convenience (automatically registering the converters) with import performance and best practices (importing pandas shouldn’t have the side effect of overwriting any custom converters you’ve already set). In the future we hope to have most of the datetime formatting functionality in matplotlib, with just the pandas-specific converters in pandas. We’ll then gracefully deprecate the automatic registration of converters in favor of users explicitly registering them when they want them.

New features

Improvements to the Parquet IO functionality

Other Enhancements

Deprecations

Performance Improvements

  • Improved performance of plotting large series/dataframes (GH18236).

Bug Fixes

Conversion

  • Bug in TimedeltaIndex subtraction could incorrectly overflow when NaT is present (GH17791)
  • Bug in DatetimeIndex subtracting datetimelike from DatetimeIndex could fail to overflow (GH18020)
  • Bug in IntervalIndex.copy() when copying and IntervalIndex with non-default closed (GH18339)
  • Bug in DataFrame.to_dict() where columns of datetime that are tz-aware were not converted to required arrays when used with orient='records', raising TypeError (GH18372)
  • Bug in DateTimeIndex and date_range() where mismatching tz-aware start and end timezones would not raise an err if end.tzinfo is None (GH18431)
  • Bug in Series.fillna() which raised when passed a long integer on Python 2 (GH18159).

Indexing

  • Bug in a boolean comparison of a datetime.datetime and a datetime64[ns] dtype Series (GH17965)
  • Bug where a MultiIndex with more than a million records was not raising AttributeError when trying to access a missing attribute (GH18165)
  • Bug in IntervalIndex constructor when a list of intervals is passed with non-default closed (GH18334)
  • Bug in Index.putmask when an invalid mask passed (GH18368)
  • Bug in masked assignment of a timedelta64[ns] dtype Series, incorrectly coerced to float (GH18493)

I/O

  • Bug in class:~pandas.io.stata.StataReader not converting date/time columns with display formatting addressed (GH17990). Previously columns with display formatting were normally left as ordinal numbers and not converted to datetime objects.
  • Bug in read_csv() when reading a compressed UTF-16 encoded file (GH18071)
  • Bug in read_csv() for handling null values in index columns when specifying na_filter=False (GH5239)
  • Bug in read_csv() when reading numeric category fields with high cardinality (GH18186)
  • Bug in DataFrame.to_csv() when the table had MultiIndex columns, and a list of strings was passed in for header (GH5539)
  • Bug in parsing integer datetime-like columns with specified format in read_sql (GH17855).
  • Bug in DataFrame.to_msgpack() when serializing data of the numpy.bool_ datatype (GH18390)
  • Bug in read_json() not decoding when reading line deliminted JSON from S3 (GH17200)
  • Bug in pandas.io.json.json_normalize() to avoid modification of meta (GH18610)
  • Bug in to_latex() where repeated multi-index values were not printed even though a higher level index differed from the previous row (GH14484)
  • Bug when reading NaN-only categorical columns in HDFStore (GH18413)
  • Bug in DataFrame.to_latex() with longtable=True where a latex multicolumn always spanned over three columns (GH17959)

Plotting

  • Bug in DataFrame.plot() and Series.plot() with DatetimeIndex where a figure generated by them is not pickleable in Python 3 (GH18439)

Groupby/Resample/Rolling

  • Bug in DataFrame.resample(...).apply(...) when there is a callable that returns different columns (GH15169)
  • Bug in DataFrame.resample(...) when there is a time change (DST) and resampling frequecy is 12h or higher (GH15549)
  • Bug in pd.DataFrameGroupBy.count() when counting over a datetimelike column (GH13393)
  • Bug in rolling.var where calculation is inaccurate with a zero-valued array (GH18430)

Reshaping

  • Error message in pd.merge_asof() for key datatype mismatch now includes datatype of left and right key (GH18068)
  • Bug in pd.concat when empty and non-empty DataFrames or Series are concatenated (GH18178 GH18187)
  • Bug in DataFrame.filter(...) when unicode is passed as a condition in Python 2 (GH13101)
  • Bug when merging empty DataFrames when np.seterr(divide='raise') is set (GH17776)

Numeric

  • Bug in pd.Series.rolling.skew() and rolling.kurt() with all equal values has floating issue (GH18044)

Categorical

  • Bug in DataFrame.astype() where casting to ‘category’ on an empty DataFrame causes a segmentation fault (GH18004)
  • Error messages in the testing module have been improved when items have different CategoricalDtype (GH18069)
  • CategoricalIndex can now correctly take a pd.api.types.CategoricalDtype as its dtype (GH18116)
  • Bug in Categorical.unique() returning read-only codes array when all categories were NaN (GH18051)
  • Bug in DataFrame.groupby(axis=1) with a CategoricalIndex (GH18432)

String

v0.21.0 (October 27, 2017)

This is a major release from 0.20.3 and includes a number of API changes, deprecations, new features, enhancements, and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this version.

Highlights include:

  • Integration with Apache Parquet, including a new top-level read_parquet() function and DataFrame.to_parquet() method, see here.
  • New user-facing pandas.api.types.CategoricalDtype for specifying categoricals independent of the data, see here.
  • The behavior of sum and prod on all-NaN Series/DataFrames is now consistent and no longer depends on whether bottleneck is installed, and sum and prod on empty Series now return NaN instead of 0, see here.
  • Compatibility fixes for pypy, see here.
  • Additions to the drop, reindex and rename API to make them more consistent, see here.
  • Addition of the new methods DataFrame.infer_objects (see here) and GroupBy.pipe (see here).
  • Indexing with a list of labels, where one or more of the labels is missing, is deprecated and will raise a KeyError in a future version, see here.

Check the API Changes and deprecations before updating.

New features

Integration with Apache Parquet file format

Integration with Apache Parquet, including a new top-level read_parquet() and DataFrame.to_parquet() method, see here (GH15838, GH17438).

Apache Parquet provides a cross-language, binary file format for reading and writing data frames efficiently. Parquet is designed to faithfully serialize and de-serialize DataFrame s, supporting all of the pandas dtypes, including extension dtypes such as datetime with timezones.

This functionality depends on either the pyarrow or fastparquet library. For more details, see see the IO docs on Parquet.

infer_objects type conversion

The DataFrame.infer_objects() and Series.infer_objects() methods have been added to perform dtype inference on object columns, replacing some of the functionality of the deprecated convert_objects method. See the documentation here for more details. (GH11221)

This method only performs soft conversions on object columns, converting Python objects to native types, but not any coercive conversions. For example:

In [1]: df = pd.DataFrame({'A': [1, 2, 3],
   ...:                    'B': np.array([1, 2, 3], dtype='object'),
   ...:                    'C': ['1', '2', '3']})
   ...: 

In [2]: df.dtypes
Out[2]: 
A     int64
B    object
C    object
dtype: object

In [3]: df.infer_objects().dtypes
Out[3]: 
A     int64
B     int64
C    object
dtype: object

Note that column 'C' was not converted - only scalar numeric types will be converted to a new type. Other types of conversion should be accomplished using the to_numeric() function (or to_datetime(), to_timedelta()).

In [4]: df = df.infer_objects()

In [5]: df['C'] = pd.to_numeric(df['C'], errors='coerce')

In [6]: df.dtypes
Out[6]: 
A    int64
B    int64
C    int64
dtype: object

Improved warnings when attempting to create columns

New users are often puzzled by the relationship between column operations and attribute access on DataFrame instances (GH7175). One specific instance of this confusion is attempting to create a new column by setting an attribute on the DataFrame:

In[1]: df = pd.DataFrame({'one': [1., 2., 3.]})
In[2]: df.two = [4, 5, 6]

This does not raise any obvious exceptions, but also does not create a new column:

In[3]: df
Out[3]:
    one
0  1.0
1  2.0
2  3.0

Setting a list-like data structure into a new attribute now raises a UserWarning about the potential for unexpected behavior. See Attribute Access.

drop now also accepts index/columns keywords

The drop() method has gained index/columns keywords as an alternative to specifying the axis. This is similar to the behavior of reindex (GH12392).

For example:

In [7]: df = pd.DataFrame(np.arange(8).reshape(2,4),
   ...:                   columns=['A', 'B', 'C', 'D'])
   ...: 

In [8]: df
Out[8]: 
   A  B  C  D
0  0  1  2  3
1  4  5  6  7

In [9]: df.drop(['B', 'C'], axis=1)
Out[9]: 
   A  D
0  0  3
1  4  7

# the following is now equivalent
In [10]: df.drop(columns=['B', 'C'])
Out[10]: 
   A  D
0  0  3
1  4  7

rename, reindex now also accept axis keyword

The DataFrame.rename() and DataFrame.reindex() methods have gained the axis keyword to specify the axis to target with the operation (GH12392).

Here’s rename:

In [11]: df = pd.DataFrame({"A": [1, 2, 3], "B": [4, 5, 6]})

In [12]: df.rename(str.lower, axis='columns')
Out[12]: 
   a  b
0  1  4
1  2  5
2  3  6

In [13]: df.rename(id, axis='index')
Out[13]: 
            A  B
4458898752  1  4
4458898784  2  5
4458898816  3  6

And reindex:

In [14]: df.reindex(['A', 'B', 'C'], axis='columns')
Out[14]: 
   A  B   C
0  1  4 NaN
1  2  5 NaN
2  3  6 NaN

In [15]: df.reindex([0, 1, 3], axis='index')
Out[15]: 
     A    B
0  1.0  4.0
1  2.0  5.0
3  NaN  NaN

The “index, columns” style continues to work as before.

In [16]: df.rename(index=id, columns=str.lower)
Out[16]: 
            a  b
4458898752  1  4
4458898784  2  5
4458898816  3  6

In [17]: df.reindex(index=[0, 1, 3], columns=['A', 'B', 'C'])
Out[17]: 
     A    B   C
0  1.0  4.0 NaN
1  2.0  5.0 NaN
3  NaN  NaN NaN

We highly encourage using named arguments to avoid confusion when using either style.

CategoricalDtype for specifying categoricals

pandas.api.types.CategoricalDtype has been added to the public API and expanded to include the categories and ordered attributes. A CategoricalDtype can be used to specify the set of categories and orderedness of an array, independent of the data. This can be useful for example, when converting string data to a Categorical (GH14711, GH15078, GH16015, GH17643):

In [18]: from pandas.api.types import CategoricalDtype

In [19]: s = pd.Series(['a', 'b', 'c', 'a'])  # strings

In [20]: dtype = CategoricalDtype(categories=['a', 'b', 'c', 'd'], ordered=True)

In [21]: s.astype(dtype)
Out[21]: 
0    a
1    b
2    c
3    a
dtype: category
Categories (4, object): [a < b < c < d]

One place that deserves special mention is in read_csv(). Previously, with dtype={'col': 'category'}, the returned values and categories would always be strings.

In [22]: data = 'A,B\na,1\nb,2\nc,3'

In [23]: pd.read_csv(StringIO(data), dtype={'B': 'category'}).B.cat.categories
Out[23]: Index(['1', '2', '3'], dtype='object')

Notice the “object” dtype.

With a CategoricalDtype of all numerics, datetimes, or timedeltas, we can automatically convert to the correct type

In [24]: dtype = {'B': CategoricalDtype([1, 2, 3])}

In [25]: pd.read_csv(StringIO(data), dtype=dtype).B.cat.categories
Out[25]: Int64Index([1, 2, 3], dtype='int64')

The values have been correctly interpreted as integers.

The .dtype property of a Categorical, CategoricalIndex or a Series with categorical type will now return an instance of CategoricalDtype. While the repr has changed, str(CategoricalDtype()) is still the string 'category'. We’ll take this moment to remind users that the preferred way to detect categorical data is to use pandas.api.types.is_categorical_dtype(), and not str(dtype) == 'category'.

See the CategoricalDtype docs for more.

GroupBy objects now have a pipe method

GroupBy objects now have a pipe method, similar to the one on DataFrame and Series, that allow for functions that take a GroupBy to be composed in a clean, readable syntax. (GH17871)

For a concrete example on combining .groupby and .pipe , imagine having a DataFrame with columns for stores, products, revenue and sold quantity. We’d like to do a groupwise calculation of prices (i.e. revenue/quantity) per store and per product. We could do this in a multi-step operation, but expressing it in terms of piping can make the code more readable.

First we set the data:

In [26]: import numpy as np

In [27]: n = 1000

In [28]: df = pd.DataFrame({'Store': np.random.choice(['Store_1', 'Store_2'], n),
   ....:                    'Product': np.random.choice(['Product_1', 'Product_2', 'Product_3'], n),
   ....:                    'Revenue': (np.random.random(n)*50+10).round(2),
   ....:                    'Quantity': np.random.randint(1, 10, size=n)})
   ....: 

In [29]: df.head(2)
Out[29]: 
     Store    Product  Revenue  Quantity
0  Store_1  Product_3    54.28         3
1  Store_2  Product_2    30.91         1

Now, to find prices per store/product, we can simply do:

In [30]: (df.groupby(['Store', 'Product'])
   ....:    .pipe(lambda grp: grp.Revenue.sum()/grp.Quantity.sum())
   ....:    .unstack().round(2))
   ....: 
Out[30]: 
Product  Product_1  Product_2  Product_3
Store                                   
Store_1       6.37       6.98       7.49
Store_2       7.60       7.01       7.13

See the documentation for more.

Categorical.rename_categories accepts a dict-like

rename_categories() now accepts a dict-like argument for new_categories. The previous categories are looked up in the dictionary’s keys and replaced if found. The behavior of missing and extra keys is the same as in DataFrame.rename().

In [31]: c = pd.Categorical(['a', 'a', 'b'])

In [32]: c.rename_categories({"a": "eh", "b": "bee"})
Out[32]: 
[eh, eh, bee]
Categories (2, object): [eh, bee]

Warning

To assist with upgrading pandas, rename_categories treats Series as list-like. Typically, Series are considered to be dict-like (e.g. in .rename, .map). In a future version of pandas rename_categories will change to treat them as dict-like. Follow the warning message’s recommendations for writing future-proof code.

In [33]: c.rename_categories(pd.Series([0, 1], index=['a', 'c']))
FutureWarning: Treating Series 'new_categories' as a list-like and using the values.
In a future version, 'rename_categories' will treat Series like a dictionary.
For dict-like, use 'new_categories.to_dict()'
For list-like, use 'new_categories.values'.
Out[33]:
[0, 0, 1]
Categories (2, int64): [0, 1]

Other Enhancements

New functions or methods
New keywords
Various enhancements

Backwards incompatible API changes

Dependencies have increased minimum versions

We have updated our minimum supported versions of dependencies (GH15206, GH15543, GH15214). If installed, we now require:

Package Minimum Version Required
Numpy 1.9.0 X
Matplotlib 1.4.3  
Scipy 0.14.0  
Bottleneck 1.0.0  

Additionally, support has been dropped for Python 3.4 (GH15251).

Sum/Prod of all-NaN or empty Series/DataFrames is now consistently NaN

Note

The changes described here have been partially reverted. See the v0.22.0 Whatsnew for more.

The behavior of sum and prod on all-NaN Series/DataFrames no longer depends on whether bottleneck is installed, and return value of sum and prod on an empty Series has changed (GH9422, GH15507).

Calling sum or prod on an empty or all-NaN Series, or columns of a DataFrame, will result in NaN. See the docs.

In [33]: s = Series([np.nan])

Previously WITHOUT bottleneck installed:

In [2]: s.sum()
Out[2]: np.nan

Previously WITH bottleneck:

In [2]: s.sum()
Out[2]: 0.0

New Behavior, without regard to the bottleneck installation:

In [34]: s.sum()
Out[34]: 0.0

Note that this also changes the sum of an empty Series. Previously this always returned 0 regardless of a bottlenck installation:

In [1]: pd.Series([]).sum()
Out[1]: 0

but for consistency with the all-NaN case, this was changed to return NaN as well:

In [35]: pd.Series([]).sum()
Out[35]: 0.0

Indexing with a list with missing labels is Deprecated

Previously, selecting with a list of labels, where one or more labels were missing would always succeed, returning NaN for missing labels. This will now show a FutureWarning. In the future this will raise a KeyError (GH15747). This warning will trigger on a DataFrame or a Series for using .loc[] or [[]] when passing a list-of-labels with at least 1 missing label. See the deprecation docs.

In [36]: s = pd.Series([1, 2, 3])

In [37]: s
Out[37]: 
0    1
1    2
2    3
dtype: int64

Previous Behavior

In [4]: s.loc[[1, 2, 3]]
Out[4]:
1    2.0
2    3.0
3    NaN
dtype: float64

Current Behavior

In [4]: s.loc[[1, 2, 3]]
Passing list-likes to .loc or [] with any missing label will raise
KeyError in the future, you can use .reindex() as an alternative.

See the documentation here:
http://pandas.pydata.org/pandas-docs/stable/indexing.html#deprecate-loc-reindex-listlike

Out[4]:
1    2.0
2    3.0
3    NaN
dtype: float64

The idiomatic way to achieve selecting potentially not-found elements is via .reindex()

In [38]: s.reindex([1, 2, 3])
Out[38]: 
1    2.0
2    3.0
3    NaN
dtype: float64

Selection with all keys found is unchanged.

In [39]: s.loc[[1, 2]]
Out[39]: 
1    2
2    3
dtype: int64

NA naming Changes

In order to promote more consistency among the pandas API, we have added additional top-level functions isna() and notna() that are aliases for isnull() and notnull(). The naming scheme is now more consistent with methods like .dropna() and .fillna(). Furthermore in all cases where .isnull() and .notnull() methods are defined, these have additional methods named .isna() and .notna(), these are included for classes Categorical, Index, Series, and DataFrame. (GH15001).

The configuration option pd.options.mode.use_inf_as_null is deprecated, and pd.options.mode.use_inf_as_na is added as a replacement.

Iteration of Series/Index will now return Python scalars

Previously, when using certain iteration methods for a Series with dtype int or float, you would receive a numpy scalar, e.g. a np.int64, rather than a Python int. Issue (GH10904) corrected this for Series.tolist() and list(Series). This change makes all iteration methods consistent, in particular, for __iter__() and .map(); note that this only affects int/float dtypes. (GH13236, GH13258, GH14216).

In [40]: s = pd.Series([1, 2, 3])

In [41]: s
Out[41]: 
0    1
1    2
2    3
dtype: int64

Previously:

In [2]: type(list(s)[0])
Out[2]: numpy.int64

New Behaviour:

In [42]: type(list(s)[0])
Out[42]: int

Furthermore this will now correctly box the results of iteration for DataFrame.to_dict() as well.

In [43]: d = {'a':[1], 'b':['b']}

In [44]: df = pd.DataFrame(d)

Previously:

In [8]: type(df.to_dict()['a'][0])
Out[8]: numpy.int64

New Behaviour:

In [45]: type(df.to_dict()['a'][0])
Out[45]: int

Indexing with a Boolean Index

Previously when passing a boolean Index to .loc, if the index of the Series/DataFrame had boolean labels, you would get a label based selection, potentially duplicating result labels, rather than a boolean indexing selection (where True selects elements), this was inconsistent how a boolean numpy array indexed. The new behavior is to act like a boolean numpy array indexer. (GH17738)

Previous Behavior:

In [46]: s = pd.Series([1, 2, 3], index=[False, True, False])

In [47]: s
Out[47]: 
False    1
True     2
False    3
dtype: int64
In [59]: s.loc[pd.Index([True, False, True])]
Out[59]:
True     2
False    1
False    3
True     2
dtype: int64

Current Behavior

In [48]: s.loc[pd.Index([True, False, True])]
Out[48]: 
False    1
False    3
dtype: int64

Furthermore, previously if you had an index that was non-numeric (e.g. strings), then a boolean Index would raise a KeyError. This will now be treated as a boolean indexer.

Previously Behavior:

In [49]: s = pd.Series([1,2,3], index=['a', 'b', 'c'])

In [50]: s
Out[50]: 
a    1
b    2
c    3
dtype: int64
In [39]: s.loc[pd.Index([True, False, True])]
KeyError: "None of [Index([True, False, True], dtype='object')] are in the [index]"

Current Behavior

In [51]: s.loc[pd.Index([True, False, True])]
Out[51]: 
a    1
c    3
dtype: int64

PeriodIndex resampling

In previous versions of pandas, resampling a Series/DataFrame indexed by a PeriodIndex returned a DatetimeIndex in some cases (GH12884). Resampling to a multiplied frequency now returns a PeriodIndex (GH15944). As a minor enhancement, resampling a PeriodIndex can now handle NaT values (GH13224)

Previous Behavior:

In [1]: pi = pd.period_range('2017-01', periods=12, freq='M')

In [2]: s = pd.Series(np.arange(12), index=pi)

In [3]: resampled = s.resample('2Q').mean()

In [4]: resampled
Out[4]:
2017-03-31     1.0
2017-09-30     5.5
2018-03-31    10.0
Freq: 2Q-DEC, dtype: float64

In [5]: resampled.index
Out[5]: DatetimeIndex(['2017-03-31', '2017-09-30', '2018-03-31'], dtype='datetime64[ns]', freq='2Q-DEC')

New Behavior:

In [52]: pi = pd.period_range('2017-01', periods=12, freq='M')

In [53]: s = pd.Series(np.arange(12), index=pi)

In [54]: resampled = s.resample('2Q').mean()

In [55]: resampled
Out[55]: 
2017Q1    2.5
2017Q3    8.5
Freq: 2Q-DEC, dtype: float64

In [56]: resampled.index
Out[56]: PeriodIndex(['2017Q1', '2017Q3'], dtype='period[2Q-DEC]', freq='2Q-DEC')

Upsampling and calling .ohlc() previously returned a Series, basically identical to calling .asfreq(). OHLC upsampling now returns a DataFrame with columns open, high, low and close (GH13083). This is consistent with downsampling and DatetimeIndex behavior.

Previous Behavior:

In [1]: pi = pd.PeriodIndex(start='2000-01-01', freq='D', periods=10)

In [2]: s = pd.Series(np.arange(10), index=pi)

In [3]: s.resample('H').ohlc()
Out[3]:
2000-01-01 00:00    0.0
                ...
2000-01-10 23:00    NaN
Freq: H, Length: 240, dtype: float64

In [4]: s.resample('M').ohlc()
Out[4]:
         open  high  low  close
2000-01     0     9    0      9

New Behavior:

In [57]: pi = pd.PeriodIndex(start='2000-01-01', freq='D', periods=10)

In [58]: s = pd.Series(np.arange(10), index=pi)

In [59]: s.resample('H').ohlc()
Out[59]: 
                  open  high  low  close
2000-01-01 00:00   0.0   0.0  0.0    0.0
2000-01-01 01:00   NaN   NaN  NaN    NaN
2000-01-01 02:00   NaN   NaN  NaN    NaN
2000-01-01 03:00   NaN   NaN  NaN    NaN
2000-01-01 04:00   NaN   NaN  NaN    NaN
2000-01-01 05:00   NaN   NaN  NaN    NaN
2000-01-01 06:00   NaN   NaN  NaN    NaN
...                ...   ...  ...    ...
2000-01-10 17:00   NaN   NaN  NaN    NaN
2000-01-10 18:00   NaN   NaN  NaN    NaN
2000-01-10 19:00   NaN   NaN  NaN    NaN
2000-01-10 20:00   NaN   NaN  NaN    NaN
2000-01-10 21:00   NaN   NaN  NaN    NaN
2000-01-10 22:00   NaN   NaN  NaN    NaN
2000-01-10 23:00   NaN   NaN  NaN    NaN

[240 rows x 4 columns]

In [60]: s.resample('M').ohlc()
Out[60]: 
         open  high  low  close
2000-01     0     9    0      9

Improved error handling during item assignment in pd.eval

eval() will now raise a ValueError when item assignment malfunctions, or inplace operations are specified, but there is no item assignment in the expression (GH16732)

In [61]: arr = np.array([1, 2, 3])

Previously, if you attempted the following expression, you would get a not very helpful error message:

In [3]: pd.eval("a = 1 + 2", target=arr, inplace=True)
...
IndexError: only integers, slices (`:`), ellipsis (`...`), numpy.newaxis (`None`)
and integer or boolean arrays are valid indices

This is a very long way of saying numpy arrays don’t support string-item indexing. With this change, the error message is now this:

In [3]: pd.eval("a = 1 + 2", target=arr, inplace=True)
...
ValueError: Cannot assign expression output to target

It also used to be possible to evaluate expressions inplace, even if there was no item assignment:

In [4]: pd.eval("1 + 2", target=arr, inplace=True)
Out[4]: 3

However, this input does not make much sense because the output is not being assigned to the target. Now, a ValueError will be raised when such an input is passed in:

In [4]: pd.eval("1 + 2", target=arr, inplace=True)
...
ValueError: Cannot operate inplace if there is no assignment

Dtype Conversions

Previously assignments, .where() and .fillna() with a bool assignment, would coerce to same the type (e.g. int / float), or raise for datetimelikes. These will now preserve the bools with object dtypes. (GH16821).

In [62]: s = Series([1, 2, 3])
In [5]: s[1] = True

In [6]: s
Out[6]:
0    1
1    1
2    3
dtype: int64

New Behavior

In [63]: s[1] = True

In [64]: s
Out[64]: 
0       1
1    True
2       3
dtype: object

Previously, as assignment to a datetimelike with a non-datetimelike would coerce the non-datetime-like item being assigned (GH14145).

In [65]: s = pd.Series([pd.Timestamp('2011-01-01'), pd.Timestamp('2012-01-01')])
In [1]: s[1] = 1

In [2]: s
Out[2]:
0   2011-01-01 00:00:00.000000000
1   1970-01-01 00:00:00.000000001
dtype: datetime64[ns]

These now coerce to object dtype.

In [66]: s[1] = 1

In [67]: s
Out[67]: 
0    2011-01-01 00:00:00
1                      1
dtype: object
  • Inconsistent behavior in .where() with datetimelikes which would raise rather than coerce to object (GH16402)
  • Bug in assignment against int64 data with np.ndarray with float64 dtype may keep int64 dtype (GH14001)

MultiIndex Constructor with a Single Level

The MultiIndex constructors no longer squeezes a MultiIndex with all length-one levels down to a regular Index. This affects all the MultiIndex constructors. (GH17178)

Previous behavior:

In [2]: pd.MultiIndex.from_tuples([('a',), ('b',)])
Out[2]: Index(['a', 'b'], dtype='object')

Length 1 levels are no longer special-cased. They behave exactly as if you had length 2+ levels, so a MultiIndex is always returned from all of the MultiIndex constructors:

In [68]: pd.MultiIndex.from_tuples([('a',), ('b',)])
Out[68]: 
MultiIndex(levels=[['a', 'b']],
           labels=[[0, 1]])

UTC Localization with Series

Previously, to_datetime() did not localize datetime Series data when utc=True was passed. Now, to_datetime() will correctly localize Series with a datetime64[ns, UTC] dtype to be consistent with how list-like and Index data are handled. (GH6415).

Previous Behavior

In [69]: s = Series(['20130101 00:00:00'] * 3)
In [12]: pd.to_datetime(s, utc=True)
Out[12]:
0   2013-01-01
1   2013-01-01
2   2013-01-01
dtype: datetime64[ns]

New Behavior

In [70]: pd.to_datetime(s, utc=True)
Out[70]: 
0   2013-01-01 00:00:00+00:00
1   2013-01-01 00:00:00+00:00
2   2013-01-01 00:00:00+00:00
dtype: datetime64[ns, UTC]

Additionally, DataFrames with datetime columns that were parsed by read_sql_table() and read_sql_query() will also be localized to UTC only if the original SQL columns were timezone aware datetime columns.

Consistency of Range Functions

In previous versions, there were some inconsistencies between the various range functions: date_range(), bdate_range(), period_range(), timedelta_range(), and interval_range(). (GH17471).

One of the inconsistent behaviors occurred when the start, end and period parameters were all specified, potentially leading to ambiguous ranges. When all three parameters were passed, interval_range ignored the period parameter, period_range ignored the end parameter, and the other range functions raised. To promote consistency among the range functions, and avoid potentially ambiguous ranges, interval_range and period_range will now raise when all three parameters are passed.

Previous Behavior:

In [2]: pd.interval_range(start=0, end=4, periods=6)
Out[2]:
IntervalIndex([(0, 1], (1, 2], (2, 3]]
              closed='right',
              dtype='interval[int64]')

In [3]: pd.period_range(start='2017Q1', end='2017Q4', periods=6, freq='Q')
Out[3]: PeriodIndex(['2017Q1', '2017Q2', '2017Q3', '2017Q4', '2018Q1', '2018Q2'], dtype='period[Q-DEC]', freq='Q-DEC')

New Behavior:

In [2]: pd.interval_range(start=0, end=4, periods=6)
---------------------------------------------------------------------------
ValueError: Of the three parameters: start, end, and periods, exactly two must be specified

In [3]: pd.period_range(start='2017Q1', end='2017Q4', periods=6, freq='Q')
---------------------------------------------------------------------------
ValueError: Of the three parameters: start, end, and periods, exactly two must be specified

Additionally, the endpoint parameter end was not included in the intervals produced by interval_range. However, all other range functions include end in their output. To promote consistency among the range functions, interval_range will now include end as the right endpoint of the final interval, except if freq is specified in a way which skips end.

Previous Behavior:

In [4]: pd.interval_range(start=0, end=4)
Out[4]:
IntervalIndex([(0, 1], (1, 2], (2, 3]]
              closed='right',
              dtype='interval[int64]')

New Behavior:

In [71]: pd.interval_range(start=0, end=4)
Out[71]: 
IntervalIndex([(0, 1], (1, 2], (2, 3], (3, 4]]
              closed='right',
              dtype='interval[int64]')

No Automatic Matplotlib Converters

Pandas no longer registers our date, time, datetime, datetime64, and Period converters with matplotlib when pandas is imported. Matplotlib plot methods (plt.plot, ax.plot, …), will not nicely format the x-axis for DatetimeIndex or PeriodIndex values. You must explicitly register these methods:

Pandas built-in Series.plot and DataFrame.plot will register these converters on first-use (GH17710).

Note

This change has been temporarily reverted in pandas 0.21.1, for more details see here.

Other API Changes

  • The Categorical constructor no longer accepts a scalar for the categories keyword. (GH16022)
  • Accessing a non-existent attribute on a closed HDFStore will now raise an AttributeError rather than a ClosedFileError (GH16301)
  • read_csv() now issues a UserWarning if the names parameter contains duplicates (GH17095)
  • read_csv() now treats 'null' and 'n/a' strings as missing values by default (GH16471, GH16078)
  • pandas.HDFStore’s string representation is now faster and less detailed. For the previous behavior, use pandas.HDFStore.info(). (GH16503).
  • Compression defaults in HDF stores now follow pytables standards. Default is no compression and if complib is missing and complevel > 0 zlib is used (GH15943)
  • Index.get_indexer_non_unique() now returns a ndarray indexer rather than an Index; this is consistent with Index.get_indexer() (GH16819)
  • Removed the @slow decorator from pandas.util.testing, which caused issues for some downstream packages’ test suites. Use @pytest.mark.slow instead, which achieves the same thing (GH16850)
  • Moved definition of MergeError to the pandas.errors module.
  • The signature of Series.set_axis() and DataFrame.set_axis() has been changed from set_axis(axis, labels) to set_axis(labels, axis=0), for consistency with the rest of the API. The old signature is deprecated and will show a FutureWarning (GH14636)
  • Series.argmin() and Series.argmax() will now raise a TypeError when used with object dtypes, instead of a ValueError (GH13595)
  • Period is now immutable, and will now raise an AttributeError when a user tries to assign a new value to the ordinal or freq attributes (GH17116).
  • to_datetime() when passed a tz-aware origin= kwarg will now raise a more informative ValueError rather than a TypeError (GH16842)
  • to_datetime() now raises a ValueError when format includes %W or %U without also including day of the week and calendar year (GH16774)
  • Renamed non-functional index to index_col in read_stata() to improve API consistency (GH16342)
  • Bug in DataFrame.drop() caused boolean labels False and True to be treated as labels 0 and 1 respectively when dropping indices from a numeric index. This will now raise a ValueError (GH16877)
  • Restricted DateOffset keyword arguments. Previously, DateOffset subclasses allowed arbitrary keyword arguments which could lead to unexpected behavior. Now, only valid arguments will be accepted. (GH17176).

Deprecations

Series.select and DataFrame.select

The Series.select() and DataFrame.select() methods are deprecated in favor of using df.loc[labels.map(crit)] (GH12401)

In [72]: df = DataFrame({'A': [1, 2, 3]}, index=['foo', 'bar', 'baz'])
In [3]: df.select(lambda x: x in ['bar', 'baz'])
FutureWarning: select is deprecated and will be removed in a future release. You can use .loc[crit] as a replacement
Out[3]:
     A
bar  2
baz  3
In [73]: df.loc[df.index.map(lambda x: x in ['bar', 'baz'])]
Out[73]: 
     A
bar  2
baz  3

Series.argmax and Series.argmin

The behavior of Series.argmax() and Series.argmin() have been deprecated in favor of Series.idxmax() and Series.idxmin(), respectively (GH16830).

For compatibility with NumPy arrays, pd.Series implements argmax and argmin. Since pandas 0.13.0, argmax has been an alias for pandas.Series.idxmax(), and argmin has been an alias for pandas.Series.idxmin(). They return the label of the maximum or minimum, rather than the position.

We’ve deprecated the current behavior of Series.argmax and Series.argmin. Using either of these will emit a FutureWarning. Use Series.idxmax() if you want the label of the maximum. Use Series.values.argmax() if you want the position of the maximum. Likewise for the minimum. In a future release Series.argmax and Series.argmin will return the position of the maximum or minimum.

Removal of prior version deprecations/changes

  • read_excel() has dropped the has_index_names parameter (GH10967)
  • The pd.options.display.height configuration has been dropped (GH3663)
  • The pd.options.display.line_width configuration has been dropped (GH2881)
  • The pd.options.display.mpl_style configuration has been dropped (GH12190)
  • Index has dropped the .sym_diff() method in favor of .symmetric_difference() (GH12591)
  • Categorical has dropped the .order() and .sort() methods in favor of .sort_values() (GH12882)
  • eval() and DataFrame.eval() have changed the default of inplace from None to False (GH11149)
  • The function get_offset_name has been dropped in favor of the .freqstr attribute for an offset (GH11834)
  • pandas no longer tests for compatibility with hdf5-files created with pandas < 0.11 (GH17404).

Performance Improvements

  • Improved performance of instantiating SparseDataFrame (GH16773)
  • Series.dt no longer performs frequency inference, yielding a large speedup when accessing the attribute (GH17210)
  • Improved performance of set_categories() by not materializing the values (GH17508)
  • Timestamp.microsecond no longer re-computes on attribute access (GH17331)
  • Improved performance of the CategoricalIndex for data that is already categorical dtype (GH17513)
  • Improved performance of RangeIndex.min() and RangeIndex.max() by using RangeIndex properties to perform the computations (GH17607)

Documentation Changes

  • Several NaT method docstrings (e.g. NaT.ctime()) were incorrect (GH17327)
  • The documentation has had references to versions < v0.17 removed and cleaned up (GH17442, GH17442, GH17404 & GH17504)

Bug Fixes

Conversion

  • Bug in assignment against datetime-like data with int may incorrectly convert to datetime-like (GH14145)
  • Bug in assignment against int64 data with np.ndarray with float64 dtype may keep int64 dtype (GH14001)
  • Fixed the return type of IntervalIndex.is_non_overlapping_monotonic to be a Python bool for consistency with similar attributes/methods. Previously returned a numpy.bool_. (GH17237)
  • Bug in IntervalIndex.is_non_overlapping_monotonic when intervals are closed on both sides and overlap at a point (GH16560)
  • Bug in Series.fillna() returns frame when inplace=True and value is dict (GH16156)
  • Bug in Timestamp.weekday_name returning a UTC-based weekday name when localized to a timezone (GH17354)
  • Bug in Timestamp.replace when replacing tzinfo around DST changes (GH15683)
  • Bug in Timedelta construction and arithmetic that would not propagate the Overflow exception (GH17367)
  • Bug in astype() converting to object dtype when passed extension type classes (DatetimeTZDtype, CategoricalDtype) rather than instances. Now a TypeError is raised when a class is passed (GH17780).
  • Bug in to_numeric() in which elements were not always being coerced to numeric when errors='coerce' (GH17007, GH17125)
  • Bug in DataFrame and Series constructors where range objects are converted to int32 dtype on Windows instead of int64 (GH16804)

Indexing

  • When called with a null slice (e.g. df.iloc[:]), the .iloc and .loc indexers return a shallow copy of the original object. Previously they returned the original object. (GH13873).
  • When called on an unsorted MultiIndex, the loc indexer now will raise UnsortedIndexError only if proper slicing is used on non-sorted levels (GH16734).
  • Fixes regression in 0.20.3 when indexing with a string on a TimedeltaIndex (GH16896).
  • Fixed TimedeltaIndex.get_loc() handling of np.timedelta64 inputs (GH16909).
  • Fix MultiIndex.sort_index() ordering when ascending argument is a list, but not all levels are specified, or are in a different order (GH16934).
  • Fixes bug where indexing with np.inf caused an OverflowError to be raised (GH16957)
  • Bug in reindexing on an empty CategoricalIndex (GH16770)
  • Fixes DataFrame.loc for setting with alignment and tz-aware DatetimeIndex (GH16889)
  • Avoids IndexError when passing an Index or Series to .iloc with older numpy (GH17193)
  • Allow unicode empty strings as placeholders in multilevel columns in Python 2 (GH17099)
  • Bug in .iloc when used with inplace addition or assignment and an int indexer on a MultiIndex causing the wrong indexes to be read from and written to (GH17148)
  • Bug in .isin() in which checking membership in empty Series objects raised an error (GH16991)
  • Bug in CategoricalIndex reindexing in which specified indices containing duplicates were not being respected (GH17323)
  • Bug in intersection of RangeIndex with negative step (GH17296)
  • Bug in IntervalIndex where performing a scalar lookup fails for included right endpoints of non-overlapping monotonic decreasing indexes (GH16417, GH17271)
  • Bug in DataFrame.first_valid_index() and DataFrame.last_valid_index() when no valid entry (GH17400)
  • Bug in Series.rename() when called with a callable, incorrectly alters the name of the Series, rather than the name of the Index. (GH17407)
  • Bug in String.str_get() raises IndexError instead of inserting NaNs when using a negative index. (GH17704)

I/O

  • Bug in read_hdf() when reading a timezone aware index from fixed format HDFStore (GH17618)
  • Bug in read_csv() in which columns were not being thoroughly de-duplicated (GH17060)
  • Bug in read_csv() in which specified column names were not being thoroughly de-duplicated (GH17095)
  • Bug in read_csv() in which non integer values for the header argument generated an unhelpful / unrelated error message (GH16338)
  • Bug in read_csv() in which memory management issues in exception handling, under certain conditions, would cause the interpreter to segfault (GH14696, GH16798).
  • Bug in read_csv() when called with low_memory=False in which a CSV with at least one column > 2GB in size would incorrectly raise a MemoryError (GH16798).
  • Bug in read_csv() when called with a single-element list header would return a DataFrame of all NaN values (GH7757)
  • Bug in DataFrame.to_csv() defaulting to ‘ascii’ encoding in Python 3, instead of ‘utf-8’ (GH17097)
  • Bug in read_stata() where value labels could not be read when using an iterator (GH16923)
  • Bug in read_stata() where the index was not set (GH16342)
  • Bug in read_html() where import check fails when run in multiple threads (GH16928)
  • Bug in read_csv() where automatic delimiter detection caused a TypeError to be thrown when a bad line was encountered rather than the correct error message (GH13374)
  • Bug in DataFrame.to_html() with notebook=True where DataFrames with named indices or non-MultiIndex indices had undesired horizontal or vertical alignment for column or row labels, respectively (GH16792)
  • Bug in DataFrame.to_html() in which there was no validation of the justify parameter (GH17527)
  • Bug in HDFStore.select() when reading a contiguous mixed-data table featuring VLArray (GH17021)
  • Bug in to_json() where several conditions (including objects with unprintable symbols, objects with deep recursion, overlong labels) caused segfaults instead of raising the appropriate exception (GH14256)

Plotting

  • Bug in plotting methods using secondary_y and fontsize not setting secondary axis font size (GH12565)
  • Bug when plotting timedelta and datetime dtypes on y-axis (GH16953)
  • Line plots no longer assume monotonic x data when calculating xlims, they show the entire lines now even for unsorted x data. (GH11310, GH11471)
  • With matplotlib 2.0.0 and above, calculation of x limits for line plots is left to matplotlib, so that its new default settings are applied. (GH15495)
  • Bug in Series.plot.bar or DataFrame.plot.bar with y not respecting user-passed color (GH16822)
  • Bug causing plotting.parallel_coordinates to reset the random seed when using random colors (GH17525)

Groupby/Resample/Rolling

  • Bug in DataFrame.resample(...).size() where an empty DataFrame did not return a Series (GH14962)
  • Bug in infer_freq() causing indices with 2-day gaps during the working week to be wrongly inferred as business daily (GH16624)
  • Bug in .rolling(...).quantile() which incorrectly used different defaults than Series.quantile() and DataFrame.quantile() (GH9413, GH16211)
  • Bug in groupby.transform() that would coerce boolean dtypes back to float (GH16875)
  • Bug in Series.resample(...).apply() where an empty Series modified the source index and did not return the name of a Series (GH14313)
  • Bug in .rolling(...).apply(...) with a DataFrame with a DatetimeIndex, a window of a timedelta-convertible and min_periods >= 1 (GH15305)
  • Bug in DataFrame.groupby where index and column keys were not recognized correctly when the number of keys equaled the number of elements on the groupby axis (GH16859)
  • Bug in groupby.nunique() with TimeGrouper which cannot handle NaT correctly (GH17575)
  • Bug in DataFrame.groupby where a single level selection from a MultiIndex unexpectedly sorts (GH17537)
  • Bug in DataFrame.groupby where spurious warning is raised when Grouper object is used to override ambiguous column name (GH17383)
  • Bug in TimeGrouper differs when passes as a list and as a scalar (GH17530)

Sparse

  • Bug in SparseSeries raises AttributeError when a dictionary is passed in as data (GH16905)
  • Bug in SparseDataFrame.fillna() not filling all NaNs when frame was instantiated from SciPy sparse matrix (GH16112)
  • Bug in SparseSeries.unstack() and SparseDataFrame.stack() (GH16614, GH15045)
  • Bug in make_sparse() treating two numeric/boolean data, which have same bits, as same when array dtype is object (GH17574)
  • SparseArray.all() and SparseArray.any() are now implemented to handle SparseArray, these were used but not implemented (GH17570)

Reshaping

  • Joining/Merging with a non unique PeriodIndex raised a TypeError (GH16871)
  • Bug in crosstab() where non-aligned series of integers were casted to float (GH17005)
  • Bug in merging with categorical dtypes with datetimelikes incorrectly raised a TypeError (GH16900)
  • Bug when using isin() on a large object series and large comparison array (GH16012)
  • Fixes regression from 0.20, Series.aggregate() and DataFrame.aggregate() allow dictionaries as return values again (GH16741)
  • Fixes dtype of result with integer dtype input, from pivot_table() when called with margins=True (GH17013)
  • Bug in crosstab() where passing two Series with the same name raised a KeyError (GH13279)
  • Series.argmin(), Series.argmax(), and their counterparts on DataFrame and groupby objects work correctly with floating point data that contains infinite values (GH13595).
  • Bug in unique() where checking a tuple of strings raised a TypeError (GH17108)
  • Bug in concat() where order of result index was unpredictable if it contained non-comparable elements (GH17344)
  • Fixes regression when sorting by multiple columns on a datetime64 dtype Series with NaT values (GH16836)
  • Bug in pivot_table() where the result’s columns did not preserve the categorical dtype of columns when dropna was False (GH17842)
  • Bug in DataFrame.drop_duplicates where dropping with non-unique column names raised a ValueError (GH17836)
  • Bug in unstack() which, when called on a list of levels, would discard the fillna argument (GH13971)
  • Bug in the alignment of range objects and other list-likes with DataFrame leading to operations being performed row-wise instead of column-wise (GH17901)

Numeric

  • Bug in .clip() with axis=1 and a list-like for threshold is passed; previously this raised ValueError (GH15390)
  • Series.clip() and DataFrame.clip() now treat NA values for upper and lower arguments as None instead of raising ValueError (GH17276).

Categorical

  • Bug in Series.isin() when called with a categorical (GH16639)
  • Bug in the categorical constructor with empty values and categories causing the .categories to be an empty Float64Index rather than an empty Index with object dtype (GH17248)
  • Bug in categorical operations with Series.cat not preserving the original Series’ name (GH17509)
  • Bug in DataFrame.merge() failing for categorical columns with boolean/int data types (GH17187)
  • Bug in constructing a Categorical/CategoricalDtype when the specified categories are of categorical type (GH17884).

PyPy

  • Compatibility with PyPy in read_csv() with usecols=[<unsorted ints>] and read_json() (GH17351)
  • Split tests into cases for CPython and PyPy where needed, which highlights the fragility of index matching with float('nan'), np.nan and NAT (GH17351)
  • Fix DataFrame.memory_usage() to support PyPy. Objects on PyPy do not have a fixed size, so an approximation is used instead (GH17228)

Other

  • Bug where some inplace operators were not being wrapped and produced a copy when invoked (GH12962)
  • Bug in eval() where the inplace parameter was being incorrectly handled (GH16732)

v0.20.3 (July 7, 2017)

This is a minor bug-fix release in the 0.20.x series and includes some small regression fixes and bug fixes. We recommend that all users upgrade to this version.

Bug Fixes

  • Fixed a bug in failing to compute rolling computations of a column-MultiIndexed DataFrame (GH16789, GH16825)
  • Fixed a pytest marker failing downstream packages’ tests suites (GH16680)

Conversion

  • Bug in pickle compat prior to the v0.20.x series, when UTC is a timezone in a Series/DataFrame/Index (GH16608)
  • Bug in Series construction when passing a Series with dtype='category' (GH16524).
  • Bug in DataFrame.astype() when passing a Series as the dtype kwarg. (GH16717).

Indexing

  • Bug in Float64Index causing an empty array instead of None to be returned from .get(np.nan) on a Series whose index did not contain any NaN s (GH8569)
  • Bug in MultiIndex.isin causing an error when passing an empty iterable (GH16777)
  • Fixed a bug in a slicing DataFrame/Series that have a TimedeltaIndex (GH16637)

I/O

  • Bug in read_csv() in which files weren’t opened as binary files by the C engine on Windows, causing EOF characters mid-field, which would fail (GH16039, GH16559, GH16675)
  • Bug in read_hdf() in which reading a Series saved to an HDF file in ‘fixed’ format fails when an explicit mode='r' argument is supplied (GH16583)
  • Bug in DataFrame.to_latex() where bold_rows was wrongly specified to be True by default, whereas in reality row labels remained non-bold whatever parameter provided. (GH16707)
  • Fixed an issue with DataFrame.style() where generated element ids were not unique (GH16780)
  • Fixed loading a DataFrame with a PeriodIndex, from a format='fixed' HDFStore, in Python 3, that was written in Python 2 (GH16781)

Plotting

  • Fixed regression that prevented RGB and RGBA tuples from being used as color arguments (GH16233)
  • Fixed an issue with DataFrame.plot.scatter() that incorrectly raised a KeyError when categorical data is used for plotting (GH16199)

Reshaping

  • PeriodIndex / TimedeltaIndex.join was missing the sort= kwarg (GH16541)
  • Bug in joining on a MultiIndex with a category dtype for a level (GH16627).
  • Bug in merge() when merging/joining with multiple categorical columns (GH16767)

Categorical

  • Bug in DataFrame.sort_values not respecting the kind parameter with categorical data (GH16793)

v0.20.2 (June 4, 2017)

This is a minor bug-fix release in the 0.20.x series and includes some small regression fixes, bug fixes and performance improvements. We recommend that all users upgrade to this version.

Enhancements

  • Unblocked access to additional compression types supported in pytables: ‘blosc:blosclz, ‘blosc:lz4’, ‘blosc:lz4hc’, ‘blosc:snappy’, ‘blosc:zlib’, ‘blosc:zstd’ (GH14478)
  • Series provides a to_latex method (GH16180)
  • A new groupby method ngroup(), parallel to the existing cumcount(), has been added to return the group order (GH11642); see here.

Performance Improvements

  • Performance regression fix when indexing with a list-like (GH16285)
  • Performance regression fix for MultiIndexes (GH16319, GH16346)
  • Improved performance of .clip() with scalar arguments (GH15400)
  • Improved performance of groupby with categorical groupers (GH16413)
  • Improved performance of MultiIndex.remove_unused_levels() (GH16556)

Bug Fixes

  • Silenced a warning on some Windows environments about “tput: terminal attributes: No such device or address” when detecting the terminal size. This fix only applies to python 3 (GH16496)
  • Bug in using pathlib.Path or py.path.local objects with io functions (GH16291)
  • Bug in Index.symmetric_difference() on two equal MultiIndex’s, results in a TypeError (GH13490)
  • Bug in DataFrame.update() with overwrite=False and NaN values (GH15593)
  • Passing an invalid engine to read_csv() now raises an informative ValueError rather than UnboundLocalError. (GH16511)
  • Bug in unique() on an array of tuples (GH16519)
  • Bug in cut() when labels are set, resulting in incorrect label ordering (GH16459)
  • Fixed a compatibility issue with IPython 6.0’s tab completion showing deprecation warnings on Categoricals (GH16409)

Conversion

  • Bug in to_numeric() in which empty data inputs were causing a segfault of the interpreter (GH16302)
  • Silence numpy warnings when broadcasting DataFrame to Series with comparison ops (GH16378, GH16306)

Indexing

  • Bug in DataFrame.reset_index(level=) with single level index (GH16263)
  • Bug in partial string indexing with a monotonic, but not strictly-monotonic, index incorrectly reversing the slice bounds (GH16515)
  • Bug in MultiIndex.remove_unused_levels() that would not return a MultiIndex equal to the original. (GH16556)

I/O

  • Bug in read_csv() when comment is passed in a space delimited text file (GH16472)
  • Bug in read_csv() not raising an exception with nonexistent columns in usecols when it had the correct length (GH14671)
  • Bug that would force importing of the clipboard routines unnecessarily, potentially causing an import error on startup (GH16288)
  • Bug that raised IndexError when HTML-rendering an empty DataFrame (GH15953)
  • Bug in read_csv() in which tarfile object inputs were raising an error in Python 2.x for the C engine (GH16530)
  • Bug where DataFrame.to_html() ignored the index_names parameter (GH16493)
  • Bug where pd.read_hdf() returns numpy strings for index names (GH13492)
  • Bug in HDFStore.select_as_multiple() where start/stop arguments were not respected (GH16209)

Plotting

  • Bug in DataFrame.plot with a single column and a list-like color (GH3486)
  • Bug in plot where NaT in DatetimeIndex results in Timestamp.min (GH12405)
  • Bug in DataFrame.boxplot where figsize keyword was not respected for non-grouped boxplots (GH11959)

Groupby/Resample/Rolling

  • Bug in creating a time-based rolling window on an empty DataFrame (GH15819)
  • Bug in rolling.cov() with offset window (GH16058)
  • Bug in .resample() and .groupby() when aggregating on integers (GH16361)

Sparse

  • Bug in construction of SparseDataFrame from scipy.sparse.dok_matrix (GH16179)

Reshaping

  • Bug in DataFrame.stack with unsorted levels in MultiIndex columns (GH16323)
  • Bug in pd.wide_to_long() where no error was raised when i was not a unique identifier (GH16382)
  • Bug in Series.isin(..) with a list of tuples (GH16394)
  • Bug in construction of a DataFrame with mixed dtypes including an all-NaT column. (GH16395)
  • Bug in DataFrame.agg() and Series.agg() with aggregating on non-callable attributes (GH16405)

Numeric

  • Bug in .interpolate(), where limit_direction was not respected when limit=None (default) was passed (GH16282)

Categorical

  • Fixed comparison operations considering the order of the categories when both categoricals are unordered (GH16014)

Other

  • Bug in DataFrame.drop() with an empty-list with non-unique indices (GH16270)

v0.20.1 (May 5, 2017)

This is a major release from 0.19.2 and includes a number of API changes, deprecations, new features, enhancements, and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this version.

Highlights include:

  • New .agg() API for Series/DataFrame similar to the groupby-rolling-resample API’s, see here
  • Integration with the feather-format, including a new top-level pd.read_feather() and DataFrame.to_feather() method, see here.
  • The .ix indexer has been deprecated, see here
  • Panel has been deprecated, see here
  • Addition of an IntervalIndex and Interval scalar type, see here
  • Improved user API when grouping by index levels in .groupby(), see here
  • Improved support for UInt64 dtypes, see here
  • A new orient for JSON serialization, orient='table', that uses the Table Schema spec and that gives the possibility for a more interactive repr in the Jupyter Notebook, see here
  • Experimental support for exporting styled DataFrames (DataFrame.style) to Excel, see here
  • Window binary corr/cov operations now return a MultiIndexed DataFrame rather than a Panel, as Panel is now deprecated, see here
  • Support for S3 handling now uses s3fs, see here
  • Google BigQuery support now uses the pandas-gbq library, see here

Warning

Pandas has changed the internal structure and layout of the codebase. This can affect imports that are not from the top-level pandas.* namespace, please see the changes here.

Check the API Changes and deprecations before updating.

Note

This is a combined release for 0.20.0 and and 0.20.1. Version 0.20.1 contains one additional change for backwards-compatibility with downstream projects using pandas’ utils routines. (GH16250)

What’s new in v0.20.0

New features

agg API for DataFrame/Series

Series & DataFrame have been enhanced to support the aggregation API. This is a familiar API from groupby, window operations, and resampling. This allows aggregation operations in a concise way by using agg() and transform(). The full documentation is here (GH1623).

Here is a sample

In [1]: df = pd.DataFrame(np.random.randn(10, 3), columns=['A', 'B', 'C'],
   ...:                  index=pd.date_range('1/1/2000', periods=10))
   ...: 

In [2]: df.iloc[3:7] = np.nan

In [3]: df
Out[3]: 
                   A         B         C
2000-01-01  1.682600  0.413582  1.689516
2000-01-02 -2.099110 -1.180182  1.595661
2000-01-03 -0.419048  0.522165 -1.208946
2000-01-04       NaN       NaN       NaN
2000-01-05       NaN       NaN       NaN
2000-01-06       NaN       NaN       NaN
2000-01-07       NaN       NaN       NaN
2000-01-08  0.955435 -0.133009  2.011466
2000-01-09  0.578780  0.897126 -0.980013
2000-01-10 -0.045748  0.361601 -0.208039

One can operate using string function names, callables, lists, or dictionaries of these.

Using a single function is equivalent to .apply.

In [4]: df.agg('sum')
Out[4]: 
A    0.652908
B    0.881282
C    2.899645
dtype: float64

Multiple aggregations with a list of functions.

In [5]: df.agg(['sum', 'min'])
Out[5]: 
            A         B         C
sum  0.652908  0.881282  2.899645
min -2.099110 -1.180182 -1.208946

Using a dict provides the ability to apply specific aggregations per column. You will get a matrix-like output of all of the aggregators. The output has one column per unique function. Those functions applied to a particular column will be NaN:

In [6]: df.agg({'A' : ['sum', 'min'], 'B' : ['min', 'max']})
Out[6]: 
            A         B
max       NaN  0.897126
min -2.099110 -1.180182
sum  0.652908       NaN

The API also supports a .transform() function for broadcasting results.

In [7]: df.transform(['abs', lambda x: x - x.min()])
Out[7]: 
                   A                   B                   C          
                 abs  <lambda>       abs  <lambda>       abs  <lambda>
2000-01-01  1.682600  3.781710  0.413582  1.593764  1.689516  2.898461
2000-01-02  2.099110  0.000000  1.180182  0.000000  1.595661  2.804606
2000-01-03  0.419048  1.680062  0.522165  1.702346  1.208946  0.000000
2000-01-04       NaN       NaN       NaN       NaN       NaN       NaN
2000-01-05       NaN       NaN       NaN       NaN       NaN       NaN
2000-01-06       NaN       NaN       NaN       NaN       NaN       NaN
2000-01-07       NaN       NaN       NaN       NaN       NaN       NaN
2000-01-08  0.955435  3.054545  0.133009  1.047173  2.011466  3.220412
2000-01-09  0.578780  2.677890  0.897126  2.077307  0.980013  0.228932
2000-01-10  0.045748  2.053362  0.361601  1.541782  0.208039  1.000907

When presented with mixed dtypes that cannot be aggregated, .agg() will only take the valid aggregations. This is similar to how groupby .agg() works. (GH15015)

In [8]: df = pd.DataFrame({'A': [1, 2, 3],
   ...:                    'B': [1., 2., 3.],
   ...:                    'C': ['foo', 'bar', 'baz'],
   ...:                    'D': pd.date_range('20130101', periods=3)})
   ...: 

In [9]: df.dtypes
Out[9]: 
A             int64
B           float64
C            object
D    datetime64[ns]
dtype: object
In [10]: df.agg(['min', 'sum'])
Out[10]: 
     A    B          C          D
min  1  1.0        bar 2013-01-01
sum  6  6.0  foobarbaz        NaT

dtype keyword for data IO

The 'python' engine for read_csv(), as well as the read_fwf() function for parsing fixed-width text files and read_excel() for parsing Excel files, now accept the dtype keyword argument for specifying the types of specific columns (GH14295). See the io docs for more information.

In [11]: data = "a  b\n1  2\n3  4"

In [12]: pd.read_fwf(StringIO(data)).dtypes
Out[12]: 
a    int64
b    int64
dtype: object

In [13]: pd.read_fwf(StringIO(data), dtype={'a':'float64', 'b':'object'}).dtypes
Out[13]: 
a    float64
b     object
dtype: object

.to_datetime() has gained an origin parameter

to_datetime() has gained a new parameter, origin, to define a reference date from where to compute the resulting timestamps when parsing numerical values with a specific unit specified. (GH11276, GH11745)

For example, with 1960-01-01 as the starting date:

In [14]: pd.to_datetime([1, 2, 3], unit='D', origin=pd.Timestamp('1960-01-01'))
Out[14]: DatetimeIndex(['1960-01-02', '1960-01-03', '1960-01-04'], dtype='datetime64[ns]', freq=None)

The default is set at origin='unix', which defaults to 1970-01-01 00:00:00, which is commonly called ‘unix epoch’ or POSIX time. This was the previous default, so this is a backward compatible change.

In [15]: pd.to_datetime([1, 2, 3], unit='D')
Out[15]: DatetimeIndex(['1970-01-02', '1970-01-03', '1970-01-04'], dtype='datetime64[ns]', freq=None)

Groupby Enhancements

Strings passed to DataFrame.groupby() as the by parameter may now reference either column names or index level names. Previously, only column names could be referenced. This allows to easily group by a column and index level at the same time. (GH5677)

In [16]: arrays = [['bar', 'bar', 'baz', 'baz', 'foo', 'foo', 'qux', 'qux'],
   ....:           ['one', 'two', 'one', 'two', 'one', 'two', 'one', 'two']]
   ....: 

In [17]: index = pd.MultiIndex.from_arrays(arrays, names=['first', 'second'])

In [18]: df = pd.DataFrame({'A': [1, 1, 1, 1, 2, 2, 3, 3],
   ....:                    'B': np.arange(8)},
   ....:                   index=index)
   ....: 

In [19]: df
Out[19]: 
              A  B
first second      
bar   one     1  0
      two     1  1
baz   one     1  2
      two     1  3
foo   one     2  4
      two     2  5
qux   one     3  6
      two     3  7

In [20]: df.groupby(['second', 'A']).sum()
Out[20]: 
          B
second A   
one    1  2
       2  4
       3  6
two    1  4
       2  5
       3  7

Better support for compressed URLs in read_csv

The compression code was refactored (GH12688). As a result, reading dataframes from URLs in read_csv() or read_table() now supports additional compression methods: xz, bz2, and zip (GH14570). Previously, only gzip compression was supported. By default, compression of URLs and paths are now inferred using their file extensions. Additionally, support for bz2 compression in the python 2 C-engine improved (GH14874).

In [21]: url = 'https://github.com/{repo}/raw/{branch}/{path}'.format(
   ....:     repo = 'pandas-dev/pandas',
   ....:     branch = 'master',
   ....:     path = 'pandas/tests/io/parser/data/salaries.csv.bz2',
   ....: )
   ....: 

In [22]: df = pd.read_table(url, compression='infer')  # default, infer compression

In [23]: df = pd.read_table(url, compression='bz2')  # explicitly specify compression

In [24]: df.head(2)
Out[24]: 
       S  X  E  M
0  13876  1  1  1
1  11608  1  3  0

Pickle file I/O now supports compression

read_pickle(), DataFrame.to_pickle() and Series.to_pickle() can now read from and write to compressed pickle files. Compression methods can be an explicit parameter or be inferred from the file extension. See the docs here.

In [25]: df = pd.DataFrame({
   ....:     'A': np.random.randn(1000),
   ....:     'B': 'foo',
   ....:     'C': pd.date_range('20130101', periods=1000, freq='s')})
   ....: 

Using an explicit compression type

In [26]: df.to_pickle("data.pkl.compress", compression="gzip")

In [27]: rt = pd.read_pickle("data.pkl.compress", compression="gzip")

In [28]: rt.head()
Out[28]: 
          A    B                   C
0  1.578227  foo 2013-01-01 00:00:00
1 -0.230575  foo 2013-01-01 00:00:01
2  0.695530  foo 2013-01-01 00:00:02
3 -0.466001  foo 2013-01-01 00:00:03
4 -0.154972  foo 2013-01-01 00:00:04

The default is to infer the compression type from the extension (compression='infer'):

In [29]: df.to_pickle("data.pkl.gz")

In [30]: rt = pd.read_pickle("data.pkl.gz")

In [31]: rt.head()
Out[31]: 
          A    B                   C
0  1.578227  foo 2013-01-01 00:00:00
1 -0.230575  foo 2013-01-01 00:00:01
2  0.695530  foo 2013-01-01 00:00:02
3 -0.466001  foo 2013-01-01 00:00:03
4 -0.154972  foo 2013-01-01 00:00:04

In [32]: df["A"].to_pickle("s1.pkl.bz2")

In [33]: rt = pd.read_pickle("s1.pkl.bz2")

In [34]: rt.head()
Out[34]: 
0    1.578227
1   -0.230575
2    0.695530
3   -0.466001
4   -0.154972
Name: A, dtype: float64

UInt64 Support Improved

Pandas has significantly improved support for operations involving unsigned, or purely non-negative, integers. Previously, handling these integers would result in improper rounding or data-type casting, leading to incorrect results. Notably, a new numerical index, UInt64Index, has been created (GH14937)

In [35]: idx = pd.UInt64Index([1, 2, 3])

In [36]: df = pd.DataFrame({'A': ['a', 'b', 'c']}, index=idx)

In [37]: df.index
Out[37]: UInt64Index([1, 2, 3], dtype='uint64')
  • Bug in converting object elements of array-like objects to unsigned 64-bit integers (GH4471, GH14982)
  • Bug in Series.unique() in which unsigned 64-bit integers were causing overflow (GH14721)
  • Bug in DataFrame construction in which unsigned 64-bit integer elements were being converted to objects (GH14881)
  • Bug in pd.read_csv() in which unsigned 64-bit integer elements were being improperly converted to the wrong data types (GH14983)
  • Bug in pd.unique() in which unsigned 64-bit integers were causing overflow (GH14915)
  • Bug in pd.value_counts() in which unsigned 64-bit integers were being erroneously truncated in the output (GH14934)

GroupBy on Categoricals

In previous versions, .groupby(..., sort=False) would fail with a ValueError when grouping on a categorical series with some categories not appearing in the data. (GH13179)

In [38]: chromosomes = np.r_[np.arange(1, 23).astype(str), ['X', 'Y']]

In [39]: df = pd.DataFrame({
   ....:     'A': np.random.randint(100),
   ....:     'B': np.random.randint(100),
   ....:     'C': np.random.randint(100),
   ....:     'chromosomes': pd.Categorical(np.random.choice(chromosomes, 100),
   ....:                                   categories=chromosomes,
   ....:                                   ordered=True)})
   ....: 

In [40]: df
Out[40]: 
     A   B   C chromosomes
0   80  36  94          12
1   80  36  94           X
2   80  36  94          19
3   80  36  94          22
4   80  36  94          17
5   80  36  94           6
6   80  36  94          13
..  ..  ..  ..         ...
93  80  36  94          21
94  80  36  94          20
95  80  36  94          11
96  80  36  94          16
97  80  36  94          21
98  80  36  94          18
99  80  36  94           8

[100 rows x 4 columns]

Previous Behavior:

In [3]: df[df.chromosomes != '1'].groupby('chromosomes', sort=False).sum()
---------------------------------------------------------------------------
ValueError: items in new_categories are not the same as in old categories

New Behavior:

In [41]: df[df.chromosomes != '1'].groupby('chromosomes', sort=False).sum()
Out[41]: 
               A    B    C
chromosomes               
2            320  144  376
3            400  180  470
4            240  108  282
5            240  108  282
6            400  180  470
7            400  180  470
8            480  216  564
...          ...  ...  ...
19           400  180  470
20           160   72  188
21           480  216  564
22           160   72  188
X            400  180  470
Y            320  144  376
1              0    0    0

[24 rows x 3 columns]

Table Schema Output

The new orient 'table' for DataFrame.to_json() will generate a Table Schema compatible string representation of the data.

In [42]: df = pd.DataFrame(
   ....:     {'A': [1, 2, 3],
   ....:      'B': ['a', 'b', 'c'],
   ....:      'C': pd.date_range('2016-01-01', freq='d', periods=3),
   ....:     }, index=pd.Index(range(3), name='idx'))
   ....: 

In [43]: df
Out[43]: 
     A  B          C
idx                 
0    1  a 2016-01-01
1    2  b 2016-01-02
2    3  c 2016-01-03

In [44]: df.to_json(orient='table')
Out[44]: '{"schema": {"fields":[{"name":"idx","type":"integer"},{"name":"A","type":"integer"},{"name":"B","type":"string"},{"name":"C","type":"datetime"}],"primaryKey":["idx"],"pandas_version":"0.20.0"}, "data": [{"idx":0,"A":1,"B":"a","C":"2016-01-01T00:00:00.000Z"},{"idx":1,"A":2,"B":"b","C":"2016-01-02T00:00:00.000Z"},{"idx":2,"A":3,"B":"c","C":"2016-01-03T00:00:00.000Z"}]}'

See IO: Table Schema for more information.

Additionally, the repr for DataFrame and Series can now publish this JSON Table schema representation of the Series or DataFrame if you are using IPython (or another frontend like nteract using the Jupyter messaging protocol). This gives frontends like the Jupyter notebook and nteract more flexiblity in how they display pandas objects, since they have more information about the data. You must enable this by setting the display.html.table_schema option to True.

SciPy sparse matrix from/to SparseDataFrame

Pandas now supports creating sparse dataframes directly from scipy.sparse.spmatrix instances. See the documentation for more information. (GH4343)

All sparse formats are supported, but matrices that are not in COOrdinate format will be converted, copying data as needed.

In [45]: from scipy.sparse import csr_matrix

In [46]: arr = np.random.random(size=(1000, 5))

In [47]: arr[arr < .9] = 0

In [48]: sp_arr = csr_matrix(arr)

In [49]: sp_arr
Out[49]: 
<1000x5 sparse matrix of type '<class 'numpy.float64'>'
	with 521 stored elements in Compressed Sparse Row format>

In [50]: sdf = pd.SparseDataFrame(sp_arr)

In [51]: sdf
Out[51]: 
      0         1        2         3         4
0   NaN       NaN      NaN       NaN       NaN
1   NaN       NaN      NaN  0.955103       NaN
2   NaN       NaN      NaN  0.900469       NaN
3   NaN       NaN      NaN       NaN       NaN
4   NaN  0.924771      NaN       NaN       NaN
5   NaN       NaN      NaN       NaN       NaN
6   NaN       NaN      NaN       NaN       NaN
..   ..       ...      ...       ...       ...
993 NaN       NaN      NaN       NaN       NaN
994 NaN       NaN      NaN       NaN  0.972191
995 NaN  0.979898  0.97901       NaN       NaN
996 NaN       NaN      NaN       NaN       NaN
997 NaN       NaN      NaN       NaN       NaN
998 NaN       NaN      NaN       NaN       NaN
999 NaN       NaN      NaN       NaN       NaN

[1000 rows x 5 columns]

To convert a SparseDataFrame back to sparse SciPy matrix in COO format, you can use:

In [52]: sdf.to_coo()
Out[52]: 
<1000x5 sparse matrix of type '<class 'numpy.float64'>'
	with 521 stored elements in COOrdinate format>

Excel output for styled DataFrames

Experimental support has been added to export DataFrame.style formats to Excel using the openpyxl engine. (GH15530)

For example, after running the following, styled.xlsx renders as below:

In [53]: np.random.seed(24)

In [54]: df = pd.DataFrame({'A': np.linspace(1, 10, 10)})

In [55]: df = pd.concat([df, pd.DataFrame(np.random.RandomState(24).randn(10, 4),
   ....:                                  columns=list('BCDE'))],
   ....:                axis=1)
   ....: 

In [56]: df.iloc[0, 2] = np.nan

In [57]: df
Out[57]: 
      A         B         C         D         E
0   1.0  1.329212       NaN -0.316280 -0.990810
1   2.0 -1.070816 -1.438713  0.564417  0.295722
2   3.0 -1.626404  0.219565  0.678805  1.889273
3   4.0  0.961538  0.104011 -0.481165  0.850229
4   5.0  1.453425  1.057737  0.165562  0.515018
5   6.0 -1.336936  0.562861  1.392855 -0.063328
6   7.0  0.121668  1.207603 -0.002040  1.627796
7   8.0  0.354493  1.037528 -0.385684  0.519818
8   9.0  1.686583 -1.325963  1.428984 -2.089354
9  10.0 -0.129820  0.631523 -0.586538  0.290720

In [58]: styled = df.style.\
   ....:     applymap(lambda val: 'color: %s' % 'red' if val < 0 else 'black').\
   ....:     highlight_max()
   ....: 

In [59]: styled.to_excel('styled.xlsx', engine='openpyxl')
_images/style-excel.png

See the Style documentation for more detail.

IntervalIndex

pandas has gained an IntervalIndex with its own dtype, interval as well as the Interval scalar type. These allow first-class support for interval notation, specifically as a return type for the categories in cut() and qcut(). The IntervalIndex allows some unique indexing, see the docs. (GH7640, GH8625)

Warning

These indexing behaviors of the IntervalIndex are provisional and may change in a future version of pandas. Feedback on usage is welcome.

Previous behavior:

The returned categories were strings, representing Intervals

In [1]: c = pd.cut(range(4), bins=2)

In [2]: c
Out[2]:
[(-0.003, 1.5], (-0.003, 1.5], (1.5, 3], (1.5, 3]]
Categories (2, object): [(-0.003, 1.5] < (1.5, 3]]

In [3]: c.categories
Out[3]: Index(['(-0.003, 1.5]', '(1.5, 3]'], dtype='object')

New behavior:

In [60]: c = pd.cut(range(4), bins=2)

In [61]: c
Out[61]: 
[(-0.003, 1.5], (-0.003, 1.5], (1.5, 3.0], (1.5, 3.0]]
Categories (2, interval[float64]): [(-0.003, 1.5] < (1.5, 3.0]]

In [62]: c.categories
Out[62]: 
IntervalIndex([(-0.003, 1.5], (1.5, 3.0]]
              closed='right',
              dtype='interval[float64]')

Furthermore, this allows one to bin other data with these same bins, with NaN representing a missing value similar to other dtypes.

In [63]: pd.cut([0, 3, 5, 1], bins=c.categories)
Out[63]: 
[(-0.003, 1.5], (1.5, 3.0], NaN, (-0.003, 1.5]]
Categories (2, interval[float64]): [(-0.003, 1.5] < (1.5, 3.0]]

An IntervalIndex can also be used in Series and DataFrame as the index.

In [64]: df = pd.DataFrame({'A': range(4),
   ....:                    'B': pd.cut([0, 3, 1, 1], bins=c.categories)}
   ....:                  ).set_index('B')
   ....: 

In [65]: df
Out[65]: 
               A
B               
(-0.003, 1.5]  0
(1.5, 3.0]     1
(-0.003, 1.5]  2
(-0.003, 1.5]  3

Selecting via a specific interval:

In [66]: df.loc[pd.Interval(1.5, 3.0)]
Out[66]: 
A    1
Name: (1.5, 3.0], dtype: int64

Selecting via a scalar value that is contained in the intervals.

In [67]: df.loc[0]
Out[67]: 
               A
B               
(-0.003, 1.5]  0
(-0.003, 1.5]  2
(-0.003, 1.5]  3

Other Enhancements

  • DataFrame.rolling() now accepts the parameter closed='right'|'left'|'both'|'neither' to choose the rolling window-endpoint closedness. See the documentation (GH13965)
  • Integration with the feather-format, including a new top-level pd.read_feather() and DataFrame.to_feather() method, see here.
  • Series.str.replace() now accepts a callable, as replacement, which is passed to re.sub (GH15055)
  • Series.str.replace() now accepts a compiled regular expression as a pattern (GH15446)
  • Series.sort_index accepts parameters kind and na_position (GH13589, GH14444)
  • DataFrame and DataFrame.groupby() have gained a nunique() method to count the distinct values over an axis (GH14336, GH15197).
  • DataFrame has gained a melt() method, equivalent to pd.melt(), for unpivoting from a wide to long format (GH12640).
  • pd.read_excel() now preserves sheet order when using sheetname=None (GH9930)
  • Multiple offset aliases with decimal points are now supported (e.g. 0.5min is parsed as 30s) (GH8419)
  • .isnull() and .notnull() have been added to Index object to make them more consistent with the Series API (GH15300)
  • New UnsortedIndexError (subclass of KeyError) raised when indexing/slicing into an unsorted MultiIndex (GH11897). This allows differentiation between errors due to lack of sorting or an incorrect key. See here
  • MultiIndex has gained a .to_frame() method to convert to a DataFrame (GH12397)
  • pd.cut and pd.qcut now support datetime64 and timedelta64 dtypes (GH14714, GH14798)
  • pd.qcut has gained the duplicates='raise'|'drop' option to control whether to raise on duplicated edges (GH7751)
  • Series provides a to_excel method to output Excel files (GH8825)
  • The usecols argument in pd.read_csv() now accepts a callable function as a value (GH14154)
  • The skiprows argument in pd.read_csv() now accepts a callable function as a value (GH10882)
  • The nrows and chunksize arguments in pd.read_csv() are supported if both are passed (GH6774, GH15755)
  • DataFrame.plot now prints a title above each subplot if suplots=True and title is a list of strings (GH14753)
  • DataFrame.plot can pass the matplotlib 2.0 default color cycle as a single string as color parameter, see here. (GH15516)
  • Series.interpolate() now supports timedelta as an index type with method='time' (GH6424)
  • Addition of a level keyword to DataFrame/Series.rename to rename labels in the specified level of a MultiIndex (GH4160).
  • DataFrame.reset_index() will now interpret a tuple index.name as a key spanning across levels of columns, if this is a MultiIndex (GH16164)
  • Timedelta.isoformat method added for formatting Timedeltas as an ISO 8601 duration. See the Timedelta docs (GH15136)
  • .select_dtypes() now allows the string datetimetz to generically select datetimes with tz (GH14910)
  • The .to_latex() method will now accept multicolumn and multirow arguments to use the accompanying LaTeX enhancements
  • pd.merge_asof() gained the option direction='backward'|'forward'|'nearest' (GH14887)
  • Series/DataFrame.asfreq() have gained a fill_value parameter, to fill missing values (GH3715).
  • Series/DataFrame.resample.asfreq have gained a fill_value parameter, to fill missing values during resampling (GH3715).
  • pandas.util.hash_pandas_object() has gained the ability to hash a MultiIndex (GH15224)
  • Series/DataFrame.squeeze() have gained the axis parameter. (GH15339)
  • DataFrame.to_excel() has a new freeze_panes parameter to turn on Freeze Panes when exporting to Excel (GH15160)
  • pd.read_html() will parse multiple header rows, creating a MutliIndex header. (GH13434).
  • HTML table output skips colspan or rowspan attribute if equal to 1. (GH15403)
  • pandas.io.formats.style.Styler template now has blocks for easier extension, see the example notebook (GH15649)
  • Styler.render() now accepts **kwargs to allow user-defined variables in the template (GH15649)
  • Compatibility with Jupyter notebook 5.0; MultiIndex column labels are left-aligned and MultiIndex row-labels are top-aligned (GH15379)
  • TimedeltaIndex now has a custom date-tick formatter specifically designed for nanosecond level precision (GH8711)
  • pd.api.types.union_categoricals gained the ignore_ordered argument to allow ignoring the ordered attribute of unioned categoricals (GH13410). See the categorical union docs for more information.
  • DataFrame.to_latex() and DataFrame.to_string() now allow optional header aliases. (GH15536)
  • Re-enable the parse_dates keyword of pd.read_excel() to parse string columns as dates (GH14326)
  • Added .empty property to subclasses of Index. (GH15270)
  • Enabled floor division for Timedelta and TimedeltaIndex (GH15828)
  • pandas.io.json.json_normalize() gained the option errors='ignore'|'raise'; the default is errors='raise' which is backward compatible. (GH14583)
  • pandas.io.json.json_normalize() with an empty list will return an empty DataFrame (GH15534)
  • pandas.io.json.json_normalize() has gained a sep option that accepts str to separate joined fields; the default is “.”, which is backward compatible. (GH14883)
  • MultiIndex.remove_unused_levels() has been added to facilitate removing unused levels. (GH15694)
  • pd.read_csv() will now raise a ParserError error whenever any parsing error occurs (GH15913, GH15925)
  • pd.read_csv() now supports the error_bad_lines and warn_bad_lines arguments for the Python parser (GH15925)
  • The display.show_dimensions option can now also be used to specify whether the length of a Series should be shown in its repr (GH7117).
  • parallel_coordinates() has gained a sort_labels keyword argument that sorts class labels and the colors assigned to them (GH15908)
  • Options added to allow one to turn on/off using bottleneck and numexpr, see here (GH16157)
  • DataFrame.style.bar() now accepts two more options to further customize the bar chart. Bar alignment is set with align='left'|'mid'|'zero', the default is “left”, which is backward compatible; You can now pass a list of color=[color_negative, color_positive]. (GH14757)

Backwards incompatible API changes

Possible incompatibility for HDF5 formats created with pandas < 0.13.0

pd.TimeSeries was deprecated officially in 0.17.0, though has already been an alias since 0.13.0. It has been dropped in favor of pd.Series. (GH15098).

This may cause HDF5 files that were created in prior versions to become unreadable if pd.TimeSeries was used. This is most likely to be for pandas < 0.13.0. If you find yourself in this situation. You can use a recent prior version of pandas to read in your HDF5 files, then write them out again after applying the procedure below.

In [2]: s = pd.TimeSeries([1,2,3], index=pd.date_range('20130101', periods=3))

In [3]: s
Out[3]:
2013-01-01    1
2013-01-02    2
2013-01-03    3
Freq: D, dtype: int64

In [4]: type(s)
Out[4]: pandas.core.series.TimeSeries

In [5]: s = pd.Series(s)

In [6]: s
Out[6]:
2013-01-01    1
2013-01-02    2
2013-01-03    3
Freq: D, dtype: int64

In [7]: type(s)
Out[7]: pandas.core.series.Series

Map on Index types now return other Index types

map on an Index now returns an Index, not a numpy array (GH12766)

In [68]: idx = Index([1, 2])

In [69]: idx
Out[69]: Int64Index([1, 2], dtype='int64')

In [70]: mi = MultiIndex.from_tuples([(1, 2), (2, 4)])

In [71]: mi
Out[71]: 
MultiIndex(levels=[[1, 2], [2, 4]],
           labels=[[0, 1], [0, 1]])

Previous Behavior:

In [5]: idx.map(lambda x: x * 2)
Out[5]: array([2, 4])

In [6]: idx.map(lambda x: (x, x * 2))
Out[6]: array([(1, 2), (2, 4)], dtype=object)

In [7]: mi.map(lambda x: x)
Out[7]: array([(1, 2), (2, 4)], dtype=object)

In [8]: mi.map(lambda x: x[0])
Out[8]: array([1, 2])

New Behavior:

In [72]: idx.map(lambda x: x * 2)
Out[72]: Int64Index([2, 4], dtype='int64')

In [73]: idx.map(lambda x: (x, x * 2))
Out[73]: 
MultiIndex(levels=[[1, 2], [2, 4]],
           labels=[[0, 1], [0, 1]])

In [74]: mi.map(lambda x: x)
Out[74]: 
MultiIndex(levels=[[1, 2], [2, 4]],
           labels=[[0, 1], [0, 1]])

In [75]: mi.map(lambda x: x[0])
Out[75]: Int64Index([1, 2], dtype='int64')

map on a Series with datetime64 values may return int64 dtypes rather than int32

In [76]: s = Series(date_range('2011-01-02T00:00', '2011-01-02T02:00', freq='H').tz_localize('Asia/Tokyo'))

In [77]: s
Out[77]: 
0   2011-01-02 00:00:00+09:00
1   2011-01-02 01:00:00+09:00
2   2011-01-02 02:00:00+09:00
dtype: datetime64[ns, Asia/Tokyo]

Previous Behavior:

In [9]: s.map(lambda x: x.hour)
Out[9]:
0    0
1    1
2    2
dtype: int32

New Behavior:

In [78]: s.map(lambda x: x.hour)
Out[78]: 
0    0
1    1
2    2
dtype: int64

Accessing datetime fields of Index now return Index

The datetime-related attributes (see here for an overview) of DatetimeIndex, PeriodIndex and TimedeltaIndex previously returned numpy arrays. They will now return a new Index object, except in the case of a boolean field, where the result will still be a boolean ndarray. (GH15022)

Previous behaviour:

In [1]: idx = pd.date_range("2015-01-01", periods=5, freq='10H')

In [2]: idx.hour
Out[2]: array([ 0, 10, 20,  6, 16], dtype=int32)

New Behavior:

In [79]: idx = pd.date_range("2015-01-01", periods=5, freq='10H')

In [80]: idx.hour
Out[80]: Int64Index([0, 10, 20, 6, 16], dtype='int64')

This has the advantage that specific Index methods are still available on the result. On the other hand, this might have backward incompatibilities: e.g. compared to numpy arrays, Index objects are not mutable. To get the original ndarray, you can always convert explicitly using np.asarray(idx.hour).

pd.unique will now be consistent with extension types

In prior versions, using Series.unique() and pandas.unique() on Categorical and tz-aware data-types would yield different return types. These are now made consistent. (GH15903)

  • Datetime tz-aware

    Previous behaviour:

    # Series
    In [5]: pd.Series([pd.Timestamp('20160101', tz='US/Eastern'),
                       pd.Timestamp('20160101', tz='US/Eastern')]).unique()
    Out[5]: array([Timestamp('2016-01-01 00:00:00-0500', tz='US/Eastern')], dtype=object)
    
    In [6]: pd.unique(pd.Series([pd.Timestamp('20160101', tz='US/Eastern'),
                                 pd.Timestamp('20160101', tz='US/Eastern')]))
    Out[6]: array(['2016-01-01T05:00:00.000000000'], dtype='datetime64[ns]')
    
    # Index
    In [7]: pd.Index([pd.Timestamp('20160101', tz='US/Eastern'),
                      pd.Timestamp('20160101', tz='US/Eastern')]).unique()
    Out[7]: DatetimeIndex(['2016-01-01 00:00:00-05:00'], dtype='datetime64[ns, US/Eastern]', freq=None)
    
    In [8]: pd.unique([pd.Timestamp('20160101', tz='US/Eastern'),
                       pd.Timestamp('20160101', tz='US/Eastern')])
    Out[8]: array(['2016-01-01T05:00:00.000000000'], dtype='datetime64[ns]')
    

    New Behavior:

    # Series, returns an array of Timestamp tz-aware
    In [81]: pd.Series([pd.Timestamp('20160101', tz='US/Eastern'),
       ....:           pd.Timestamp('20160101', tz='US/Eastern')]).unique()
       ....: 
    Out[81]: array([Timestamp('2016-01-01 00:00:00-0500', tz='US/Eastern')], dtype=object)
    
    In [82]: pd.unique(pd.Series([pd.Timestamp('20160101', tz='US/Eastern'),
       ....:                      pd.Timestamp('20160101', tz='US/Eastern')]))
       ....: 
    Out[82]: array([Timestamp('2016-01-01 00:00:00-0500', tz='US/Eastern')], dtype=object)
    
    # Index, returns a DatetimeIndex
    In [83]: pd.Index([pd.Timestamp('20160101', tz='US/Eastern'),
       ....:           pd.Timestamp('20160101', tz='US/Eastern')]).unique()
       ....: 
    Out[83]: DatetimeIndex(['2016-01-01 05:00:00-05:00'], dtype='datetime64[ns, US/Eastern]', freq=None)
    
    In [84]: pd.unique(pd.Index([pd.Timestamp('20160101', tz='US/Eastern'),
       ....:                     pd.Timestamp('20160101', tz='US/Eastern')]))
       ....: 
    Out[84]: DatetimeIndex(['2016-01-01 00:00:00-05:00'], dtype='datetime64[ns, US/Eastern]', freq=None)
    
  • Categoricals

    Previous behaviour:

    In [1]: pd.Series(list('baabc'), dtype='category').unique()
    Out[1]:
    [b, a, c]
    Categories (3, object): [b, a, c]
    
    In [2]: pd.unique(pd.Series(list('baabc'), dtype='category'))
    Out[2]: array(['b', 'a', 'c'], dtype=object)
    

    New Behavior:

    # returns a Categorical
    In [85]: pd.Series(list('baabc'), dtype='category').unique()
    Out[85]: 
    [b, a, c]
    Categories (3, object): [b, a, c]
    
    In [86]: pd.unique(pd.Series(list('baabc'), dtype='category'))
    Out[86]: 
    [b, a, c]
    Categories (3, object): [b, a, c]
    

S3 File Handling

pandas now uses s3fs for handling S3 connections. This shouldn’t break any code. However, since s3fs is not a required dependency, you will need to install it separately, like boto in prior versions of pandas. (GH11915).

Partial String Indexing Changes

DatetimeIndex Partial String Indexing now works as an exact match, provided that string resolution coincides with index resolution, including a case when both are seconds (GH14826). See Slice vs. Exact Match for details.

In [87]: df = DataFrame({'a': [1, 2, 3]}, DatetimeIndex(['2011-12-31 23:59:59',
   ....:                                                 '2012-01-01 00:00:00',
   ....:                                                 '2012-01-01 00:00:01']))
   ....: 

Previous Behavior:

In [4]: df['2011-12-31 23:59:59']
Out[4]:
                       a
2011-12-31 23:59:59  1

In [5]: df['a']['2011-12-31 23:59:59']
Out[5]:
2011-12-31 23:59:59    1
Name: a, dtype: int64

New Behavior:

In [4]: df['2011-12-31 23:59:59']
KeyError: '2011-12-31 23:59:59'

In [5]: df['a']['2011-12-31 23:59:59']
Out[5]: 1

Concat of different float dtypes will not automatically upcast

Previously, concat of multiple objects with different float dtypes would automatically upcast results to a dtype of float64. Now the smallest acceptable dtype will be used (GH13247)

In [88]: df1 = pd.DataFrame(np.array([1.0], dtype=np.float32, ndmin=2))

In [89]: df1.dtypes
Out[89]: 
0    float32
dtype: object

In [90]: df2 = pd.DataFrame(np.array([np.nan], dtype=np.float32, ndmin=2))

In [91]: df2.dtypes
Out[91]: 
0    float32
dtype: object

Previous Behavior:

In [7]: pd.concat([df1, df2]).dtypes
Out[7]:
0    float64
dtype: object

New Behavior:

In [92]: pd.concat([df1, df2]).dtypes
Out[92]: 
0    float32
dtype: object

Pandas Google BigQuery support has moved

pandas has split off Google BigQuery support into a separate package pandas-gbq. You can conda install pandas-gbq -c conda-forge or pip install pandas-gbq to get it. The functionality of read_gbq() and DataFrame.to_gbq() remain the same with the currently released version of pandas-gbq=0.1.4. Documentation is now hosted here (GH15347)

Memory Usage for Index is more Accurate

In previous versions, showing .memory_usage() on a pandas structure that has an index, would only include actual index values and not include structures that facilitated fast indexing. This will generally be different for Index and MultiIndex and less-so for other index types. (GH15237)

Previous Behavior:

In [8]: index = Index(['foo', 'bar', 'baz'])

In [9]: index.memory_usage(deep=True)
Out[9]: 180

In [10]: index.get_loc('foo')
Out[10]: 0

In [11]: index.memory_usage(deep=True)
Out[11]: 180

New Behavior:

In [8]: index = Index(['foo', 'bar', 'baz'])

In [9]: index.memory_usage(deep=True)
Out[9]: 180

In [10]: index.get_loc('foo')
Out[10]: 0

In [11]: index.memory_usage(deep=True)
Out[11]: 260

DataFrame.sort_index changes

In certain cases, calling .sort_index() on a MultiIndexed DataFrame would return the same DataFrame without seeming to sort. This would happen with a lexsorted, but non-monotonic levels. (GH15622, GH15687, GH14015, GH13431, GH15797)

This is unchanged from prior versions, but shown for illustration purposes:

In [93]: df = DataFrame(np.arange(6), columns=['value'], index=MultiIndex.from_product([list('BA'), range(3)]))

In [94]: df
Out[94]: 
     value
B 0      0
  1      1
  2      2
A 0      3
  1      4
  2      5
In [95]: df.index.is_lexsorted()
Out[95]: False

In [96]: df.index.is_monotonic
Out[96]: False

Sorting works as expected

In [97]: df.sort_index()
Out[97]: 
     value
A 0      3
  1      4
  2      5
B 0      0
  1      1
  2      2
In [98]: df.sort_index().index.is_lexsorted()
Out[98]: True

In [99]: df.sort_index().index.is_monotonic
Out[99]: True

However, this example, which has a non-monotonic 2nd level, doesn’t behave as desired.

In [100]: df = pd.DataFrame(
   .....:         {'value': [1, 2, 3, 4]},
   .....:          index=pd.MultiIndex(levels=[['a', 'b'], ['bb', 'aa']],
   .....:                              labels=[[0, 0, 1, 1], [0, 1, 0, 1]]))
   .....: 

In [101]: df
Out[101]: 
      value
a bb      1
  aa      2
b bb      3
  aa      4

Previous Behavior:

In [11]: df.sort_index()
Out[11]:
      value
a bb      1
  aa      2
b bb      3
  aa      4

In [14]: df.sort_index().index.is_lexsorted()
Out[14]: True

In [15]: df.sort_index().index.is_monotonic
Out[15]: False

New Behavior:

In [102]: df.sort_index()
Out[102]: 
      value
a aa      2
  bb      1
b aa      4
  bb      3

In [103]: df.sort_index().index.is_lexsorted()
Out[103]: True

In [104]: df.sort_index().index.is_monotonic
Out[104]: True

Groupby Describe Formatting

The output formatting of groupby.describe() now labels the describe() metrics in the columns instead of the index. This format is consistent with groupby.agg() when applying multiple functions at once. (GH4792)

Previous Behavior:

In [1]: df = pd.DataFrame({'A': [1, 1, 2, 2], 'B': [1, 2, 3, 4]})

In [2]: df.groupby('A').describe()
Out[2]:
                B
A
1 count  2.000000
  mean   1.500000
  std    0.707107
  min    1.000000
  25%    1.250000
  50%    1.500000
  75%    1.750000
  max    2.000000
2 count  2.000000
  mean   3.500000
  std    0.707107
  min    3.000000
  25%    3.250000
  50%    3.500000
  75%    3.750000
  max    4.000000

In [3]: df.groupby('A').agg([np.mean, np.std, np.min, np.max])
Out[3]:
     B
  mean       std amin amax
A
1  1.5  0.707107    1    2
2  3.5  0.707107    3    4

New Behavior:

In [105]: df = pd.DataFrame({'A': [1, 1, 2, 2], 'B': [1, 2, 3, 4]})

In [106]: df.groupby('A').describe()
Out[106]: 
      B                                          
  count mean       std  min   25%  50%   75%  max
A                                                
1   2.0  1.5  0.707107  1.0  1.25  1.5  1.75  2.0
2   2.0  3.5  0.707107  3.0  3.25  3.5  3.75  4.0

In [107]: df.groupby('A').agg([np.mean, np.std, np.min, np.max])
Out[107]: 
     B                    
  mean       std amin amax
A                         
1  1.5  0.707107    1    2
2  3.5  0.707107    3    4

Window Binary Corr/Cov operations return a MultiIndex DataFrame

A binary window operation, like .corr() or .cov(), when operating on a .rolling(..), .expanding(..), or .ewm(..) object, will now return a 2-level MultiIndexed DataFrame rather than a Panel, as Panel is now deprecated, see here. These are equivalent in function, but a MultiIndexed DataFrame enjoys more support in pandas. See the section on Windowed Binary Operations for more information. (GH15677)

In [108]: np.random.seed(1234)

In [109]: df = pd.DataFrame(np.random.rand(100, 2),
   .....:                   columns=pd.Index(['A', 'B'], name='bar'),
   .....:                   index=pd.date_range('20160101',
   .....:                                       periods=100, freq='D', name='foo'))
   .....: 

In [110]: df.tail()
Out[110]: 
bar                A         B
foo                           
2016-04-05  0.640880  0.126205
2016-04-06  0.171465  0.737086
2016-04-07  0.127029  0.369650
2016-04-08  0.604334  0.103104
2016-04-09  0.802374  0.945553

Previous Behavior:

In [2]: df.rolling(12).corr()
Out[2]:
<class 'pandas.core.panel.Panel'>
Dimensions: 100 (items) x 2 (major_axis) x 2 (minor_axis)
Items axis: 2016-01-01 00:00:00 to 2016-04-09 00:00:00
Major_axis axis: A to B
Minor_axis axis: A to B

New Behavior:

In [111]: res = df.rolling(12).corr()

In [112]: res.tail()
Out[112]: 
bar                    A         B
foo        bar                    
2016-04-07 B   -0.132090  1.000000
2016-04-08 A    1.000000 -0.145775
           B   -0.145775  1.000000
2016-04-09 A    1.000000  0.119645
           B    0.119645  1.000000

Retrieving a correlation matrix for a cross-section

In [113]: df.rolling(12).corr().loc['2016-04-07']
Out[113]: 
bar                   A        B
foo        bar                  
2016-04-07 A    1.00000 -0.13209
           B   -0.13209  1.00000

HDFStore where string comparison

In previous versions most types could be compared to string column in a HDFStore usually resulting in an invalid comparison, returning an empty result frame. These comparisons will now raise a TypeError (GH15492)

In [114]: df = pd.DataFrame({'unparsed_date': ['2014-01-01', '2014-01-01']})

In [115]: df.to_hdf('store.h5', 'key', format='table', data_columns=True)

In [116]: df.dtypes
Out[116]: 
unparsed_date    object
dtype: object

Previous Behavior:

In [4]: pd.read_hdf('store.h5', 'key', where='unparsed_date > ts')
File "<string>", line 1
  (unparsed_date > 1970-01-01 00:00:01.388552400)
                        ^
SyntaxError: invalid token

New Behavior:

In [18]: ts = pd.Timestamp('2014-01-01')

In [19]: pd.read_hdf('store.h5', 'key', where='unparsed_date > ts')
TypeError: Cannot compare 2014-01-01 00:00:00 of
type <class 'pandas.tslib.Timestamp'> to string column

Index.intersection and inner join now preserve the order of the left Index

Index.intersection() now preserves the order of the calling Index (left) instead of the other Index (right) (GH15582). This affects inner joins, DataFrame.join() and merge(), and the .align method.

  • Index.intersection

    In [117]: left = pd.Index([2, 1, 0])
    
    In [118]: left
    Out[118]: Int64Index([2, 1, 0], dtype='int64')
    
    In [119]: right = pd.Index([1, 2, 3])
    
    In [120]: right
    Out[120]: Int64Index([1, 2, 3], dtype='int64')
    

    Previous Behavior:

    In [4]: left.intersection(right)
    Out[4]: Int64Index([1, 2], dtype='int64')
    

    New Behavior:

    In [121]: left.intersection(right)
    Out[121]: Int64Index([2, 1], dtype='int64')
    
  • DataFrame.join and pd.merge

    In [122]: left = pd.DataFrame({'a': [20, 10, 0]}, index=[2, 1, 0])
    
    In [123]: left
    Out[123]: 
        a
    2  20
    1  10
    0   0
    
    In [124]: right = pd.DataFrame({'b': [100, 200, 300]}, index=[1, 2, 3])
    
    In [125]: right
    Out[125]: 
         b
    1  100
    2  200
    3  300
    

    Previous Behavior:

    In [4]: left.join(right, how='inner')
    Out[4]:
        a    b
    1  10  100
    2  20  200
    

    New Behavior:

    In [126]: left.join(right, how='inner')
    Out[126]: 
        a    b
    2  20  200
    1  10  100
    

Pivot Table always returns a DataFrame

The documentation for pivot_table() states that a DataFrame is always returned. Here a bug is fixed that allowed this to return a Series under certain circumstance. (GH4386)

In [127]: df = DataFrame({'col1': [3, 4, 5],
   .....:                 'col2': ['C', 'D', 'E'],
   .....:                 'col3': [1, 3, 9]})
   .....: 

In [128]: df
Out[128]: 
   col1 col2  col3
0     3    C     1
1     4    D     3
2     5    E     9

Previous Behavior:

In [2]: df.pivot_table('col1', index=['col3', 'col2'], aggfunc=np.sum)
Out[2]:
col3  col2
1     C       3
3     D       4
9     E       5
Name: col1, dtype: int64

New Behavior:

In [129]: df.pivot_table('col1', index=['col3', 'col2'], aggfunc=np.sum)
Out[129]: 
           col1
col3 col2      
1    C        3
3    D        4
9    E        5

Other API Changes

  • numexpr version is now required to be >= 2.4.6 and it will not be used at all if this requisite is not fulfilled (GH15213).
  • CParserError has been renamed to ParserError in pd.read_csv() and will be removed in the future (GH12665)
  • SparseArray.cumsum() and SparseSeries.cumsum() will now always return SparseArray and SparseSeries respectively (GH12855)
  • DataFrame.applymap() with an empty DataFrame will return a copy of the empty DataFrame instead of a Series (GH8222)
  • Series.map() now respects default values of dictionary subclasses with a __missing__ method, such as collections.Counter (GH15999)
  • .loc has compat with .ix for accepting iterators, and NamedTuples (GH15120)
  • interpolate() and fillna() will raise a ValueError if the limit keyword argument is not greater than 0. (GH9217)
  • pd.read_csv() will now issue a ParserWarning whenever there are conflicting values provided by the dialect parameter and the user (GH14898)
  • pd.read_csv() will now raise a ValueError for the C engine if the quote character is larger than than one byte (GH11592)
  • inplace arguments now require a boolean value, else a ValueError is thrown (GH14189)
  • pandas.api.types.is_datetime64_ns_dtype will now report True on a tz-aware dtype, similar to pandas.api.types.is_datetime64_any_dtype
  • DataFrame.asof() will return a null filled Series instead the scalar NaN if a match is not found (GH15118)
  • Specific support for copy.copy() and copy.deepcopy() functions on NDFrame objects (GH15444)
  • Series.sort_values() accepts a one element list of bool for consistency with the behavior of DataFrame.sort_values() (GH15604)
  • .merge() and .join() on category dtype columns will now preserve the category dtype when possible (GH10409)
  • SparseDataFrame.default_fill_value will be 0, previously was nan in the return from pd.get_dummies(..., sparse=True) (GH15594)
  • The default behaviour of Series.str.match has changed from extracting groups to matching the pattern. The extracting behaviour was deprecated since pandas version 0.13.0 and can be done with the Series.str.extract method (GH5224). As a consequence, the as_indexer keyword is ignored (no longer needed to specify the new behaviour) and is deprecated.
  • NaT will now correctly report False for datetimelike boolean operations such as is_month_start (GH15781)
  • NaT will now correctly return np.nan for Timedelta and Period accessors such as days and quarter (GH15782)
  • NaT will now returns NaT for tz_localize and tz_convert methods (GH15830)
  • DataFrame and Panel constructors with invalid input will now raise ValueError rather than PandasError, if called with scalar inputs and not axes (GH15541)
  • DataFrame and Panel constructors with invalid input will now raise ValueError rather than pandas.core.common.PandasError, if called with scalar inputs and not axes; The exception PandasError is removed as well. (GH15541)
  • The exception pandas.core.common.AmbiguousIndexError is removed as it is not referenced (GH15541)

Reorganization of the library: Privacy Changes

Modules Privacy Has Changed

Some formerly public python/c/c++/cython extension modules have been moved and/or renamed. These are all removed from the public API. Furthermore, the pandas.core, pandas.compat, and pandas.util top-level modules are now considered to be PRIVATE. If indicated, a deprecation warning will be issued if you reference theses modules. (GH12588)

Previous Location New Location Deprecated
pandas.lib pandas._libs.lib X
pandas.tslib pandas._libs.tslib X
pandas.computation pandas.core.computation X
pandas.msgpack pandas.io.msgpack  
pandas.index pandas._libs.index  
pandas.algos pandas._libs.algos  
pandas.hashtable pandas._libs.hashtable  
pandas.indexes pandas.core.indexes  
pandas.json pandas._libs.json / pandas.io.json X
pandas.parser pandas._libs.parsers X
pandas.formats pandas.io.formats  
pandas.sparse pandas.core.sparse  
pandas.tools pandas.core.reshape X
pandas.types pandas.core.dtypes X
pandas.io.sas.saslib pandas.io.sas._sas  
pandas._join pandas._libs.join  
pandas._hash pandas._libs.hashing  
pandas._period pandas._libs.period  
pandas._sparse pandas._libs.sparse  
pandas._testing pandas._libs.testing  
pandas._window pandas._libs.window  

Some new subpackages are created with public functionality that is not directly exposed in the top-level namespace: pandas.errors, pandas.plotting and pandas.testing (more details below). Together with pandas.api.types and certain functions in the pandas.io and pandas.tseries submodules, these are now the public subpackages.

Further changes:

  • The function union_categoricals() is now importable from pandas.api.types, formerly from pandas.types.concat (GH15998)
  • The type import pandas.tslib.NaTType is deprecated and can be replaced by using type(pandas.NaT) (GH16146)
  • The public functions in pandas.tools.hashing deprecated from that locations, but are now importable from pandas.util (GH16223)
  • The modules in pandas.util: decorators, print_versions, doctools, validators, depr_module are now private. Only the functions exposed in pandas.util itself are public (GH16223)

pandas.errors

We are adding a standard public module for all pandas exceptions & warnings pandas.errors. (GH14800). Previously these exceptions & warnings could be imported from pandas.core.common or pandas.io.common. These exceptions and warnings will be removed from the *.common locations in a future release. (GH15541)

The following are now part of this API:

['DtypeWarning',
 'EmptyDataError',
 'OutOfBoundsDatetime',
 'ParserError',
 'ParserWarning',
 'PerformanceWarning',
 'UnsortedIndexError',
 'UnsupportedFunctionCall']

pandas.testing

We are adding a standard module that exposes the public testing functions in pandas.testing (GH9895). Those functions can be used when writing tests for functionality using pandas objects.

The following testing functions are now part of this API:

pandas.plotting

A new public pandas.plotting module has been added that holds plotting functionality that was previously in either pandas.tools.plotting or in the top-level namespace. See the deprecations sections for more details.

Other Development Changes

  • Building pandas for development now requires cython >= 0.23 (GH14831)
  • Require at least 0.23 version of cython to avoid problems with character encodings (GH14699)
  • Switched the test framework to use pytest (GH13097)
  • Reorganization of tests directory layout (GH14854, GH15707).

Deprecations

Deprecate .ix

The .ix indexer is deprecated, in favor of the more strict .iloc and .loc indexers. .ix offers a lot of magic on the inference of what the user wants to do. To wit, .ix can decide to index positionally OR via labels, depending on the data type of the index. This has caused quite a bit of user confusion over the years. The full indexing documentation is here. (GH14218)

The recommended methods of indexing are:

  • .loc if you want to label index
  • .iloc if you want to positionally index.

Using .ix will now show a DeprecationWarning with a link to some examples of how to convert code here.

In [130]: df = pd.DataFrame({'A': [1, 2, 3],
   .....:                    'B': [4, 5, 6]},
   .....:                   index=list('abc'))
   .....: 

In [131]: df
Out[131]: 
   A  B
a  1  4
b  2  5
c  3  6

Previous Behavior, where you wish to get the 0th and the 2nd elements from the index in the ‘A’ column.

In [3]: df.ix[[0, 2], 'A']
Out[3]:
a    1
c    3
Name: A, dtype: int64

Using .loc. Here we will select the appropriate indexes from the index, then use label indexing.

In [132]: df.loc[df.index[[0, 2]], 'A']
Out[132]: 
a    1
c    3
Name: A, dtype: int64

Using .iloc. Here we will get the location of the ‘A’ column, then use positional indexing to select things.

In [133]: df.iloc[[0, 2], df.columns.get_loc('A')]
Out[133]: 
a    1
c    3
Name: A, dtype: int64

Deprecate Panel

Panel is deprecated and will be removed in a future version. The recommended way to represent 3-D data are with a MultiIndex on a DataFrame via the to_frame() or with the xarray package. Pandas provides a to_xarray() method to automate this conversion. For more details see Deprecate Panel documentation. (GH13563).

In [134]: p = tm.makePanel()

In [135]: p
Out[135]: 
<class 'pandas.core.panel.Panel'>
Dimensions: 3 (items) x 3 (major_axis) x 4 (minor_axis)
Items axis: ItemA to ItemC
Major_axis axis: 2000-01-03 00:00:00 to 2000-01-05 00:00:00
Minor_axis axis: A to D

Convert to a MultiIndex DataFrame

In [136]: p.to_frame()
Out[136]: 
                     ItemA     ItemB     ItemC
major      minor                              
2000-01-03 A      0.628776 -1.409432  0.209395
           B      0.988138 -1.347533 -0.896581
           C     -0.938153  1.272395 -0.161137
           D     -0.223019 -0.591863 -1.051539
2000-01-04 A      0.186494  1.422986 -0.592886
           B     -0.072608  0.363565  1.104352
           C     -1.239072 -1.449567  0.889157
           D      2.123692 -0.414505 -0.319561
2000-01-05 A      0.952478 -2.147855 -1.473116
           B     -0.550603 -0.014752 -0.431550
           C      0.139683 -1.195524  0.288377
           D      0.122273 -1.425795 -0.619993

Convert to an xarray DataArray

In [137]: p.to_xarray()
Out[137]: 
<xarray.DataArray (items: 3, major_axis: 3, minor_axis: 4)>
array([[[ 0.628776,  0.988138, -0.938153, -0.223019],
        [ 0.186494, -0.072608, -1.239072,  2.123692],
        [ 0.952478, -0.550603,  0.139683,  0.122273]],

       [[-1.409432, -1.347533,  1.272395, -0.591863],
        [ 1.422986,  0.363565, -1.449567, -0.414505],
        [-2.147855, -0.014752, -1.195524, -1.425795]],

       [[ 0.209395, -0.896581, -0.161137, -1.051539],
        [-0.592886,  1.104352,  0.889157, -0.319561],
        [-1.473116, -0.43155 ,  0.288377, -0.619993]]])
Coordinates:
  * items       (items) object 'ItemA' 'ItemB' 'ItemC'
  * major_axis  (major_axis) datetime64[ns] 2000-01-03 2000-01-04 2000-01-05
  * minor_axis  (minor_axis) object 'A' 'B' 'C' 'D'

Deprecate groupby.agg() with a dictionary when renaming

The .groupby(..).agg(..), .rolling(..).agg(..), and .resample(..).agg(..) syntax can accept a variable of inputs, including scalars, list, and a dict of column names to scalars or lists. This provides a useful syntax for constructing multiple (potentially different) aggregations.

However, .agg(..) can also accept a dict that allows ‘renaming’ of the result columns. This is a complicated and confusing syntax, as well as not consistent between Series and DataFrame. We are deprecating this ‘renaming’ functionaility.

  • We are deprecating passing a dict to a grouped/rolled/resampled Series. This allowed one to rename the resulting aggregation, but this had a completely different meaning than passing a dictionary to a grouped DataFrame, which accepts column-to-aggregations.
  • We are deprecating passing a dict-of-dicts to a grouped/rolled/resampled DataFrame in a similar manner.

This is an illustrative example:

In [138]: df = pd.DataFrame({'A': [1, 1, 1, 2, 2],
   .....:                    'B': range(5),
   .....:                    'C': range(5)})
   .....: 

In [139]: df
Out[139]: 
   A  B  C
0  1  0  0
1  1  1  1
2  1  2  2
3  2  3  3
4  2  4  4

Here is a typical useful syntax for computing different aggregations for different columns. This is a natural, and useful syntax. We aggregate from the dict-to-list by taking the specified columns and applying the list of functions. This returns a MultiIndex for the columns (this is not deprecated).

In [140]: df.groupby('A').agg({'B': 'sum', 'C': 'min'})
Out[140]: 
   B  C
A      
1  3  0
2  7  3

Here’s an example of the first deprecation, passing a dict to a grouped Series. This is a combination aggregation & renaming:

In [6]: df.groupby('A').B.agg({'foo': 'count'})
FutureWarning: using a dict on a Series for aggregation
is deprecated and will be removed in a future version

Out[6]:
   foo
A
1    3
2    2

You can accomplish the same operation, more idiomatically by:

In [141]: df.groupby('A').B.agg(['count']).rename(columns={'count': 'foo'})
Out[141]: 
   foo
A     
1    3
2    2

Here’s an example of the second deprecation, passing a dict-of-dict to a grouped DataFrame:

In [23]: (df.groupby('A')
            .agg({'B': {'foo': 'sum'}, 'C': {'bar': 'min'}})
         )
FutureWarning: using a dict with renaming is deprecated and
will be removed in a future version

Out[23]:
     B   C
   foo bar
A
1   3   0
2   7   3

You can accomplish nearly the same by:

In [142]: (df.groupby('A')
   .....:    .agg({'B': 'sum', 'C': 'min'})
   .....:    .rename(columns={'B': 'foo', 'C': 'bar'})
   .....: )
   .....: 
Out[142]: 
   foo  bar
A          
1    3    0
2    7    3

Deprecate .plotting

The pandas.tools.plotting module has been deprecated, in favor of the top level pandas.plotting module. All the public plotting functions are now available from pandas.plotting (GH12548).

Furthermore, the top-level pandas.scatter_matrix and pandas.plot_params are deprecated. Users can import these from pandas.plotting as well.

Previous script:

pd.tools.plotting.scatter_matrix(df)
pd.scatter_matrix(df)

Should be changed to:

pd.plotting.scatter_matrix(df)

Other Deprecations

  • SparseArray.to_dense() has deprecated the fill parameter, as that parameter was not being respected (GH14647)
  • SparseSeries.to_dense() has deprecated the sparse_only parameter (GH14647)
  • Series.repeat() has deprecated the reps parameter in favor of repeats (GH12662)
  • The Series constructor and .astype method have deprecated accepting timestamp dtypes without a frequency (e.g. np.datetime64) for the dtype parameter (GH15524)
  • Index.repeat() and MultiIndex.repeat() have deprecated the n parameter in favor of repeats (GH12662)
  • Categorical.searchsorted() and Series.searchsorted() have deprecated the v parameter in favor of value (GH12662)
  • TimedeltaIndex.searchsorted(), DatetimeIndex.searchsorted(), and PeriodIndex.searchsorted() have deprecated the key parameter in favor of value (GH12662)
  • DataFrame.astype() has deprecated the raise_on_error parameter in favor of errors (GH14878)
  • Series.sortlevel and DataFrame.sortlevel have been deprecated in favor of Series.sort_index and DataFrame.sort_index (GH15099)
  • importing concat from pandas.tools.merge has been deprecated in favor of imports from the pandas namespace. This should only affect explicit imports (GH15358)
  • Series/DataFrame/Panel.consolidate() been deprecated as a public method. (GH15483)
  • The as_indexer keyword of Series.str.match() has been deprecated (ignored keyword) (GH15257).
  • The following top-level pandas functions have been deprecated and will be removed in a future version (GH13790, GH15940)
    • pd.pnow(), replaced by Period.now()
    • pd.Term, is removed, as it is not applicable to user code. Instead use in-line string expressions in the where clause when searching in HDFStore
    • pd.Expr, is removed, as it is not applicable to user code.
    • pd.match(), is removed.
    • pd.groupby(), replaced by using the .groupby() method directly on a Series/DataFrame
    • pd.get_store(), replaced by a direct call to pd.HDFStore(...)
  • is_any_int_dtype, is_floating_dtype, and is_sequence are deprecated from pandas.api.types (GH16042)

Removal of prior version deprecations/changes

  • The pandas.rpy module is removed. Similar functionality can be accessed through the rpy2 project. See the R interfacing docs for more details.
  • The pandas.io.ga module with a google-analytics interface is removed (GH11308). Similar functionality can be found in the Google2Pandas package.
  • pd.to_datetime and pd.to_timedelta have dropped the coerce parameter in favor of errors (GH13602)
  • pandas.stats.fama_macbeth, pandas.stats.ols, pandas.stats.plm and pandas.stats.var, as well as the top-level pandas.fama_macbeth and pandas.ols routines are removed. Similar functionaility can be found in the statsmodels package. (GH11898)
  • The TimeSeries and SparseTimeSeries classes, aliases of Series and SparseSeries, are removed (GH10890, GH15098).
  • Series.is_time_series is dropped in favor of Series.index.is_all_dates (GH15098)
  • The deprecated irow, icol, iget and iget_value methods are removed in favor of iloc and iat as explained here (GH10711).
  • The deprecated DataFrame.iterkv() has been removed in favor of DataFrame.iteritems() (GH10711)
  • The Categorical constructor has dropped the name parameter (GH10632)
  • Categorical has dropped support for NaN categories (GH10748)
  • The take_last parameter has been dropped from duplicated(), drop_duplicates(), nlargest(), and nsmallest() methods (GH10236, GH10792, GH10920)
  • Series, Index, and DataFrame have dropped the sort and order methods (GH10726)
  • Where clauses in pytables are only accepted as strings and expressions types and not other data-types (GH12027)
  • DataFrame has dropped the combineAdd and combineMult methods in favor of add and mul respectively (GH10735)

Performance Improvements

  • Improved performance of pd.wide_to_long() (GH14779)
  • Improved performance of pd.factorize() by releasing the GIL with object dtype when inferred as strings (GH14859, GH16057)
  • Improved performance of timeseries plotting with an irregular DatetimeIndex (or with compat_x=True) (GH15073).
  • Improved performance of groupby().cummin() and groupby().cummax() (GH15048, GH15109, GH15561, GH15635)
  • Improved performance and reduced memory when indexing with a MultiIndex (GH15245)
  • When reading buffer object in read_sas() method without specified format, filepath string is inferred rather than buffer object. (GH14947)
  • Improved performance of .rank() for categorical data (GH15498)
  • Improved performance when using .unstack() (GH15503)
  • Improved performance of merge/join on category columns (GH10409)
  • Improved performance of drop_duplicates() on bool columns (GH12963)
  • Improve performance of pd.core.groupby.GroupBy.apply when the applied function used the .name attribute of the group DataFrame (GH15062).
  • Improved performance of iloc indexing with a list or array (GH15504).
  • Improved performance of Series.sort_index() with a monotonic index (GH15694)
  • Improved performance in pd.read_csv() on some platforms with buffered reads (GH16039)

Bug Fixes

Conversion

  • Bug in Timestamp.replace now raises TypeError when incorrect argument names are given; previously this raised ValueError (GH15240)
  • Bug in Timestamp.replace with compat for passing long integers (GH15030)
  • Bug in Timestamp returning UTC based time/date attributes when a timezone was provided (GH13303, GH6538)
  • Bug in Timestamp incorrectly localizing timezones during construction (GH11481, GH15777)
  • Bug in TimedeltaIndex addition where overflow was being allowed without error (GH14816)
  • Bug in TimedeltaIndex raising a ValueError when boolean indexing with loc (GH14946)
  • Bug in catching an overflow in Timestamp + Timedelta/Offset operations (GH15126)
  • Bug in DatetimeIndex.round() and Timestamp.round() floating point accuracy when rounding by milliseconds or less (GH14440, GH15578)
  • Bug in astype() where inf values were incorrectly converted to integers. Now raises error now with astype() for Series and DataFrames (GH14265)
  • Bug in DataFrame(..).apply(to_numeric) when values are of type decimal.Decimal. (GH14827)
  • Bug in describe() when passing a numpy array which does not contain the median to the percentiles keyword argument (GH14908)
  • Cleaned up PeriodIndex constructor, including raising on floats more consistently (GH13277)
  • Bug in using __deepcopy__ on empty NDFrame objects (GH15370)
  • Bug in .replace() may result in incorrect dtypes. (GH12747, GH15765)
  • Bug in Series.replace and DataFrame.replace which failed on empty replacement dicts (GH15289)
  • Bug in Series.replace which replaced a numeric by string (GH15743)
  • Bug in Index construction with NaN elements and integer dtype specified (GH15187)
  • Bug in Series construction with a datetimetz (GH14928)
  • Bug in Series.dt.round() inconsistent behaviour on NaT ‘s with different arguments (GH14940)
  • Bug in Series constructor when both copy=True and dtype arguments are provided (GH15125)
  • Incorrect dtyped Series was returned by comparison methods (e.g., lt, gt, …) against a constant for an empty DataFrame (GH15077)
  • Bug in Series.ffill() with mixed dtypes containing tz-aware datetimes. (GH14956)
  • Bug in DataFrame.fillna() where the argument downcast was ignored when fillna value was of type dict (GH15277)
  • Bug in .asfreq(), where frequency was not set for empty Series (GH14320)
  • Bug in DataFrame construction with nulls and datetimes in a list-like (GH15869)
  • Bug in DataFrame.fillna() with tz-aware datetimes (GH15855)
  • Bug in is_string_dtype, is_timedelta64_ns_dtype, and is_string_like_dtype in which an error was raised when None was passed in (GH15941)
  • Bug in the return type of pd.unique on a Categorical, which was returning an ndarray and not a Categorical (GH15903)
  • Bug in Index.to_series() where the index was not copied (and so mutating later would change the original), (GH15949)
  • Bug in indexing with partial string indexing with a len-1 DataFrame (GH16071)
  • Bug in Series construction where passing invalid dtype didn’t raise an error. (GH15520)

Indexing

  • Bug in Index power operations with reversed operands (GH14973)
  • Bug in DataFrame.sort_values() when sorting by multiple columns where one column is of type int64 and contains NaT (GH14922)
  • Bug in DataFrame.reindex() in which method was ignored when passing columns (GH14992)
  • Bug in DataFrame.loc with indexing a MultiIndex with a Series indexer (GH14730, GH15424)
  • Bug in DataFrame.loc with indexing a MultiIndex with a numpy array (GH15434)
  • Bug in Series.asof which raised if the series contained all np.nan (GH15713)
  • Bug in .at when selecting from a tz-aware column (GH15822)
  • Bug in Series.where() and DataFrame.where() where array-like conditionals were being rejected (GH15414)
  • Bug in Series.where() where TZ-aware data was converted to float representation (GH15701)
  • Bug in .loc that would not return the correct dtype for scalar access for a DataFrame (GH11617)
  • Bug in output formatting of a MultiIndex when names are integers (GH12223, GH15262)
  • Bug in Categorical.searchsorted() where alphabetical instead of the provided categorical order was used (GH14522)
  • Bug in Series.iloc where a Categorical object for list-like indexes input was returned, where a Series was expected. (GH14580)
  • Bug in DataFrame.isin comparing datetimelike to empty frame (GH15473)
  • Bug in .reset_index() when an all NaN level of a MultiIndex would fail (GH6322)
  • Bug in .reset_index() when raising error for index name already present in MultiIndex columns (GH16120)
  • Bug in creating a MultiIndex with tuples and not passing a list of names; this will now raise ValueError (GH15110)
  • Bug in the HTML display with with a MultiIndex and truncation (GH14882)
  • Bug in the display of .info() where a qualifier (+) would always be displayed with a MultiIndex that contains only non-strings (GH15245)
  • Bug in pd.concat() where the names of MultiIndex of resulting DataFrame are not handled correctly when None is presented in the names of MultiIndex of input DataFrame (GH15787)
  • Bug in DataFrame.sort_index() and Series.sort_index() where na_position doesn’t work with a MultiIndex (GH14784, GH16604)
  • Bug in in pd.concat() when combining objects with a CategoricalIndex (GH16111)
  • Bug in indexing with a scalar and a CategoricalIndex (GH16123)

I/O

  • Bug in pd.to_numeric() in which float and unsigned integer elements were being improperly casted (GH14941, GH15005)
  • Bug in pd.read_fwf() where the skiprows parameter was not being respected during column width inference (GH11256)
  • Bug in pd.read_csv() in which the dialect parameter was not being verified before processing (GH14898)
  • Bug in pd.read_csv() in which missing data was being improperly handled with usecols (GH6710)
  • Bug in pd.read_csv() in which a file containing a row with many columns followed by rows with fewer columns would cause a crash (GH14125)
  • Bug in pd.read_csv() for the C engine where usecols were being indexed incorrectly with parse_dates (GH14792)
  • Bug in pd.read_csv() with parse_dates when multiline headers are specified (GH15376)
  • Bug in pd.read_csv() with float_precision='round_trip' which caused a segfault when a text entry is parsed (GH15140)
  • Bug in pd.read_csv() when an index was specified and no values were specified as null values (GH15835)
  • Bug in pd.read_csv() in which certain invalid file objects caused the Python interpreter to crash (GH15337)
  • Bug in pd.read_csv() in which invalid values for nrows and chunksize were allowed (GH15767)
  • Bug in pd.read_csv() for the Python engine in which unhelpful error messages were being raised when parsing errors occurred (GH15910)
  • Bug in pd.read_csv() in which the skipfooter parameter was not being properly validated (GH15925)
  • Bug in pd.to_csv() in which there was numeric overflow when a timestamp index was being written (GH15982)
  • Bug in pd.util.hashing.hash_pandas_object() in which hashing of categoricals depended on the ordering of categories, instead of just their values. (GH15143)
  • Bug in .to_json() where lines=True and contents (keys or values) contain escaped characters (GH15096)
  • Bug in .to_json() causing single byte ascii characters to be expanded to four byte unicode (GH15344)
  • Bug in .to_json() for the C engine where rollover was not correctly handled for case where frac is odd and diff is exactly 0.5 (GH15716, GH15864)
  • Bug in pd.read_json() for Python 2 where lines=True and contents contain non-ascii unicode characters (GH15132)
  • Bug in pd.read_msgpack() in which Series categoricals were being improperly processed (GH14901)
  • Bug in pd.read_msgpack() which did not allow loading of a dataframe with an index of type CategoricalIndex (GH15487)
  • Bug in pd.read_msgpack() when deserializing a CategoricalIndex (GH15487)
  • Bug in DataFrame.to_records() with converting a DatetimeIndex with a timezone (GH13937)
  • Bug in DataFrame.to_records() which failed with unicode characters in column names (GH11879)
  • Bug in .to_sql() when writing a DataFrame with numeric index names (GH15404).
  • Bug in DataFrame.to_html() with index=False and max_rows raising in IndexError (GH14998)
  • Bug in pd.read_hdf() passing a Timestamp to the where parameter with a non date column (GH15492)
  • Bug in DataFrame.to_stata() and StataWriter which produces incorrectly formatted files to be produced for some locales (GH13856)
  • Bug in StataReader and StataWriter which allows invalid encodings (GH15723)
  • Bug in the Series repr not showing the length when the output was truncated (GH15962).

Plotting

  • Bug in DataFrame.hist where plt.tight_layout caused an AttributeError (use matplotlib >= 2.0.1) (GH9351)
  • Bug in DataFrame.boxplot where fontsize was not applied to the tick labels on both axes (GH15108)
  • Bug in the date and time converters pandas registers with matplotlib not handling multiple dimensions (GH16026)
  • Bug in pd.scatter_matrix() could accept either color or c, but not both (GH14855)

Groupby/Resample/Rolling

  • Bug in .groupby(..).resample() when passed the on= kwarg. (GH15021)
  • Properly set __name__ and __qualname__ for Groupby.* functions (GH14620)
  • Bug in GroupBy.get_group() failing with a categorical grouper (GH15155)
  • Bug in .groupby(...).rolling(...) when on is specified and using a DatetimeIndex (GH15130, GH13966)
  • Bug in groupby operations with timedelta64 when passing numeric_only=False (GH5724)
  • Bug in groupby.apply() coercing object dtypes to numeric types, when not all values were numeric (GH14423, GH15421, GH15670)
  • Bug in resample, where a non-string loffset argument would not be applied when resampling a timeseries (GH13218)
  • Bug in DataFrame.groupby().describe() when grouping on Index containing tuples (GH14848)
  • Bug in groupby().nunique() with a datetimelike-grouper where bins counts were incorrect (GH13453)
  • Bug in groupby.transform() that would coerce the resultant dtypes back to the original (GH10972, GH11444)
  • Bug in groupby.agg() incorrectly localizing timezone on datetime (GH15426, GH10668, GH13046)
  • Bug in .rolling/expanding() functions where count() was not counting np.Inf, nor handling object dtypes (GH12541)
  • Bug in .rolling() where pd.Timedelta or datetime.timedelta was not accepted as a window argument (GH15440)
  • Bug in Rolling.quantile function that caused a segmentation fault when called with a quantile value outside of the range [0, 1] (GH15463)
  • Bug in DataFrame.resample().median() if duplicate column names are present (GH14233)

Sparse

  • Bug in SparseSeries.reindex on single level with list of length 1 (GH15447)
  • Bug in repr-formatting a SparseDataFrame after a value was set on (a copy of) one of its series (GH15488)
  • Bug in SparseDataFrame construction with lists not coercing to dtype (GH15682)
  • Bug in sparse array indexing in which indices were not being validated (GH15863)

Reshaping

  • Bug in pd.merge_asof() where left_index or right_index caused a failure when multiple by was specified (GH15676)
  • Bug in pd.merge_asof() where left_index/right_index together caused a failure when tolerance was specified (GH15135)
  • Bug in DataFrame.pivot_table() where dropna=True would not drop all-NaN columns when the columns was a category dtype (GH15193)
  • Bug in pd.melt() where passing a tuple value for value_vars caused a TypeError (GH15348)
  • Bug in pd.pivot_table() where no error was raised when values argument was not in the columns (GH14938)
  • Bug in pd.concat() in which concatenating with an empty dataframe with join='inner' was being improperly handled (GH15328)
  • Bug with sort=True in DataFrame.join and pd.merge when joining on indexes (GH15582)
  • Bug in DataFrame.nsmallest and DataFrame.nlargest where identical values resulted in duplicated rows (GH15297)
  • Bug in pandas.pivot_table() incorrectly raising UnicodeError when passing unicode input for margins keyword (GH13292)

Numeric

  • Bug in .rank() which incorrectly ranks ordered categories (GH15420)
  • Bug in .corr() and .cov() where the column and index were the same object (GH14617)
  • Bug in .mode() where mode was not returned if was only a single value (GH15714)
  • Bug in pd.cut() with a single bin on an all 0s array (GH15428)
  • Bug in pd.qcut() with a single quantile and an array with identical values (GH15431)
  • Bug in pandas.tools.utils.cartesian_product() with large input can cause overflow on windows (GH15265)
  • Bug in .eval() which caused multiline evals to fail with local variables not on the first line (GH15342)

Other

  • Compat with SciPy 0.19.0 for testing on .interpolate() (GH15662)
  • Compat for 32-bit platforms for .qcut/cut; bins will now be int64 dtype (GH14866)
  • Bug in interactions with Qt when a QtApplication already exists (GH14372)
  • Avoid use of np.finfo() during import pandas removed to mitigate deadlock on Python GIL misuse (GH14641)

v0.19.2 (December 24, 2016)

This is a minor bug-fix release in the 0.19.x series and includes some small regression fixes, bug fixes and performance improvements. We recommend that all users upgrade to this version.

Highlights include:

Enhancements

The pd.merge_asof(), added in 0.19.0, gained some improvements:

  • pd.merge_asof() gained left_index/right_index and left_by/right_by arguments (GH14253)
  • pd.merge_asof() can take multiple columns in by parameter and has specialized dtypes for better performance (GH13936)

Performance Improvements

  • Performance regression with PeriodIndex (GH14822)
  • Performance regression in indexing with getitem (GH14930)
  • Improved performance of .replace() (GH12745)
  • Improved performance Series creation with a datetime index and dictionary data (GH14894)

Bug Fixes

  • Compat with python 3.6 for pickling of some offsets (GH14685)
  • Compat with python 3.6 for some indexing exception types (GH14684, GH14689)
  • Compat with python 3.6 for deprecation warnings in the test suite (GH14681)
  • Compat with python 3.6 for Timestamp pickles (GH14689)
  • Compat with dateutil==2.6.0; segfault reported in the testing suite (GH14621)
  • Allow nanoseconds in Timestamp.replace as a kwarg (GH14621)
  • Bug in pd.read_csv in which aliasing was being done for na_values when passed in as a dictionary (GH14203)
  • Bug in pd.read_csv in which column indices for a dict-like na_values were not being respected (GH14203)
  • Bug in pd.read_csv where reading files fails, if the number of headers is equal to the number of lines in the file (GH14515)
  • Bug in pd.read_csv for the Python engine in which an unhelpful error message was being raised when multi-char delimiters were not being respected with quotes (GH14582)
  • Fix bugs (GH14734, GH13654) in pd.read_sas and pandas.io.sas.sas7bdat.SAS7BDATReader that caused problems when reading a SAS file incrementally.
  • Bug in pd.read_csv for the Python engine in which an unhelpful error message was being raised when skipfooter was not being respected by Python’s CSV library (GH13879)
  • Bug in .fillna() in which timezone aware datetime64 values were incorrectly rounded (GH14872)
  • Bug in .groupby(..., sort=True) of a non-lexsorted MultiIndex when grouping with multiple levels (GH14776)
  • Bug in pd.cut with negative values and a single bin (GH14652)
  • Bug in pd.to_numeric where a 0 was not unsigned on a downcast='unsigned' argument (GH14401)
  • Bug in plotting regular and irregular timeseries using shared axes (sharex=True or ax.twinx()) (GH13341, GH14322).
  • Bug in not propagating exceptions in parsing invalid datetimes, noted in python 3.6 (GH14561)
  • Bug in resampling a DatetimeIndex in local TZ, covering a DST change, which would raise AmbiguousTimeError (GH14682)
  • Bug in indexing that transformed RecursionError into KeyError or IndexingError (GH14554)
  • Bug in HDFStore when writing a MultiIndex when using data_columns=True (GH14435)
  • Bug in HDFStore.append() when writing a Series and passing a min_itemsize argument containing a value for the index (GH11412)
  • Bug when writing to a HDFStore in table format with a min_itemsize value for the index and without asking to append (GH10381)
  • Bug in Series.groupby.nunique() raising an IndexError for an empty Series (GH12553)
  • Bug in DataFrame.nlargest and DataFrame.nsmallest when the index had duplicate values (GH13412)
  • Bug in clipboard functions on linux with python2 with unicode and separators (GH13747)
  • Bug in clipboard functions on Windows 10 and python 3 (GH14362, GH12807)
  • Bug in .to_clipboard() and Excel compat (GH12529)
  • Bug in DataFrame.combine_first() for integer columns (GH14687).
  • Bug in pd.read_csv() in which the dtype parameter was not being respected for empty data (GH14712)
  • Bug in pd.read_csv() in which the nrows parameter was not being respected for large input when using the C engine for parsing (GH7626)
  • Bug in pd.merge_asof() could not handle timezone-aware DatetimeIndex when a tolerance was specified (GH14844)
  • Explicit check in to_stata and StataWriter for out-of-range values when writing doubles (GH14618)
  • Bug in .plot(kind='kde') which did not drop missing values to generate the KDE Plot, instead generating an empty plot. (GH14821)
  • Bug in unstack() if called with a list of column(s) as an argument, regardless of the dtypes of all columns, they get coerced to object (GH11847)

v0.19.1 (November 3, 2016)

This is a minor bug-fix release from 0.19.0 and includes some small regression fixes, bug fixes and performance improvements. We recommend that all users upgrade to this version.

What’s new in v0.19.1

Performance Improvements

  • Fixed performance regression in factorization of Period data (GH14338)
  • Fixed performance regression in Series.asof(where) when where is a scalar (GH14461)
  • Improved performance in DataFrame.asof(where) when where is a scalar (GH14461)
  • Improved performance in .to_json() when lines=True (GH14408)
  • Improved performance in certain types of loc indexing with a MultiIndex (GH14551).

Bug Fixes

  • Source installs from PyPI will now again work without cython installed, as in previous versions (GH14204)
  • Compat with Cython 0.25 for building (GH14496)
  • Fixed regression where user-provided file handles were closed in read_csv (c engine) (GH14418).
  • Fixed regression in DataFrame.quantile when missing values where present in some columns (GH14357).
  • Fixed regression in Index.difference where the freq of a DatetimeIndex was incorrectly set (GH14323)
  • Added back pandas.core.common.array_equivalent with a deprecation warning (GH14555).
  • Bug in pd.read_csv for the C engine in which quotation marks were improperly parsed in skipped rows (GH14459)
  • Bug in pd.read_csv for Python 2.x in which Unicode quote characters were no longer being respected (GH14477)
  • Fixed regression in Index.append when categorical indices were appended (GH14545).
  • Fixed regression in pd.DataFrame where constructor fails when given dict with None value (GH14381)
  • Fixed regression in DatetimeIndex._maybe_cast_slice_bound when index is empty (GH14354).
  • Bug in localizing an ambiguous timezone when a boolean is passed (GH14402)
  • Bug in TimedeltaIndex addition with a Datetime-like object where addition overflow in the negative direction was not being caught (GH14068, GH14453)
  • Bug in string indexing against data with object Index may raise AttributeError (GH14424)
  • Corrrecly raise ValueError on empty input to pd.eval() and df.query() (GH13139)
  • Bug in RangeIndex.intersection when result is a empty set (GH14364).
  • Bug in groupby-transform broadcasting that could cause incorrect dtype coercion (GH14457)
  • Bug in Series.__setitem__ which allowed mutating read-only arrays (GH14359).
  • Bug in DataFrame.insert where multiple calls with duplicate columns can fail (GH14291)
  • pd.merge() will raise ValueError with non-boolean parameters in passed boolean type arguments (GH14434)
  • Bug in Timestamp where dates very near the minimum (1677-09) could underflow on creation (GH14415)
  • Bug in pd.concat where names of the keys were not propagated to the resulting MultiIndex (GH14252)
  • Bug in pd.concat where axis cannot take string parameters 'rows' or 'columns' (GH14369)
  • Bug in pd.concat with dataframes heterogeneous in length and tuple keys (GH14438)
  • Bug in MultiIndex.set_levels where illegal level values were still set after raising an error (GH13754)
  • Bug in DataFrame.to_json where lines=True and a value contained a } character (GH14391)
  • Bug in df.groupby causing an AttributeError when grouping a single index frame by a column and the index level (GH14327)
  • Bug in df.groupby where TypeError raised when pd.Grouper(key=...) is passed in a list (GH14334)
  • Bug in pd.pivot_table may raise TypeError or ValueError when index or columns is not scalar and values is not specified (GH14380)

v0.19.0 (October 2, 2016)

This is a major release from 0.18.1 and includes number of API changes, several new features, enhancements, and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this version.

Highlights include:

  • merge_asof() for asof-style time-series joining, see here
  • .rolling() is now time-series aware, see here
  • read_csv() now supports parsing Categorical data, see here
  • A function union_categorical() has been added for combining categoricals, see here
  • PeriodIndex now has its own period dtype, and changed to be more consistent with other Index classes. See here
  • Sparse data structures gained enhanced support of int and bool dtypes, see here
  • Comparison operations with Series no longer ignores the index, see here for an overview of the API changes.
  • Introduction of a pandas development API for utility functions, see here.
  • Deprecation of Panel4D and PanelND. We recommend to represent these types of n-dimensional data with the xarray package.
  • Removal of the previously deprecated modules pandas.io.data, pandas.io.wb, pandas.tools.rplot.

Warning

pandas >= 0.19.0 will no longer silence numpy ufunc warnings upon import, see here.

New features

merge_asof for asof-style time-series joining

A long-time requested feature has been added through the merge_asof() function, to support asof style joining of time-series (GH1870, GH13695, GH13709, GH13902). Full documentation is here.

The merge_asof() performs an asof merge, which is similar to a left-join except that we match on nearest key rather than equal keys.

In [1]: left = pd.DataFrame({'a': [1, 5, 10],
   ...:                      'left_val': ['a', 'b', 'c']})
   ...: 

In [2]: right = pd.DataFrame({'a': [1, 2, 3, 6, 7],
   ...:                      'right_val': [1, 2, 3, 6, 7]})
   ...: 

In [3]: left
Out[3]: 
    a left_val
0   1        a
1   5        b
2  10        c

In [4]: right
Out[4]: 
   a  right_val
0  1          1
1  2          2
2  3          3
3  6          6
4  7          7

We typically want to match exactly when possible, and use the most recent value otherwise.

In [5]: pd.merge_asof(left, right, on='a')
Out[5]: 
    a left_val  right_val
0   1        a          1
1   5        b          3
2  10        c          7

We can also match rows ONLY with prior data, and not an exact match.

In [6]: pd.merge_asof(left, right, on='a', allow_exact_matches=False)
Out[6]: 
    a left_val  right_val
0   1        a        NaN
1   5        b        3.0
2  10        c        7.0

In a typical time-series example, we have trades and quotes and we want to asof-join them. This also illustrates using the by parameter to group data before merging.

In [7]: trades = pd.DataFrame({
   ...:     'time': pd.to_datetime(['20160525 13:30:00.023',
   ...:                             '20160525 13:30:00.038',
   ...:                             '20160525 13:30:00.048',
   ...:                             '20160525 13:30:00.048',
   ...:                             '20160525 13:30:00.048']),
   ...:     'ticker': ['MSFT', 'MSFT',
   ...:                'GOOG', 'GOOG', 'AAPL'],
   ...:     'price': [51.95, 51.95,
   ...:               720.77, 720.92, 98.00],
   ...:     'quantity': [75, 155,
   ...:                  100, 100, 100]},
   ...:     columns=['time', 'ticker', 'price', 'quantity'])
   ...: 

In [8]: quotes = pd.DataFrame({
   ...:     'time': pd.to_datetime(['20160525 13:30:00.023',
   ...:                             '20160525 13:30:00.023',
   ...:                             '20160525 13:30:00.030',
   ...:                             '20160525 13:30:00.041',
   ...:                             '20160525 13:30:00.048',
   ...:                             '20160525 13:30:00.049',
   ...:                             '20160525 13:30:00.072',
   ...:                             '20160525 13:30:00.075']),
   ...:     'ticker': ['GOOG', 'MSFT', 'MSFT',
   ...:                'MSFT', 'GOOG', 'AAPL', 'GOOG',
   ...:                'MSFT'],
   ...:     'bid': [720.50, 51.95, 51.97, 51.99,
   ...:             720.50, 97.99, 720.50, 52.01],
   ...:     'ask': [720.93, 51.96, 51.98, 52.00,
   ...:             720.93, 98.01, 720.88, 52.03]},
   ...:     columns=['time', 'ticker', 'bid', 'ask'])
   ...: 
In [9]: trades
Out[9]: 
                     time ticker   price  quantity
0 2016-05-25 13:30:00.023   MSFT   51.95        75
1 2016-05-25 13:30:00.038   MSFT   51.95       155
2 2016-05-25 13:30:00.048   GOOG  720.77       100
3 2016-05-25 13:30:00.048   GOOG  720.92       100
4 2016-05-25 13:30:00.048   AAPL   98.00       100

In [10]: quotes
Out[10]: 
                     time ticker     bid     ask
0 2016-05-25 13:30:00.023   GOOG  720.50  720.93
1 2016-05-25 13:30:00.023   MSFT   51.95   51.96
2 2016-05-25 13:30:00.030   MSFT   51.97   51.98
3 2016-05-25 13:30:00.041   MSFT   51.99   52.00
4 2016-05-25 13:30:00.048   GOOG  720.50  720.93
5 2016-05-25 13:30:00.049   AAPL   97.99   98.01
6 2016-05-25 13:30:00.072   GOOG  720.50  720.88
7 2016-05-25 13:30:00.075   MSFT   52.01   52.03

An asof merge joins on the on, typically a datetimelike field, which is ordered, and in this case we are using a grouper in the by field. This is like a left-outer join, except that forward filling happens automatically taking the most recent non-NaN value.

In [11]: pd.merge_asof(trades, quotes,
   ....:               on='time',
   ....:               by='ticker')
   ....: 
Out[11]: 
                     time ticker   price  quantity     bid     ask
0 2016-05-25 13:30:00.023   MSFT   51.95        75   51.95   51.96
1 2016-05-25 13:30:00.038   MSFT   51.95       155   51.97   51.98
2 2016-05-25 13:30:00.048   GOOG  720.77       100  720.50  720.93
3 2016-05-25 13:30:00.048   GOOG  720.92       100  720.50  720.93
4 2016-05-25 13:30:00.048   AAPL   98.00       100     NaN     NaN

This returns a merged DataFrame with the entries in the same order as the original left passed DataFrame (trades in this case), with the fields of the quotes merged.

.rolling() is now time-series aware

.rolling() objects are now time-series aware and can accept a time-series offset (or convertible) for the window argument (GH13327, GH12995). See the full documentation here.

In [12]: dft = pd.DataFrame({'B': [0, 1, 2, np.nan, 4]},
   ....:                    index=pd.date_range('20130101 09:00:00', periods=5, freq='s'))
   ....: 

In [13]: dft
Out[13]: 
                       B
2013-01-01 09:00:00  0.0
2013-01-01 09:00:01  1.0
2013-01-01 09:00:02  2.0
2013-01-01 09:00:03  NaN
2013-01-01 09:00:04  4.0

This is a regular frequency index. Using an integer window parameter works to roll along the window frequency.

In [14]: dft.rolling(2).sum()
Out[14]: 
                       B
2013-01-01 09:00:00  NaN
2013-01-01 09:00:01  1.0
2013-01-01 09:00:02  3.0
2013-01-01 09:00:03  NaN
2013-01-01 09:00:04  NaN

In [15]: dft.rolling(2, min_periods=1).sum()
Out[15]: 
                       B
2013-01-01 09:00:00  0.0
2013-01-01 09:00:01  1.0
2013-01-01 09:00:02  3.0
2013-01-01 09:00:03  2.0
2013-01-01 09:00:04  4.0

Specifying an offset allows a more intuitive specification of the rolling frequency.

In [16]: dft.rolling('2s').sum()
Out[16]: 
                       B
2013-01-01 09:00:00  0.0
2013-01-01 09:00:01  1.0
2013-01-01 09:00:02  3.0
2013-01-01 09:00:03  2.0
2013-01-01 09:00:04  4.0

Using a non-regular, but still monotonic index, rolling with an integer window does not impart any special calculation.

In [17]: dft = DataFrame({'B': [0, 1, 2, np.nan, 4]},
   ....:                 index = pd.Index([pd.Timestamp('20130101 09:00:00'),
   ....:                                   pd.Timestamp('20130101 09:00:02'),
   ....:                                   pd.Timestamp('20130101 09:00:03'),
   ....:                                   pd.Timestamp('20130101 09:00:05'),
   ....:                                   pd.Timestamp('20130101 09:00:06')],
   ....:                                  name='foo'))
   ....: 

In [18]: dft
Out[18]: 
                       B
foo                     
2013-01-01 09:00:00  0.0
2013-01-01 09:00:02  1.0
2013-01-01 09:00:03  2.0
2013-01-01 09:00:05  NaN
2013-01-01 09:00:06  4.0

In [19]: dft.rolling(2).sum()
Out[19]: 
                       B
foo                     
2013-01-01 09:00:00  NaN
2013-01-01 09:00:02  1.0
2013-01-01 09:00:03  3.0
2013-01-01 09:00:05  NaN
2013-01-01 09:00:06  NaN

Using the time-specification generates variable windows for this sparse data.

In [20]: dft.rolling('2s').sum()
Out[20]: 
                       B
foo                     
2013-01-01 09:00:00  0.0
2013-01-01 09:00:02  1.0
2013-01-01 09:00:03  3.0
2013-01-01 09:00:05  NaN
2013-01-01 09:00:06  4.0

Furthermore, we now allow an optional on parameter to specify a column (rather than the default of the index) in a DataFrame.

In [21]: dft = dft.reset_index()

In [22]: dft
Out[22]: 
                  foo    B
0 2013-01-01 09:00:00  0.0
1 2013-01-01 09:00:02  1.0
2 2013-01-01 09:00:03  2.0
3 2013-01-01 09:00:05  NaN
4 2013-01-01 09:00:06  4.0

In [23]: dft.rolling('2s', on='foo').sum()
Out[23]: 
                  foo    B
0 2013-01-01 09:00:00  0.0
1 2013-01-01 09:00:02  1.0
2 2013-01-01 09:00:03  3.0
3 2013-01-01 09:00:05  NaN
4 2013-01-01 09:00:06  4.0

read_csv has improved support for duplicate column names

Duplicate column names are now supported in read_csv() whether they are in the file or passed in as the names parameter (GH7160, GH9424)

In [24]: data = '0,1,2\n3,4,5'

In [25]: names = ['a', 'b', 'a']

Previous behavior:

In [2]: pd.read_csv(StringIO(data), names=names)
Out[2]:
   a  b  a
0  2  1  2
1  5  4  5

The first a column contained the same data as the second a column, when it should have contained the values [0, 3].

New behavior:

In [26]: pd.read_csv(StringIO(data), names=names)
Out[26]: 
   a  b  a.1
0  0  1    2
1  3  4    5

read_csv supports parsing Categorical directly

The read_csv() function now supports parsing a Categorical column when specified as a dtype (GH10153). Depending on the structure of the data, this can result in a faster parse time and lower memory usage compared to converting to Categorical after parsing. See the io docs here.

In [27]: data = 'col1,col2,col3\na,b,1\na,b,2\nc,d,3'

In [28]: pd.read_csv(StringIO(data))
Out[28]: 
  col1 col2  col3
0    a    b     1
1    a    b     2
2    c    d     3

In [29]: pd.read_csv(StringIO(data)).dtypes
Out[29]: 
col1    object
col2    object
col3     int64
dtype: object

In [30]: pd.read_csv(StringIO(data), dtype='category').dtypes
Out[30]: 
col1    category
col2    category
col3    category
dtype: object

Individual columns can be parsed as a Categorical using a dict specification

In [31]: pd.read_csv(StringIO(data), dtype={'col1': 'category'}).dtypes
Out[31]: 
col1    category
col2      object
col3       int64
dtype: object

Note

The resulting categories will always be parsed as strings (object dtype). If the categories are numeric they can be converted using the to_numeric() function, or as appropriate, another converter such as to_datetime().

In [32]: df = pd.read_csv(StringIO(data), dtype='category')

In [33]: df.dtypes
Out[33]: 
col1    category
col2    category
col3    category
dtype: object

In [34]: df['col3']
Out[34]: 
0    1
1    2
2    3
Name: col3, dtype: category
Categories (3, object): [1, 2, 3]

In [35]: df['col3'].cat.categories = pd.to_numeric(df['col3'].cat.categories)

In [36]: df['col3']
Out[36]: 
0    1
1    2
2    3
Name: col3, dtype: category
Categories (3, int64): [1, 2, 3]

Categorical Concatenation

  • A function union_categoricals() has been added for combining categoricals, see Unioning Categoricals (GH13361, GH13763, GH13846, GH14173)

    In [37]: from pandas.api.types import union_categoricals
    
    In [38]: a = pd.Categorical(["b", "c"])
    
    In [39]: b = pd.Categorical(["a", "b"])
    
    In [40]: union_categoricals([a, b])
    Out[40]: 
    [b, c, a, b]
    Categories (3, object): [b, c, a]
    
  • concat and append now can concat category dtypes with different categories as object dtype (GH13524)

    In [41]: s1 = pd.Series(['a', 'b'], dtype='category')
    
    In [42]: s2 = pd.Series(['b', 'c'], dtype='category')
    

    Previous behavior:

    In [1]: pd.concat([s1, s2])
    ValueError: incompatible categories in categorical concat
    

    New behavior:

    In [43]: pd.concat([s1, s2])
    Out[43]: 
    0    a
    1    b
    0    b
    1    c
    dtype: object
    

Semi-Month Offsets

Pandas has gained new frequency offsets, SemiMonthEnd (‘SM’) and SemiMonthBegin (‘SMS’). These provide date offsets anchored (by default) to the 15th and end of month, and 15th and 1st of month respectively. (GH1543)

In [44]: from pandas.tseries.offsets import SemiMonthEnd, SemiMonthBegin

SemiMonthEnd:

In [45]: Timestamp('2016-01-01') + SemiMonthEnd()
Out[45]: Timestamp('2016-01-15 00:00:00')

In [46]: pd.date_range('2015-01-01', freq='SM', periods=4)
Out[46]: DatetimeIndex(['2015-01-15', '2015-01-31', '2015-02-15', '2015-02-28'], dtype='datetime64[ns]', freq='SM-15')

SemiMonthBegin:

In [47]: Timestamp('2016-01-01') + SemiMonthBegin()
Out[47]: Timestamp('2016-01-15 00:00:00')

In [48]: pd.date_range('2015-01-01', freq='SMS', periods=4)
Out[48]: DatetimeIndex(['2015-01-01', '2015-01-15', '2015-02-01', '2015-02-15'], dtype='datetime64[ns]', freq='SMS-15')

Using the anchoring suffix, you can also specify the day of month to use instead of the 15th.

In [49]: pd.date_range('2015-01-01', freq='SMS-16', periods=4)
Out[49]: DatetimeIndex(['2015-01-01', '2015-01-16', '2015-02-01', '2015-02-16'], dtype='datetime64[ns]', freq='SMS-16')

In [50]: pd.date_range('2015-01-01', freq='SM-14', periods=4)
Out[50]: DatetimeIndex(['2015-01-14', '2015-01-31', '2015-02-14', '2015-02-28'], dtype='datetime64[ns]', freq='SM-14')

New Index methods

The following methods and options are added to Index, to be more consistent with the Series and DataFrame API.

Index now supports the .where() function for same shape indexing (GH13170)

In [51]: idx = pd.Index(['a', 'b', 'c'])

In [52]: idx.where([True, False, True])
Out[52]: Index(['a', nan, 'c'], dtype='object')

Index now supports .dropna() to exclude missing values (GH6194)

In [53]: idx = pd.Index([1, 2, np.nan, 4])

In [54]: idx.dropna()
Out[54]: Float64Index([1.0, 2.0, 4.0], dtype='float64')

For MultiIndex, values are dropped if any level is missing by default. Specifying how='all' only drops values where all levels are missing.

In [55]: midx = pd.MultiIndex.from_arrays([[1, 2, np.nan, 4],
   ....:                                     [1, 2, np.nan, np.nan]])
   ....: 

In [56]: midx
Out[56]: 
MultiIndex(levels=[[1, 2, 4], [1, 2]],
           labels=[[0, 1, -1, 2], [0, 1, -1, -1]])

In [57]: midx.dropna()
Out[57]: 
MultiIndex(levels=[[1, 2, 4], [1, 2]],
           labels=[[0, 1], [0, 1]])

In [58]: midx.dropna(how='all')
Out[58]: 
MultiIndex(levels=[[1, 2, 4], [1, 2]],
           labels=[[0, 1, 2], [0, 1, -1]])

Index now supports .str.extractall() which returns a DataFrame, see the docs here (GH10008, GH13156)

In [59]: idx = pd.Index(["a1a2", "b1", "c1"])

In [60]: idx.str.extractall("[ab](?P<digit>\d)")
Out[60]: 
        digit
  match      
0 0         1
  1         2
1 0         1

Index.astype() now accepts an optional boolean argument copy, which allows optional copying if the requirements on dtype are satisfied (GH13209)

Google BigQuery Enhancements

  • The read_gbq() method has gained the dialect argument to allow users to specify whether to use BigQuery’s legacy SQL or BigQuery’s standard SQL. See the docs for more details (GH13615).
  • The to_gbq() method now allows the DataFrame column order to differ from the destination table schema (GH11359).

Fine-grained numpy errstate

Previous versions of pandas would permanently silence numpy’s ufunc error handling when pandas was imported. Pandas did this in order to silence the warnings that would arise from using numpy ufuncs on missing data, which are usually represented as NaN s. Unfortunately, this silenced legitimate warnings arising in non-pandas code in the application. Starting with 0.19.0, pandas will use the numpy.errstate context manager to silence these warnings in a more fine-grained manner, only around where these operations are actually used in the pandas codebase. (GH13109, GH13145)

After upgrading pandas, you may see new RuntimeWarnings being issued from your code. These are likely legitimate, and the underlying cause likely existed in the code when using previous versions of pandas that simply silenced the warning. Use numpy.errstate around the source of the RuntimeWarning to control how these conditions are handled.

get_dummies now returns integer dtypes

The pd.get_dummies function now returns dummy-encoded columns as small integers, rather than floats (GH8725). This should provide an improved memory footprint.

Previous behavior:

In [1]: pd.get_dummies(['a', 'b', 'a', 'c']).dtypes

Out[1]:
a    float64
b    float64
c    float64
dtype: object

New behavior:

In [61]: pd.get_dummies(['a', 'b', 'a', 'c']).dtypes
Out[61]: 
a    uint8
b    uint8
c    uint8
dtype: object

Downcast values to smallest possible dtype in to_numeric

pd.to_numeric() now accepts a downcast parameter, which will downcast the data if possible to smallest specified numerical dtype (GH13352)

In [62]: s = ['1', 2, 3]

In [63]: pd.to_numeric(s, downcast='unsigned')
Out[63]: array([1, 2, 3], dtype=uint8)

In [64]: pd.to_numeric(s, downcast='integer')
Out[64]: array([1, 2, 3], dtype=int8)

pandas development API

As part of making pandas API more uniform and accessible in the future, we have created a standard sub-package of pandas, pandas.api to hold public API’s. We are starting by exposing type introspection functions in pandas.api.types. More sub-packages and officially sanctioned API’s will be published in future versions of pandas (GH13147, GH13634)

The following are now part of this API:

In [65]: import pprint

In [66]: from pandas.api import types

In [67]: funcs = [ f for f in dir(types) if not f.startswith('_') ]

In [68]: pprint.pprint(funcs)
['CategoricalDtype',
 'DatetimeTZDtype',
 'IntervalDtype',
 'PeriodDtype',
 'infer_dtype',
 'is_any_int_dtype',
 'is_array_like',
 'is_bool',
 'is_bool_dtype',
 'is_categorical',
 'is_categorical_dtype',
 'is_complex',
 'is_complex_dtype',
 'is_datetime64_any_dtype',
 'is_datetime64_dtype',
 'is_datetime64_ns_dtype',
 'is_datetime64tz_dtype',
 'is_datetimetz',
 'is_dict_like',
 'is_dtype_equal',
 'is_extension_type',
 'is_file_like',
 'is_float',
 'is_float_dtype',
 'is_floating_dtype',
 'is_hashable',
 'is_int64_dtype',
 'is_integer',
 'is_integer_dtype',
 'is_interval',
 'is_interval_dtype',
 'is_iterator',
 'is_list_like',
 'is_named_tuple',
 'is_number',
 'is_numeric_dtype',
 'is_object_dtype',
 'is_period',
 'is_period_dtype',
 'is_re',
 'is_re_compilable',
 'is_scalar',
 'is_sequence',
 'is_signed_integer_dtype',
 'is_sparse',
 'is_string_dtype',
 'is_timedelta64_dtype',
 'is_timedelta64_ns_dtype',
 'is_unsigned_integer_dtype',
 'pandas_dtype',
 'union_categoricals']

Note

Calling these functions from the internal module pandas.core.common will now show a DeprecationWarning (GH13990)

Other enhancements

  • Timestamp can now accept positional and keyword parameters similar to datetime.datetime() (GH10758, GH11630)

    In [69]: pd.Timestamp(2012, 1, 1)
    Out[69]: Timestamp('2012-01-01 00:00:00')
    
    In [70]: pd.Timestamp(year=2012, month=1, day=1, hour=8, minute=30)
    Out[70]: Timestamp('2012-01-01 08:30:00')
    
  • The .resample() function now accepts a on= or level= parameter for resampling on a datetimelike column or MultiIndex level (GH13500)

    In [71]: df = pd.DataFrame({'date': pd.date_range('2015-01-01', freq='W', periods=5),
       ....:                    'a': np.arange(5)},
       ....:                   index=pd.MultiIndex.from_arrays([
       ....:                            [1,2,3,4,5],
       ....:                            pd.date_range('2015-01-01', freq='W', periods=5)],
       ....:                        names=['v','d']))
       ....: 
    
    In [72]: df
    Out[72]: 
                       date  a
    v d                       
    1 2015-01-04 2015-01-04  0
    2 2015-01-11 2015-01-11  1
    3 2015-01-18 2015-01-18  2
    4 2015-01-25 2015-01-25  3
    5 2015-02-01 2015-02-01  4
    
    In [73]: df.resample('M', on='date').sum()
    Out[73]: 
                a
    date         
    2015-01-31  6
    2015-02-28  4
    
    In [74]: df.resample('M', level='d').sum()
    Out[74]: 
                a
    d            
    2015-01-31  6
    2015-02-28  4
    
  • The .get_credentials() method of GbqConnector can now first try to fetch the application default credentials. See the docs for more details (GH13577).

  • The .tz_localize() method of DatetimeIndex and Timestamp has gained the errors keyword, so you can potentially coerce nonexistent timestamps to NaT. The default behavior remains to raising a NonExistentTimeError (GH13057)

  • .to_hdf/read_hdf() now accept path objects (e.g. pathlib.Path, py.path.local) for the file path (GH11773)

  • The pd.read_csv() with engine='python' has gained support for the decimal (GH12933), na_filter (GH13321) and the memory_map option (GH13381).

  • Consistent with the Python API, pd.read_csv() will now interpret +inf as positive infinity (GH13274)

  • The pd.read_html() has gained support for the na_values, converters, keep_default_na options (GH13461)

  • Categorical.astype() now accepts an optional boolean argument copy, effective when dtype is categorical (GH13209)

  • DataFrame has gained the .asof() method to return the last non-NaN values according to the selected subset (GH13358)

  • The DataFrame constructor will now respect key ordering if a list of OrderedDict objects are passed in (GH13304)

  • pd.read_html() has gained support for the decimal option (GH12907)

  • Series has gained the properties .is_monotonic, .is_monotonic_increasing, .is_monotonic_decreasing, similar to Index (GH13336)

  • DataFrame.to_sql() now allows a single value as the SQL type for all columns (GH11886).

  • Series.append now supports the ignore_index option (GH13677)

  • .to_stata() and StataWriter can now write variable labels to Stata dta files using a dictionary to make column names to labels (GH13535, GH13536)

  • .to_stata() and StataWriter will automatically convert datetime64[ns] columns to Stata format %tc, rather than raising a ValueError (GH12259)

  • read_stata() and StataReader raise with a more explicit error message when reading Stata files with repeated value labels when convert_categoricals=True (GH13923)

  • DataFrame.style will now render sparsified MultiIndexes (GH11655)

  • DataFrame.style will now show column level names (e.g. DataFrame.columns.names) (GH13775)

  • DataFrame has gained support to re-order the columns based on the values in a row using df.sort_values(by='...', axis=1) (GH10806)

    In [75]: df = pd.DataFrame({'A': [2, 7], 'B': [3, 5], 'C': [4, 8]},
       ....:                   index=['row1', 'row2'])
       ....: 
    
    In [76]: df
    Out[76]: 
          A  B  C
    row1  2  3  4
    row2  7  5  8
    
    In [77]: df.sort_values(by='row2', axis=1)
    Out[77]: 
          B  A  C
    row1  3  2  4
    row2  5  7  8
    
  • Added documentation to I/O regarding the perils of reading in columns with mixed dtypes and how to handle it (GH13746)

  • to_html() now has a border argument to control the value in the opening <table> tag. The default is the value of the html.border option, which defaults to 1. This also affects the notebook HTML repr, but since Jupyter’s CSS includes a border-width attribute, the visual effect is the same. (GH11563).

  • Raise ImportError in the sql functions when sqlalchemy is not installed and a connection string is used (GH11920).

  • Compatibility with matplotlib 2.0. Older versions of pandas should also work with matplotlib 2.0 (GH13333)

  • Timestamp, Period, DatetimeIndex, PeriodIndex and .dt accessor have gained a .is_leap_year property to check whether the date belongs to a leap year. (GH13727)

  • astype() will now accept a dict of column name to data types mapping as the dtype argument. (GH12086)

  • The pd.read_json and DataFrame.to_json has gained support for reading and writing json lines with lines option see Line delimited json (GH9180)

  • read_excel() now supports the true_values and false_values keyword arguments (GH13347)

  • groupby() will now accept a scalar and a single-element list for specifying level on a non-MultiIndex grouper. (GH13907)

  • Non-convertible dates in an excel date column will be returned without conversion and the column will be object dtype, rather than raising an exception (GH10001).

  • pd.Timedelta(None) is now accepted and will return NaT, mirroring pd.Timestamp (GH13687)

  • pd.read_stata() can now handle some format 111 files, which are produced by SAS when generating Stata dta files (GH11526)

  • Series and Index now support divmod which will return a tuple of series or indices. This behaves like a standard binary operator with regards to broadcasting rules (GH14208).

API changes

Series.tolist() will now return Python types

Series.tolist() will now return Python types in the output, mimicking NumPy .tolist() behavior (GH10904)

In [78]: s = pd.Series([1,2,3])

Previous behavior:

In [7]: type(s.tolist()[0])
Out[7]:
 <class 'numpy.int64'>

New behavior:

In [79]: type(s.tolist()[0])
Out[79]: int

Series operators for different indexes

Following Series operators have been changed to make all operators consistent, including DataFrame (GH1134, GH4581, GH13538)

  • Series comparison operators now raise ValueError when index are different.
  • Series logical operators align both index of left and right hand side.

Warning

Until 0.18.1, comparing Series with the same length, would succeed even if the .index are different (the result ignores .index). As of 0.19.0, this will raises ValueError to be more strict. This section also describes how to keep previous behavior or align different indexes, using the flexible comparison methods like .eq.

As a result, Series and DataFrame operators behave as below:

Arithmetic operators

Arithmetic operators align both index (no changes).

In [80]: s1 = pd.Series([1, 2, 3], index=list('ABC'))

In [81]: s2 = pd.Series([2, 2, 2], index=list('ABD'))

In [82]: s1 + s2
Out[82]: 
A    3.0
B    4.0
C    NaN
D    NaN
dtype: float64

In [83]: df1 = pd.DataFrame([1, 2, 3], index=list('ABC'))

In [84]: df2 = pd.DataFrame([2, 2, 2], index=list('ABD'))

In [85]: df1 + df2
Out[85]: 
     0
A  3.0
B  4.0
C  NaN
D  NaN
Comparison operators

Comparison operators raise ValueError when .index are different.

Previous Behavior (Series):

Series compared values ignoring the .index as long as both had the same length:

In [1]: s1 == s2
Out[1]:
A    False
B     True
C    False
dtype: bool

New behavior (Series):

In [2]: s1 == s2
Out[2]:
ValueError: Can only compare identically-labeled Series objects

Note

To achieve the same result as previous versions (compare values based on locations ignoring .index), compare both .values.

In [86]: s1.values == s2.values
Out[86]: array([False,  True, False], dtype=bool)

If you want to compare Series aligning its .index, see flexible comparison methods section below:

In [87]: s1.eq(s2)
Out[87]: 
A    False
B     True
C    False
D    False
dtype: bool

Current Behavior (DataFrame, no change):

In [3]: df1 == df2
Out[3]:
ValueError: Can only compare identically-labeled DataFrame objects
Logical operators

Logical operators align both .index of left and right hand side.

Previous behavior (Series), only left hand side index was kept:

In [4]: s1 = pd.Series([True, False, True], index=list('ABC'))
In [5]: s2 = pd.Series([True, True, True], index=list('ABD'))
In [6]: s1 & s2
Out[6]:
A     True
B    False
C    False
dtype: bool

New behavior (Series):

In [88]: s1 = pd.Series([True, False, True], index=list('ABC'))

In [89]: s2 = pd.Series([True, True, True], index=list('ABD'))

In [90]: s1 & s2
Out[90]: 
A     True
B    False
C    False
D    False
dtype: bool

Note

Series logical operators fill a NaN result with False.

Note

To achieve the same result as previous versions (compare values based on only left hand side index), you can use reindex_like:

In [91]: s1 & s2.reindex_like(s1)
Out[91]: 
A     True
B    False
C    False
dtype: bool

Current Behavior (DataFrame, no change):

In [92]: df1 = pd.DataFrame([True, False, True], index=list('ABC'))

In [93]: df2 = pd.DataFrame([True, True, True], index=list('ABD'))

In [94]: df1 & df2
Out[94]: 
       0
A   True
B  False
C    NaN
D    NaN
Flexible comparison methods

Series flexible comparison methods like eq, ne, le, lt, ge and gt now align both index. Use these operators if you want to compare two Series which has the different index.

In [95]: s1 = pd.Series([1, 2, 3], index=['a', 'b', 'c'])

In [96]: s2 = pd.Series([2, 2, 2], index=['b', 'c', 'd'])

In [97]: s1.eq(s2)
Out[97]: 
a    False
b     True
c    False
d    False
dtype: bool

In [98]: s1.ge(s2)
Out[98]: 
a    False
b     True
c     True
d    False
dtype: bool

Previously, this worked the same as comparison operators (see above).

Series type promotion on assignment

A Series will now correctly promote its dtype for assignment with incompat values to the current dtype (GH13234)

In [99]: s = pd.Series()

Previous behavior:

In [2]: s["a"] = pd.Timestamp("2016-01-01")

In [3]: s["b"] = 3.0
TypeError: invalid type promotion

New behavior:

In [100]: s["a"] = pd.Timestamp("2016-01-01")

In [101]: s["b"] = 3.0

In [102]: s
Out[102]: 
a    2016-01-01 00:00:00
b                      3
dtype: object

In [103]: s.dtype
Out[103]: dtype('O')

.to_datetime() changes

Previously if .to_datetime() encountered mixed integers/floats and strings, but no datetimes with errors='coerce' it would convert all to NaT.

Previous behavior:

In [2]: pd.to_datetime([1, 'foo'], errors='coerce')
Out[2]: DatetimeIndex(['NaT', 'NaT'], dtype='datetime64[ns]', freq=None)

Current behavior:

This will now convert integers/floats with the default unit of ns.

In [104]: pd.to_datetime([1, 'foo'], errors='coerce')
Out[104]: DatetimeIndex(['1970-01-01 00:00:00.000000001', 'NaT'], dtype='datetime64[ns]', freq=None)

Bug fixes related to .to_datetime():

  • Bug in pd.to_datetime() when passing integers or floats, and no unit and errors='coerce' (GH13180).
  • Bug in pd.to_datetime() when passing invalid datatypes (e.g. bool); will now respect the errors keyword (GH13176)
  • Bug in pd.to_datetime() which overflowed on int8, and int16 dtypes (GH13451)
  • Bug in pd.to_datetime() raise AttributeError with NaN and the other string is not valid when errors='ignore' (GH12424)
  • Bug in pd.to_datetime() did not cast floats correctly when unit was specified, resulting in truncated datetime (GH13834)

Merging changes

Merging will now preserve the dtype of the join keys (GH8596)

In [105]: df1 = pd.DataFrame({'key': [1], 'v1': [10]})

In [106]: df1
Out[106]: 
   key  v1
0    1  10

In [107]: df2 = pd.DataFrame({'key': [1, 2], 'v1': [20, 30]})

In [108]: df2
Out[108]: 
   key  v1
0    1  20
1    2  30

Previous behavior:

In [5]: pd.merge(df1, df2, how='outer')
Out[5]:
   key    v1
0  1.0  10.0
1  1.0  20.0
2  2.0  30.0

In [6]: pd.merge(df1, df2, how='outer').dtypes
Out[6]:
key    float64
v1     float64
dtype: object

New behavior:

We are able to preserve the join keys

In [109]: pd.merge(df1, df2, how='outer')
Out[109]: 
   key  v1
0    1  10
1    1  20
2    2  30

In [110]: pd.merge(df1, df2, how='outer').dtypes
Out[110]: 
key    int64
v1     int64
dtype: object

Of course if you have missing values that are introduced, then the resulting dtype will be upcast, which is unchanged from previous.

In [111]: pd.merge(df1, df2, how='outer', on='key')
Out[111]: 
   key  v1_x  v1_y
0    1  10.0    20
1    2   NaN    30

In [112]: pd.merge(df1, df2, how='outer', on='key').dtypes
Out[112]: 
key       int64
v1_x    float64
v1_y      int64
dtype: object

.describe() changes

Percentile identifiers in the index of a .describe() output will now be rounded to the least precision that keeps them distinct (GH13104)

In [113]: s = pd.Series([0, 1, 2, 3, 4])

In [114]: df = pd.DataFrame([0, 1, 2, 3, 4])

Previous behavior:

The percentiles were rounded to at most one decimal place, which could raise ValueError for a data frame if the percentiles were duplicated.

In [3]: s.describe(percentiles=[0.0001, 0.0005, 0.001, 0.999, 0.9995, 0.9999])
Out[3]:
count     5.000000
mean      2.000000
std       1.581139
min       0.000000
0.0%      0.000400
0.1%      0.002000
0.1%      0.004000
50%       2.000000
99.9%     3.996000
100.0%    3.998000
100.0%    3.999600
max       4.000000
dtype: float64

In [4]: df.describe(percentiles=[0.0001, 0.0005, 0.001, 0.999, 0.9995, 0.9999])
Out[4]:
...
ValueError: cannot reindex from a duplicate axis

New behavior:

In [115]: s.describe(percentiles=[0.0001, 0.0005, 0.001, 0.999, 0.9995, 0.9999])
Out[115]: 
count     5.000000
mean      2.000000
std       1.581139
min       0.000000
0.01%     0.000400
0.05%     0.002000
0.1%      0.004000
50%       2.000000
99.9%     3.996000
99.95%    3.998000
99.99%    3.999600
max       4.000000
dtype: float64

In [116]: df.describe(percentiles=[0.0001, 0.0005, 0.001, 0.999, 0.9995, 0.9999])
Out[116]: 
               0
count   5.000000
mean    2.000000
std     1.581139
min     0.000000
0.01%   0.000400
0.05%   0.002000
0.1%    0.004000
50%     2.000000
99.9%   3.996000
99.95%  3.998000
99.99%  3.999600
max     4.000000

Furthermore:

  • Passing duplicated percentiles will now raise a ValueError.
  • Bug in .describe() on a DataFrame with a mixed-dtype column index, which would previously raise a TypeError (GH13288)

Period changes

PeriodIndex now has period dtype

PeriodIndex now has its own period dtype. The period dtype is a pandas extension dtype like category or the timezone aware dtype (datetime64[ns, tz]) (GH13941). As a consequence of this change, PeriodIndex no longer has an integer dtype:

Previous behavior:

In [1]: pi = pd.PeriodIndex(['2016-08-01'], freq='D')

In [2]: pi
Out[2]: PeriodIndex(['2016-08-01'], dtype='int64', freq='D')

In [3]: pd.api.types.is_integer_dtype(pi)
Out[3]: True

In [4]: pi.dtype
Out[4]: dtype('int64')

New behavior:

In [117]: pi = pd.PeriodIndex(['2016-08-01'], freq='D')

In [118]: pi
Out[118]: PeriodIndex(['2016-08-01'], dtype='period[D]', freq='D')

In [119]: pd.api.types.is_integer_dtype(pi)
Out[119]: False

In [120]: pd.api.types.is_period_dtype(pi)
Out[120]: True

In [121]: pi.dtype
Out[121]: period[D]

In [122]: type(pi.dtype)
Out[122]: pandas.core.dtypes.dtypes.PeriodDtype
Period('NaT') now returns pd.NaT

Previously, Period has its own Period('NaT') representation different from pd.NaT. Now Period('NaT') has been changed to return pd.NaT. (GH12759, GH13582)

Previous behavior:

In [5]: pd.Period('NaT', freq='D')
Out[5]: Period('NaT', 'D')

New behavior:

These result in pd.NaT without providing freq option.

In [123]: pd.Period('NaT')
Out[123]: NaT

In [124]: pd.Period(None)
Out[124]: NaT

To be compatible with Period addition and subtraction, pd.NaT now supports addition and subtraction with int. Previously it raised ValueError.

Previous behavior:

In [5]: pd.NaT + 1
...
ValueError: Cannot add integral value to Timestamp without freq.

New behavior:

In [125]: pd.NaT + 1
Out[125]: NaT

In [126]: pd.NaT - 1
Out[126]: NaT
PeriodIndex.values now returns array of Period object

.values is changed to return an array of Period objects, rather than an array of integers (GH13988).

Previous behavior:

In [6]: pi = pd.PeriodIndex(['2011-01', '2011-02'], freq='M')
In [7]: pi.values
array([492, 493])

New behavior:

In [127]: pi = pd.PeriodIndex(['2011-01', '2011-02'], freq='M')

In [128]: pi.values
Out[128]: array([Period('2011-01', 'M'), Period('2011-02', 'M')], dtype=object)

Index + / - no longer used for set operations

Addition and subtraction of the base Index type and of DatetimeIndex (not the numeric index types) previously performed set operations (set union and difference). This behavior was already deprecated since 0.15.0 (in favor using the specific .union() and .difference() methods), and is now disabled. When possible, + and - are now used for element-wise operations, for example for concatenating strings or subtracting datetimes (GH8227, GH14127).

Previous behavior:

In [1]: pd.Index(['a', 'b']) + pd.Index(['a', 'c'])
FutureWarning: using '+' to provide set union with Indexes is deprecated, use '|' or .union()
Out[1]: Index(['a', 'b', 'c'], dtype='object')

New behavior: the same operation will now perform element-wise addition:

In [129]: pd.Index(['a', 'b']) + pd.Index(['a', 'c'])
Out[129]: Index(['aa', 'bc'], dtype='object')

Note that numeric Index objects already performed element-wise operations. For example, the behavior of adding two integer Indexes is unchanged. The base Index is now made consistent with this behavior.

In [130]: pd.Index([1, 2, 3]) + pd.Index([2, 3, 4])
Out[130]: Int64Index([3, 5, 7], dtype='int64')

Further, because of this change, it is now possible to subtract two DatetimeIndex objects resulting in a TimedeltaIndex:

Previous behavior:

In [1]: pd.DatetimeIndex(['2016-01-01', '2016-01-02']) - pd.DatetimeIndex(['2016-01-02', '2016-01-03'])
FutureWarning: using '-' to provide set differences with datetimelike Indexes is deprecated, use .difference()
Out[1]: DatetimeIndex(['2016-01-01'], dtype='datetime64[ns]', freq=None)

New behavior:

In [131]: pd.DatetimeIndex(['2016-01-01', '2016-01-02']) - pd.DatetimeIndex(['2016-01-02', '2016-01-03'])
Out[131]: TimedeltaIndex(['-1 days', '-1 days'], dtype='timedelta64[ns]', freq=None)

Index.difference and .symmetric_difference changes

Index.difference and Index.symmetric_difference will now, more consistently, treat NaN values as any other values. (GH13514)

In [132]: idx1 = pd.Index([1, 2, 3, np.nan])

In [133]: idx2 = pd.Index([0, 1, np.nan])

Previous behavior:

In [3]: idx1.difference(idx2)
Out[3]: Float64Index([nan, 2.0, 3.0], dtype='float64')

In [4]: idx1.symmetric_difference(idx2)
Out[4]: Float64Index([0.0, nan, 2.0, 3.0], dtype='float64')

New behavior:

In [134]: idx1.difference(idx2)
Out[134]: Float64Index([2.0, 3.0], dtype='float64')

In [135]: idx1.symmetric_difference(idx2)
Out[135]: Float64Index([0.0, 2.0, 3.0], dtype='float64')

Index.unique consistently returns Index

Index.unique() now returns unique values as an Index of the appropriate dtype. (GH13395). Previously, most Index classes returned np.ndarray, and DatetimeIndex, TimedeltaIndex and PeriodIndex returned Index to keep metadata like timezone.

Previous behavior:

In [1]: pd.Index([1, 2, 3]).unique()
Out[1]: array([1, 2, 3])

In [2]: pd.DatetimeIndex(['2011-01-01', '2011-01-02', '2011-01-03'], tz='Asia/Tokyo').unique()
Out[2]:
DatetimeIndex(['2011-01-01 00:00:00+09:00', '2011-01-02 00:00:00+09:00',
               '2011-01-03 00:00:00+09:00'],
              dtype='datetime64[ns, Asia/Tokyo]', freq=None)

New behavior:

In [136]: pd.Index([1, 2, 3]).unique()
Out[136]: Int64Index([1, 2, 3], dtype='int64')

In [137]: pd.DatetimeIndex(['2011-01-01', '2011-01-02', '2011-01-03'], tz='Asia/Tokyo').unique()
Out[137]: 
DatetimeIndex(['2011-01-01 00:00:00+09:00', '2011-01-02 00:00:00+09:00',
               '2011-01-03 00:00:00+09:00'],
              dtype='datetime64[ns, Asia/Tokyo]', freq=None)

MultiIndex constructors, groupby and set_index preserve categorical dtypes

MultiIndex.from_arrays and MultiIndex.from_product will now preserve categorical dtype in MultiIndex levels (GH13743, GH13854).

In [138]: cat = pd.Categorical(['a', 'b'], categories=list("bac"))

In [139]: lvl1 = ['foo', 'bar']

In [140]: midx = pd.MultiIndex.from_arrays([cat, lvl1])

In [141]: midx
Out[141]: 
MultiIndex(levels=[['b', 'a', 'c'], ['bar', 'foo']],
           labels=[[1, 0], [1, 0]])

Previous behavior:

In [4]: midx.levels[0]
Out[4]: Index(['b', 'a', 'c'], dtype='object')

In [5]: midx.get_level_values[0]
Out[5]: Index(['a', 'b'], dtype='object')

New behavior: the single level is now a CategoricalIndex:

In [142]: midx.levels[0]
Out[142]: CategoricalIndex(['b', 'a', 'c'], categories=['b', 'a', 'c'], ordered=False, dtype='category')

In [143]: midx.get_level_values(0)
Out[143]: CategoricalIndex(['a', 'b'], categories=['b', 'a', 'c'], ordered=False, dtype='category')

An analogous change has been made to MultiIndex.from_product. As a consequence, groupby and set_index also preserve categorical dtypes in indexes

In [144]: df = pd.DataFrame({'A': [0, 1], 'B': [10, 11], 'C': cat})

In [145]: df_grouped = df.groupby(by=['A', 'C']).first()

In [146]: df_set_idx = df.set_index(['A', 'C'])

Previous behavior:

In [11]: df_grouped.index.levels[1]
Out[11]: Index(['b', 'a', 'c'], dtype='object', name='C')
In [12]: df_grouped.reset_index().dtypes
Out[12]:
A      int64
C     object
B    float64
dtype: object

In [13]: df_set_idx.index.levels[1]
Out[13]: Index(['b', 'a', 'c'], dtype='object', name='C')
In [14]: df_set_idx.reset_index().dtypes
Out[14]:
A      int64
C     object
B      int64
dtype: object

New behavior:

In [147]: df_grouped.index.levels[1]
Out[147]: CategoricalIndex(['b', 'a', 'c'], categories=['b', 'a', 'c'], ordered=False, name='C', dtype='category')

In [148]: df_grouped.reset_index().dtypes
Out[148]: 
A       int64
C    category
B     float64
dtype: object

In [149]: df_set_idx.index.levels[1]
Out[149]: CategoricalIndex(['b', 'a', 'c'], categories=['b', 'a', 'c'], ordered=False, name='C', dtype='category')

In [150]: df_set_idx.reset_index().dtypes
Out[150]: 
A       int64
C    category
B       int64
dtype: object

read_csv will progressively enumerate chunks

When read_csv() is called with chunksize=n and without specifying an index, each chunk used to have an independently generated index from 0 to n-1. They are now given instead a progressive index, starting from 0 for the first chunk, from n for the second, and so on, so that, when concatenated, they are identical to the result of calling read_csv() without the chunksize= argument (GH12185).

In [151]: data = 'A,B\n0,1\n2,3\n4,5\n6,7'

Previous behavior:

In [2]: pd.concat(pd.read_csv(StringIO(data), chunksize=2))
Out[2]:
   A  B
0  0  1
1  2  3
0  4  5
1  6  7

New behavior:

In [152]: pd.concat(pd.read_csv(StringIO(data), chunksize=2))
Out[152]: 
   A  B
0  0  1
1  2  3
2  4  5
3  6  7

Sparse Changes

These changes allow pandas to handle sparse data with more dtypes, and for work to make a smoother experience with data handling.

int64 and bool support enhancements

Sparse data structures now gained enhanced support of int64 and bool dtype (GH667, GH13849).

Previously, sparse data were float64 dtype by default, even if all inputs were of int or bool dtype. You had to specify dtype explicitly to create sparse data with int64 dtype. Also, fill_value had to be specified explicitly because the default was np.nan which doesn’t appear in int64 or bool data.

In [1]: pd.SparseArray([1, 2, 0, 0])
Out[1]:
[1.0, 2.0, 0.0, 0.0]
Fill: nan
IntIndex
Indices: array([0, 1, 2, 3], dtype=int32)

# specifying int64 dtype, but all values are stored in sp_values because
# fill_value default is np.nan
In [2]: pd.SparseArray([1, 2, 0, 0], dtype=np.int64)
Out[2]:
[1, 2, 0, 0]
Fill: nan
IntIndex
Indices: array([0, 1, 2, 3], dtype=int32)

In [3]: pd.SparseArray([1, 2, 0, 0], dtype=np.int64, fill_value=0)
Out[3]:
[1, 2, 0, 0]
Fill: 0
IntIndex
Indices: array([0, 1], dtype=int32)

As of v0.19.0, sparse data keeps the input dtype, and uses more appropriate fill_value defaults (0 for int64 dtype, False for bool dtype).

In [153]: pd.SparseArray([1, 2, 0, 0], dtype=np.int64)
Out[153]: 
[1, 2, 0, 0]
Fill: 0
IntIndex
Indices: array([0, 1], dtype=int32)

In [154]: pd.SparseArray([True, False, False, False])
Out[154]: 
[True, False, False, False]
Fill: False
IntIndex
Indices: array([0], dtype=int32)

See the docs for more details.

Operators now preserve dtypes
  • Sparse data structure now can preserve dtype after arithmetic ops (GH13848)

    In [155]: s = pd.SparseSeries([0, 2, 0, 1], fill_value=0, dtype=np.int64)
    
    In [156]: s.dtype
    Out[156]: dtype('int64')
    
    In [157]: s + 1
    Out[157]: 
    0    1
    1    3
    2    1
    3    2
    dtype: int64
    BlockIndex
    Block locations: array([1, 3], dtype=int32)
    Block lengths: array([1, 1], dtype=int32)
    
  • Sparse data structure now support astype to convert internal dtype (GH13900)

    In [158]: s = pd.SparseSeries([1., 0., 2., 0.], fill_value=0)
    
    In [159]: s
    Out[159]: 
    0    1.0
    1    0.0
    2    2.0
    3    0.0
    dtype: float64
    BlockIndex
    Block locations: array([0, 2], dtype=int32)
    Block lengths: array([1, 1], dtype=int32)
    
    In [160]: s.astype(np.int64)
    Out[160]: 
    0    1
    1    0
    2    2
    3    0
    dtype: int64
    BlockIndex
    Block locations: array([0, 2], dtype=int32)
    Block lengths: array([1, 1], dtype=int32)
    

    astype fails if data contains values which cannot be converted to specified dtype. Note that the limitation is applied to fill_value which default is np.nan.

    In [7]: pd.SparseSeries([1., np.nan, 2., np.nan], fill_value=np.nan).astype(np.int64)
    Out[7]:
    ValueError: unable to coerce current fill_value nan to int64 dtype
    
Other sparse fixes
  • Subclassed SparseDataFrame and SparseSeries now preserve class types when slicing or transposing. (GH13787)
  • SparseArray with bool dtype now supports logical (bool) operators (GH14000)
  • Bug in SparseSeries with MultiIndex [] indexing may raise IndexError (GH13144)
  • Bug in SparseSeries with MultiIndex [] indexing result may have normal Index (GH13144)
  • Bug in SparseDataFrame in which axis=None did not default to axis=0 (GH13048)
  • Bug in SparseSeries and SparseDataFrame creation with object dtype may raise TypeError (GH11633)
  • Bug in SparseDataFrame doesn’t respect passed SparseArray or SparseSeries ‘s dtype and fill_value (GH13866)
  • Bug in SparseArray and SparseSeries don’t apply ufunc to fill_value (GH13853)
  • Bug in SparseSeries.abs incorrectly keeps negative fill_value (GH13853)
  • Bug in single row slicing on multi-type SparseDataFrame s, types were previously forced to float (GH13917)
  • Bug in SparseSeries slicing changes integer dtype to float (GH8292)
  • Bug in SparseDataFarme comparison ops may raise TypeError (GH13001)
  • Bug in SparseDataFarme.isnull raises ValueError (GH8276)
  • Bug in SparseSeries representation with bool dtype may raise IndexError (GH13110)
  • Bug in SparseSeries and SparseDataFrame of bool or int64 dtype may display its values like float64 dtype (GH13110)
  • Bug in sparse indexing using SparseArray with bool dtype may return incorrect result (GH13985)
  • Bug in SparseArray created from SparseSeries may lose dtype (GH13999)
  • Bug in SparseSeries comparison with dense returns normal Series rather than SparseSeries (GH13999)

Indexer dtype changes

Note

This change only affects 64 bit python running on Windows, and only affects relatively advanced indexing operations

Methods such as Index.get_indexer that return an indexer array, coerce that array to a “platform int”, so that it can be directly used in 3rd party library operations like numpy.take. Previously, a platform int was defined as np.int_ which corresponds to a C integer, but the correct type, and what is being used now, is np.intp, which corresponds to the C integer size that can hold a pointer (GH3033, GH13972).

These types are the same on many platform, but for 64 bit python on Windows, np.int_ is 32 bits, and np.intp is 64 bits. Changing this behavior improves performance for many operations on that platform.

Previous behavior:

In [1]: i = pd.Index(['a', 'b', 'c'])

In [2]: i.get_indexer(['b', 'b', 'c']).dtype
Out[2]: dtype('int32')

New behavior:

In [1]: i = pd.Index(['a', 'b', 'c'])

In [2]: i.get_indexer(['b', 'b', 'c']).dtype
Out[2]: dtype('int64')

Other API Changes

  • Timestamp.to_pydatetime will issue a UserWarning when warn=True, and the instance has a non-zero number of nanoseconds, previously this would print a message to stdout (GH14101).
  • Series.unique() with datetime and timezone now returns return array of Timestamp with timezone (GH13565).
  • Panel.to_sparse() will raise a NotImplementedError exception when called (GH13778).
  • Index.reshape() will raise a NotImplementedError exception when called (GH12882).
  • .filter() enforces mutual exclusion of the keyword arguments (GH12399).
  • eval’s upcasting rules for float32 types have been updated to be more consistent with NumPy’s rules. New behavior will not upcast to float64 if you multiply a pandas float32 object by a scalar float64 (GH12388).
  • An UnsupportedFunctionCall error is now raised if NumPy ufuncs like np.mean are called on groupby or resample objects (GH12811).
  • __setitem__ will no longer apply a callable rhs as a function instead of storing it. Call where directly to get the previous behavior (GH13299).
  • Calls to .sample() will respect the random seed set via numpy.random.seed(n) (GH13161)
  • Styler.apply is now more strict about the outputs your function must return. For axis=0 or axis=1, the output shape must be identical. For axis=None, the output must be a DataFrame with identical columns and index labels (GH13222).
  • Float64Index.astype(int) will now raise ValueError if Float64Index contains NaN values (GH13149)
  • TimedeltaIndex.astype(int) and DatetimeIndex.astype(int) will now return Int64Index instead of np.array (GH13209)
  • Passing Period with multiple frequencies to normal Index now returns Index with object dtype (GH13664)
  • PeriodIndex.fillna with Period has different freq now coerces to object dtype (GH13664)
  • Faceted boxplots from DataFrame.boxplot(by=col) now return a Series when return_type is not None. Previously these returned an OrderedDict. Note that when return_type=None, the default, these still return a 2-D NumPy array (GH12216, GH7096).
  • pd.read_hdf will now raise a ValueError instead of KeyError, if a mode other than r, r+ and a is supplied. (GH13623)
  • pd.read_csv(), pd.read_table(), and pd.read_hdf() raise the builtin FileNotFoundError exception for Python 3.x when called on a nonexistent file; this is back-ported as IOError in Python 2.x (GH14086)
  • More informative exceptions are passed through the csv parser. The exception type would now be the original exception type instead of CParserError (GH13652).
  • pd.read_csv() in the C engine will now issue a ParserWarning or raise a ValueError when sep encoded is more than one character long (GH14065)
  • DataFrame.values will now return float64 with a DataFrame of mixed int64 and uint64 dtypes, conforming to np.find_common_type (GH10364, GH13917)
  • .groupby.groups will now return a dictionary of Index objects, rather than a dictionary of np.ndarray or lists (GH14293)

Deprecations

  • Series.reshape and Categorical.reshape have been deprecated and will be removed in a subsequent release (GH12882, GH12882)
  • PeriodIndex.to_datetime has been deprecated in favor of PeriodIndex.to_timestamp (GH8254)
  • Timestamp.to_datetime has been deprecated in favor of Timestamp.to_pydatetime (GH8254)
  • Index.to_datetime and DatetimeIndex.to_datetime have been deprecated in favor of pd.to_datetime (GH8254)
  • pandas.core.datetools module has been deprecated and will be removed in a subsequent release (GH14094)
  • SparseList has been deprecated and will be removed in a future version (GH13784)
  • DataFrame.to_html() and DataFrame.to_latex() have dropped the colSpace parameter in favor of col_space (GH13857)
  • DataFrame.to_sql() has deprecated the flavor parameter, as it is superfluous when SQLAlchemy is not installed (GH13611)
  • Deprecated read_csv keywords:
    • compact_ints and use_unsigned have been deprecated and will be removed in a future version (GH13320)
    • buffer_lines has been deprecated and will be removed in a future version (GH13360)
    • as_recarray has been deprecated and will be removed in a future version (GH13373)
    • skip_footer has been deprecated in favor of skipfooter and will be removed in a future version (GH13349)
  • top-level pd.ordered_merge() has been renamed to pd.merge_ordered() and the original name will be removed in a future version (GH13358)
  • Timestamp.offset property (and named arg in the constructor), has been deprecated in favor of freq (GH12160)
  • pd.tseries.util.pivot_annual is deprecated. Use pivot_table as alternative, an example is here (GH736)
  • pd.tseries.util.isleapyear has been deprecated and will be removed in a subsequent release. Datetime-likes now have a .is_leap_year property (GH13727)
  • Panel4D and PanelND constructors are deprecated and will be removed in a future version. The recommended way to represent these types of n-dimensional data are with the xarray package. Pandas provides a to_xarray() method to automate this conversion (GH13564).
  • pandas.tseries.frequencies.get_standard_freq is deprecated. Use pandas.tseries.frequencies.to_offset(freq).rule_code instead (GH13874)
  • pandas.tseries.frequencies.to_offset’s freqstr keyword is deprecated in favor of freq (GH13874)
  • Categorical.from_array has been deprecated and will be removed in a future version (GH13854)

Removal of prior version deprecations/changes

  • The SparsePanel class has been removed (GH13778)
  • The pd.sandbox module has been removed in favor of the external library pandas-qt (GH13670)
  • The pandas.io.data and pandas.io.wb modules are removed in favor of the pandas-datareader package (GH13724).
  • The pandas.tools.rplot module has been removed in favor of the seaborn package (GH13855)
  • DataFrame.to_csv() has dropped the engine parameter, as was deprecated in 0.17.1 (GH11274, GH13419)
  • DataFrame.to_dict() has dropped the outtype parameter in favor of orient (GH13627, GH8486)
  • pd.Categorical has dropped setting of the ordered attribute directly in favor of the set_ordered method (GH13671)
  • pd.Categorical has dropped the levels attribute in favor of categories (GH8376)
  • DataFrame.to_sql() has dropped the mysql option for the flavor parameter (GH13611)
  • Panel.shift() has dropped the lags parameter in favor of periods (GH14041)
  • pd.Index has dropped the diff method in favor of difference (GH13669)
  • pd.DataFrame has dropped the to_wide method in favor of to_panel (GH14039)
  • Series.to_csv has dropped the nanRep parameter in favor of na_rep (GH13804)
  • Series.xs, DataFrame.xs, Panel.xs, Panel.major_xs, and Panel.minor_xs have dropped the copy parameter (GH13781)
  • str.split has dropped the return_type parameter in favor of expand (GH13701)
  • Removal of the legacy time rules (offset aliases), deprecated since 0.17.0 (this has been alias since 0.8.0) (GH13590, GH13868). Now legacy time rules raises ValueError. For the list of currently supported offsets, see here.
  • The default value for the return_type parameter for DataFrame.plot.box and DataFrame.boxplot changed from None to "axes". These methods will now return a matplotlib axes by default instead of a dictionary of artists. See here (GH6581).
  • The tquery and uquery functions in the pandas.io.sql module are removed (GH5950).

Performance Improvements

  • Improved performance of sparse IntIndex.intersect (GH13082)
  • Improved performance of sparse arithmetic with BlockIndex when the number of blocks are large, though recommended to use IntIndex in such cases (GH13082)
  • Improved performance of DataFrame.quantile() as it now operates per-block (GH11623)
  • Improved performance of float64 hash table operations, fixing some very slow indexing and groupby operations in python 3 (GH13166, GH13334)
  • Improved performance of DataFrameGroupBy.transform (GH12737)
  • Improved performance of Index and Series .duplicated (GH10235)
  • Improved performance of Index.difference (GH12044)
  • Improved performance of RangeIndex.is_monotonic_increasing and is_monotonic_decreasing (GH13749)
  • Improved performance of datetime string parsing in DatetimeIndex (GH13692)
  • Improved performance of hashing Period (GH12817)
  • Improved performance of factorize of datetime with timezone (GH13750)
  • Improved performance of by lazily creating indexing hashtables on larger Indexes (GH14266)
  • Improved performance of groupby.groups (GH14293)
  • Unnecessary materializing of a MultiIndex when introspecting for memory usage (GH14308)

Bug Fixes

  • Bug in groupby().shift(), which could cause a segfault or corruption in rare circumstances when grouping by columns with missing values (GH13813)
  • Bug in groupby().cumsum() calculating cumprod when axis=1. (GH13994)
  • Bug in pd.to_timedelta() in which the errors parameter was not being respected (GH13613)
  • Bug in io.json.json_normalize(), where non-ascii keys raised an exception (GH13213)
  • Bug when passing a not-default-indexed Series as xerr or yerr in .plot() (GH11858)
  • Bug in area plot draws legend incorrectly if subplot is enabled or legend is moved after plot (matplotlib 1.5.0 is required to draw area plot legend properly) (GH9161, GH13544)
  • Bug in DataFrame assignment with an object-dtyped Index where the resultant column is mutable to the original object. (GH13522)
  • Bug in matplotlib AutoDataFormatter; this restores the second scaled formatting and re-adds micro-second scaled formatting (GH13131)
  • Bug in selection from a HDFStore with a fixed format and start and/or stop specified will now return the selected range (GH8287)
  • Bug in Categorical.from_codes() where an unhelpful error was raised when an invalid ordered parameter was passed in (GH14058)
  • Bug in Series construction from a tuple of integers on windows not returning default dtype (int64) (GH13646)
  • Bug in TimedeltaIndex addition with a Datetime-like object where addition overflow was not being caught (GH14068)
  • Bug in .groupby(..).resample(..) when the same object is called multiple times (GH13174)
  • Bug in .to_records() when index name is a unicode string (GH13172)
  • Bug in calling .memory_usage() on object which doesn’t implement (GH12924)
  • Regression in Series.quantile with nans (also shows up in .median() and .describe() ); furthermore now names the Series with the quantile (GH13098, GH13146)
  • Bug in SeriesGroupBy.transform with datetime values and missing groups (GH13191)
  • Bug where empty Series were incorrectly coerced in datetime-like numeric operations (GH13844)
  • Bug in Categorical constructor when passed a Categorical containing datetimes with timezones (GH14190)
  • Bug in Series.str.extractall() with str index raises ValueError (GH13156)
  • Bug in Series.str.extractall() with single group and quantifier (GH13382)
  • Bug in DatetimeIndex and Period subtraction raises ValueError or AttributeError rather than TypeError (GH13078)
  • Bug in Index and Series created with NaN and NaT mixed data may not have datetime64 dtype (GH13324)
  • Bug in Index and Series may ignore np.datetime64('nat') and np.timdelta64('nat') to infer dtype (GH13324)
  • Bug in PeriodIndex and Period subtraction raises AttributeError (GH13071)
  • Bug in PeriodIndex construction returning a float64 index in some circumstances (GH13067)
  • Bug in .resample(..) with a PeriodIndex not changing its freq appropriately when empty (GH13067)
  • Bug in .resample(..) with a PeriodIndex not retaining its type or name with an empty DataFrame appropriately when empty (GH13212)
  • Bug in groupby(..).apply(..) when the passed function returns scalar values per group (GH13468).
  • Bug in groupby(..).resample(..) where passing some keywords would raise an exception (GH13235)
  • Bug in .tz_convert on a tz-aware DateTimeIndex that relied on index being sorted for correct results (GH13306)
  • Bug in .tz_localize with dateutil.tz.tzlocal may return incorrect result (GH13583)
  • Bug in DatetimeTZDtype dtype with dateutil.tz.tzlocal cannot be regarded as valid dtype (GH13583)
  • Bug in pd.read_hdf() where attempting to load an HDF file with a single dataset, that had one or more categorical columns, failed unless the key argument was set to the name of the dataset. (GH13231)
  • Bug in .rolling() that allowed a negative integer window in construction of the Rolling() object, but would later fail on aggregation (GH13383)
  • Bug in Series indexing with tuple-valued data and a numeric index (GH13509)
  • Bug in printing pd.DataFrame where unusual elements with the object dtype were causing segfaults (GH13717)
  • Bug in ranking Series which could result in segfaults (GH13445)
  • Bug in various index types, which did not propagate the name of passed index (GH12309)
  • Bug in DatetimeIndex, which did not honour the copy=True (GH13205)
  • Bug in DatetimeIndex.is_normalized returns incorrectly for normalized date_range in case of local timezones (GH13459)
  • Bug in pd.concat and .append may coerces datetime64 and timedelta to object dtype containing python built-in datetime or timedelta rather than Timestamp or Timedelta (GH13626)
  • Bug in PeriodIndex.append may raises AttributeError when the result is object dtype (GH13221)
  • Bug in CategoricalIndex.append may accept normal list (GH13626)
  • Bug in pd.concat and .append with the same timezone get reset to UTC (GH7795)
  • Bug in Series and DataFrame .append raises AmbiguousTimeError if data contains datetime near DST boundary (GH13626)
  • Bug in DataFrame.to_csv() in which float values were being quoted even though quotations were specified for non-numeric values only (GH12922, GH13259)
  • Bug in DataFrame.describe() raising ValueError with only boolean columns (GH13898)
  • Bug in MultiIndex slicing where extra elements were returned when level is non-unique (GH12896)
  • Bug in .str.replace does not raise TypeError for invalid replacement (GH13438)
  • Bug in MultiIndex.from_arrays which didn’t check for input array lengths matching (GH13599)
  • Bug in cartesian_product and MultiIndex.from_product which may raise with empty input arrays (GH12258)
  • Bug in pd.read_csv() which may cause a segfault or corruption when iterating in large chunks over a stream/file under rare circumstances (GH13703)
  • Bug in pd.read_csv() which caused errors to be raised when a dictionary containing scalars is passed in for na_values (GH12224)
  • Bug in pd.read_csv() which caused BOM files to be incorrectly parsed by not ignoring the BOM (GH4793)
  • Bug in pd.read_csv() with engine='python' which raised errors when a numpy array was passed in for usecols (GH12546)
  • Bug in pd.read_csv() where the index columns were being incorrectly parsed when parsed as dates with a thousands parameter (GH14066)
  • Bug in pd.read_csv() with engine='python' in which NaN values weren’t being detected after data was converted to numeric values (GH13314)
  • Bug in pd.read_csv() in which the nrows argument was not properly validated for both engines (GH10476)
  • Bug in pd.read_csv() with engine='python' in which infinities of mixed-case forms were not being interpreted properly (GH13274)
  • Bug in pd.read_csv() with engine='python' in which trailing NaN values were not being parsed (GH13320)
  • Bug in pd.read_csv() with engine='python' when reading from a tempfile.TemporaryFile on Windows with Python 3 (GH13398)
  • Bug in pd.read_csv() that prevents usecols kwarg from accepting single-byte unicode strings (GH13219)
  • Bug in pd.read_csv() that prevents usecols from being an empty set (GH13402)
  • Bug in pd.read_csv() in the C engine where the NULL character was not being parsed as NULL (GH14012)
  • Bug in pd.read_csv() with engine='c' in which NULL quotechar was not accepted even though quoting was specified as None (GH13411)
  • Bug in pd.read_csv() with engine='c' in which fields were not properly cast to float when quoting was specified as non-numeric (GH13411)
  • Bug in pd.read_csv() in Python 2.x with non-UTF8 encoded, multi-character separated data (GH3404)
  • Bug in pd.read_csv(), where aliases for utf-xx (e.g. UTF-xx, UTF_xx, utf_xx) raised UnicodeDecodeError (GH13549)
  • Bug in pd.read_csv, pd.read_table, pd.read_fwf, pd.read_stata and pd.read_sas where files were opened by parsers but not closed if both chunksize and iterator were None. (GH13940)
  • Bug in StataReader, StataWriter, XportReader and SAS7BDATReader where a file was not properly closed when an error was raised. (GH13940)
  • Bug in pd.pivot_table() where margins_name is ignored when aggfunc is a list (GH13354)
  • Bug in pd.Series.str.zfill, center, ljust, rjust, and pad when passing non-integers, did not raise TypeError (GH13598)
  • Bug in checking for any null objects in a TimedeltaIndex, which always returned True (GH13603)
  • Bug in Series arithmetic raises TypeError if it contains datetime-like as object dtype (GH13043)
  • Bug Series.isnull() and Series.notnull() ignore Period('NaT') (GH13737)
  • Bug Series.fillna() and Series.dropna() don’t affect to Period('NaT') (GH13737
  • Bug in .fillna(value=np.nan) incorrectly raises KeyError on a category dtyped Series (GH14021)
  • Bug in extension dtype creation where the created types were not is/identical (GH13285)
  • Bug in .resample(..) where incorrect warnings were triggered by IPython introspection (GH13618)
  • Bug in NaT - Period raises AttributeError (GH13071)
  • Bug in Series comparison may output incorrect result if rhs contains NaT (GH9005)
  • Bug in Series and Index comparison may output incorrect result if it contains NaT with object dtype (GH13592)
  • Bug in Period addition raises TypeError if Period is on right hand side (GH13069)
  • Bug in Peirod and Series or Index comparison raises TypeError (GH13200)
  • Bug in pd.set_eng_float_format() that would prevent NaN and Inf from formatting (GH11981)
  • Bug in .unstack with Categorical dtype resets .ordered to True (GH13249)
  • Clean some compile time warnings in datetime parsing (GH13607)
  • Bug in factorize raises AmbiguousTimeError if data contains datetime near DST boundary (GH13750)
  • Bug in .set_index raises AmbiguousTimeError if new index contains DST boundary and multi levels (GH12920)
  • Bug in .shift raises AmbiguousTimeError if data contains datetime near DST boundary (GH13926)
  • Bug in pd.read_hdf() returns incorrect result when a DataFrame with a categorical column and a query which doesn’t match any values (GH13792)
  • Bug in .iloc when indexing with a non lex-sorted MultiIndex (GH13797)
  • Bug in .loc when indexing with date strings in a reverse sorted DatetimeIndex (GH14316)
  • Bug in Series comparison operators when dealing with zero dim NumPy arrays (GH13006)
  • Bug in .combine_first may return incorrect dtype (GH7630, GH10567)
  • Bug in groupby where apply returns different result depending on whether first result is None or not (GH12824)
  • Bug in groupby(..).nth() where the group key is included inconsistently if called after .head()/.tail() (GH12839)
  • Bug in .to_html, .to_latex and .to_string silently ignore custom datetime formatter passed through the formatters key word (GH10690)
  • Bug in DataFrame.iterrows(), not yielding a Series subclasse if defined (GH13977)
  • Bug in pd.to_numeric when errors='coerce' and input contains non-hashable objects (GH13324)
  • Bug in invalid Timedelta arithmetic and comparison may raise ValueError rather than TypeError (GH13624)
  • Bug in invalid datetime parsing in to_datetime and DatetimeIndex may raise TypeError rather than ValueError (GH11169, GH11287)
  • Bug in Index created with tz-aware Timestamp and mismatched tz option incorrectly coerces timezone (GH13692)
  • Bug in DatetimeIndex with nanosecond frequency does not include timestamp specified with end (GH13672)
  • Bug in `Series when setting a slice with a np.timedelta64 (GH14155)
  • Bug in Index raises OutOfBoundsDatetime if datetime exceeds datetime64[ns] bounds, rather than coercing to object dtype (GH13663)
  • Bug in Index may ignore specified datetime64 or timedelta64 passed as dtype (GH13981)
  • Bug in RangeIndex can be created without no arguments rather than raises TypeError (GH13793)
  • Bug in .value_counts() raises OutOfBoundsDatetime if data exceeds datetime64[ns] bounds (GH13663)
  • Bug in DatetimeIndex may raise OutOfBoundsDatetime if input np.datetime64 has other unit than ns (GH9114)
  • Bug in Series creation with np.datetime64 which has other unit than ns as object dtype results in incorrect values (GH13876)
  • Bug in resample with timedelta data where data was casted to float (GH13119).
  • Bug in pd.isnull() pd.notnull() raise TypeError if input datetime-like has other unit than ns (GH13389)
  • Bug in pd.merge() may raise TypeError if input datetime-like has other unit than ns (GH13389)
  • Bug in HDFStore/read_hdf() discarded DatetimeIndex.name if tz was set (GH13884)
  • Bug in Categorical.remove_unused_categories() changes .codes dtype to platform int (GH13261)
  • Bug in groupby with as_index=False returns all NaN’s when grouping on multiple columns including a categorical one (GH13204)
  • Bug in df.groupby(...)[...] where getitem with Int64Index raised an error (GH13731)
  • Bug in the CSS classes assigned to DataFrame.style for index names. Previously they were assigned "col_heading level<n> col<c>" where n was the number of levels + 1. Now they are assigned "index_name level<n>", where n is the correct level for that MultiIndex.
  • Bug where pd.read_gbq() could throw ImportError: No module named discovery as a result of a naming conflict with another python package called apiclient (GH13454)
  • Bug in Index.union returns an incorrect result with a named empty index (GH13432)
  • Bugs in Index.difference and DataFrame.join raise in Python3 when using mixed-integer indexes (GH13432, GH12814)
  • Bug in subtract tz-aware datetime.datetime from tz-aware datetime64 series (GH14088)
  • Bug in .to_excel() when DataFrame contains a MultiIndex which contains a label with a NaN value (GH13511)
  • Bug in invalid frequency offset string like “D1”, “-2-3H” may not raise ValueError (GH13930)
  • Bug in concat and groupby for hierarchical frames with RangeIndex levels (GH13542).
  • Bug in Series.str.contains() for Series containing only NaN values of object dtype (GH14171)
  • Bug in agg() function on groupby dataframe changes dtype of datetime64[ns] column to float64 (GH12821)
  • Bug in using NumPy ufunc with PeriodIndex to add or subtract integer raise IncompatibleFrequency. Note that using standard operator like + or - is recommended, because standard operators use more efficient path (GH13980)
  • Bug in operations on NaT returning float instead of datetime64[ns] (GH12941)
  • Bug in Series flexible arithmetic methods (like .add()) raises ValueError when axis=None (GH13894)
  • Bug in DataFrame.to_csv() with MultiIndex columns in which a stray empty line was added (GH6618)
  • Bug in DatetimeIndex, TimedeltaIndex and PeriodIndex.equals() may return True when input isn’t Index but contains the same values (GH13107)
  • Bug in assignment against datetime with timezone may not work if it contains datetime near DST boundary (GH14146)
  • Bug in pd.eval() and HDFStore query truncating long float literals with python 2 (GH14241)
  • Bug in Index raises KeyError displaying incorrect column when column is not in the df and columns contains duplicate values (GH13822)
  • Bug in Period and PeriodIndex creating wrong dates when frequency has combined offset aliases (GH13874)
  • Bug in .to_string() when called with an integer line_width and index=False raises an UnboundLocalError exception because idx referenced before assignment.
  • Bug in eval() where the resolvers argument would not accept a list (GH14095)
  • Bugs in stack, get_dummies, make_axis_dummies which don’t preserve categorical dtypes in (multi)indexes (GH13854)
  • PeriodIndex can now accept list and array which contains pd.NaT (GH13430)
  • Bug in df.groupby where .median() returns arbitrary values if grouped dataframe contains empty bins (GH13629)
  • Bug in Index.copy() where name parameter was ignored (GH14302)

v0.18.1 (May 3, 2016)

This is a minor bug-fix release from 0.18.0 and includes a large number of bug fixes along with several new features, enhancements, and performance improvements. We recommend that all users upgrade to this version.

Highlights include:

  • .groupby(...) has been enhanced to provide convenient syntax when working with .rolling(..), .expanding(..) and .resample(..) per group, see here
  • pd.to_datetime() has gained the ability to assemble dates from a DataFrame, see here
  • Method chaining improvements, see here.
  • Custom business hour offset, see here.
  • Many bug fixes in the handling of sparse, see here
  • Expanded the Tutorials section with a feature on modern pandas, courtesy of @TomAugsburger. (GH13045).

New features

Custom Business Hour

The CustomBusinessHour is a mixture of BusinessHour and CustomBusinessDay which allows you to specify arbitrary holidays. For details, see Custom Business Hour (GH11514)

In [1]: from pandas.tseries.offsets import CustomBusinessHour

In [2]: from pandas.tseries.holiday import USFederalHolidayCalendar

In [3]: bhour_us = CustomBusinessHour(calendar=USFederalHolidayCalendar())

Friday before MLK Day

In [4]: dt = datetime(2014, 1, 17, 15)

In [5]: dt + bhour_us
Out[5]: Timestamp('2014-01-17 16:00:00')

Tuesday after MLK Day (Monday is skipped because it’s a holiday)

In [6]: dt + bhour_us * 2
Out[6]: Timestamp('2014-01-20 09:00:00')

.groupby(..) syntax with window and resample operations

.groupby(...) has been enhanced to provide convenient syntax when working with .rolling(..), .expanding(..) and .resample(..) per group, see (GH12486, GH12738).

You can now use .rolling(..) and .expanding(..) as methods on groupbys. These return another deferred object (similar to what .rolling() and .expanding() do on ungrouped pandas objects). You can then operate on these RollingGroupby objects in a similar manner.

Previously you would have to do this to get a rolling window mean per-group:

In [7]: df = pd.DataFrame({'A': [1] * 20 + [2] * 12 + [3] * 8,
   ...:                    'B': np.arange(40)})
   ...: 

In [8]: df
Out[8]: 
    A   B
0   1   0
1   1   1
2   1   2
3   1   3
4   1   4
5   1   5
6   1   6
.. ..  ..
33  3  33
34  3  34
35  3  35
36  3  36
37  3  37
38  3  38
39  3  39

[40 rows x 2 columns]
In [9]: df.groupby('A').apply(lambda x: x.rolling(4).B.mean())
Out[9]: 
A    
1  0      NaN
   1      NaN
   2      NaN
   3      1.5
   4      2.5
   5      3.5
   6      4.5
         ... 
3  33     NaN
   34     NaN
   35    33.5
   36    34.5
   37    35.5
   38    36.5
   39    37.5
Name: B, Length: 40, dtype: float64

Now you can do:

In [10]: df.groupby('A').rolling(4).B.mean()
Out[10]: 
A    
1  0      NaN
   1      NaN
   2      NaN
   3      1.5
   4      2.5
   5      3.5
   6      4.5
         ... 
3  33     NaN
   34     NaN
   35    33.5
   36    34.5
   37    35.5
   38    36.5
   39    37.5
Name: B, Length: 40, dtype: float64

For .resample(..) type of operations, previously you would have to:

In [11]: df = pd.DataFrame({'date': pd.date_range(start='2016-01-01',
   ....:                                          periods=4,
   ....:                                          freq='W'),
   ....:                    'group': [1, 1, 2, 2],
   ....:                    'val': [5, 6, 7, 8]}).set_index('date')
   ....: 

In [12]: df
Out[12]: 
            group  val
date                  
2016-01-03      1    5
2016-01-10      1    6
2016-01-17      2    7
2016-01-24      2    8
In [13]: df.groupby('group').apply(lambda x: x.resample('1D').ffill())
Out[13]: 
                  group  val
group date                  
1     2016-01-03      1    5
      2016-01-04      1    5
      2016-01-05      1    5
      2016-01-06      1    5
      2016-01-07      1    5
      2016-01-08      1    5
      2016-01-09      1    5
...                 ...  ...
2     2016-01-18      2    7
      2016-01-19      2    7
      2016-01-20      2    7
      2016-01-21      2    7
      2016-01-22      2    7
      2016-01-23      2    7
      2016-01-24      2    8

[16 rows x 2 columns]

Now you can do:

In [14]: df.groupby('group').resample('1D').ffill()
Out[14]: 
                  group  val
group date                  
1     2016-01-03      1    5
      2016-01-04      1    5
      2016-01-05      1    5
      2016-01-06      1    5
      2016-01-07      1    5
      2016-01-08      1    5
      2016-01-09      1    5
...                 ...  ...
2     2016-01-18      2    7
      2016-01-19      2    7
      2016-01-20      2    7
      2016-01-21      2    7
      2016-01-22      2    7
      2016-01-23      2    7
      2016-01-24      2    8

[16 rows x 2 columns]

Method chaininng improvements

The following methods / indexers now accept a callable. It is intended to make these more useful in method chains, see the documentation. (GH11485, GH12533)

  • .where() and .mask()
  • .loc[], iloc[] and .ix[]
  • [] indexing
.where() and .mask()

These can accept a callable for the condition and other arguments.

In [15]: df = pd.DataFrame({'A': [1, 2, 3],
   ....:                    'B': [4, 5, 6],
   ....:                    'C': [7, 8, 9]})
   ....: 

In [16]: df.where(lambda x: x > 4, lambda x: x + 10)
Out[16]: 
    A   B  C
0  11  14  7
1  12   5  8
2  13   6  9
.loc[], .iloc[], .ix[]

These can accept a callable, and a tuple of callable as a slicer. The callable can return a valid boolean indexer or anything which is valid for these indexer’s input.

# callable returns bool indexer
In [17]: df.loc[lambda x: x.A >= 2, lambda x: x.sum() > 10]
Out[17]: 
   B  C
1  5  8
2  6  9

# callable returns list of labels
In [18]: df.loc[lambda x: [1, 2], lambda x: ['A', 'B']]
Out[18]: 
   A  B
1  2  5
2  3  6
[] indexing

Finally, you can use a callable in [] indexing of Series, DataFrame and Panel. The callable must return a valid input for [] indexing depending on its class and index type.

In [19]: df[lambda x: 'A']
Out[19]: 
0    1
1    2
2    3
Name: A, dtype: int64

Using these methods / indexers, you can chain data selection operations without using temporary variable.

In [20]: bb = pd.read_csv('data/baseball.csv', index_col='id')

In [21]: (bb.groupby(['year', 'team'])
   ....:    .sum()
   ....:    .loc[lambda df: df.r > 100]
   ....: )
   ....: 
Out[21]: 
           stint    g    ab    r    h  X2b  X3b  hr    rbi    sb   cs   bb     so   ibb   hbp    sh    sf  gidp
year team                                                                                                      
2007 CIN       6  379   745  101  203   35    2  36  125.0  10.0  1.0  105  127.0  14.0   1.0   1.0  15.0  18.0
     DET       5  301  1062  162  283   54    4  37  144.0  24.0  7.0   97  176.0   3.0  10.0   4.0   8.0  28.0
     HOU       4  311   926  109  218   47    6  14   77.0  10.0  4.0   60  212.0   3.0   9.0  16.0   6.0  17.0
     LAN      11  413  1021  153  293   61    3  36  154.0   7.0  5.0  114  141.0   8.0   9.0   3.0   8.0  29.0
     NYN      13  622  1854  240  509  101    3  61  243.0  22.0  4.0  174  310.0  24.0  23.0  18.0  15.0  48.0
     SFN       5  482  1305  198  337   67    6  40  171.0  26.0  7.0  235  188.0  51.0   8.0  16.0   6.0  41.0
     TEX       2  198   729  115  200   40    4  28  115.0  21.0  4.0   73  140.0   4.0   5.0   2.0   8.0  16.0
     TOR       4  459  1408  187  378   96    2  58  223.0   4.0  2.0  190  265.0  16.0  12.0   4.0  16.0  38.0

Partial string indexing on DateTimeIndex when part of a MultiIndex

Partial string indexing now matches on DateTimeIndex when part of a MultiIndex (GH10331)

In [22]: dft2 = pd.DataFrame(np.random.randn(20, 1),
   ....:                     columns=['A'],
   ....:                     index=pd.MultiIndex.from_product([pd.date_range('20130101',
   ....:                                                                     periods=10,
   ....:                                                                     freq='12H'),
   ....:                                                      ['a', 'b']]))
   ....: 

In [23]: dft2
Out[23]: 
                              A
2013-01-01 00:00:00 a  0.156998
                    b -0.571455
2013-01-01 12:00:00 a  1.057633
                    b -0.791489
2013-01-02 00:00:00 a -0.524627
                    b  0.071878
2013-01-02 12:00:00 a  1.910759
...                         ...
2013-01-04 00:00:00 b  1.015405
2013-01-04 12:00:00 a  0.749185
                    b -0.675521
2013-01-05 00:00:00 a  0.440266
                    b  0.688972
2013-01-05 12:00:00 a -0.276646
                    b  1.924533

[20 rows x 1 columns]

In [24]: dft2.loc['2013-01-05']
Out[24]: 
                              A
2013-01-05 00:00:00 a  0.440266
                    b  0.688972
2013-01-05 12:00:00 a -0.276646
                    b  1.924533

On other levels

In [25]: idx = pd.IndexSlice

In [26]: dft2 = dft2.swaplevel(0, 1).sort_index()

In [27]: dft2
Out[27]: 
                              A
a 2013-01-01 00:00:00  0.156998
  2013-01-01 12:00:00  1.057633
  2013-01-02 00:00:00 -0.524627
  2013-01-02 12:00:00  1.910759
  2013-01-03 00:00:00  0.513082
  2013-01-03 12:00:00  1.043945
  2013-01-04 00:00:00  1.459927
...                         ...
b 2013-01-02 12:00:00  0.787965
  2013-01-03 00:00:00 -0.546416
  2013-01-03 12:00:00  2.107785
  2013-01-04 00:00:00  1.015405
  2013-01-04 12:00:00 -0.675521
  2013-01-05 00:00:00  0.688972
  2013-01-05 12:00:00  1.924533

[20 rows x 1 columns]

In [28]: dft2.loc[idx[:, '2013-01-05'], :]
Out[28]: 
                              A
a 2013-01-05 00:00:00  0.440266
  2013-01-05 12:00:00 -0.276646
b 2013-01-05 00:00:00  0.688972
  2013-01-05 12:00:00  1.924533

Assembling Datetimes

pd.to_datetime() has gained the ability to assemble datetimes from a passed in DataFrame or a dict. (GH8158).

In [29]: df = pd.DataFrame({'year': [2015, 2016],
   ....:                    'month': [2, 3],
   ....:                    'day': [4, 5],
   ....:                    'hour': [2, 3]})
   ....: 

In [30]: df
Out[30]: 
   year  month  day  hour
0  2015      2    4     2
1  2016      3    5     3

Assembling using the passed frame.

In [31]: pd.to_datetime(df)
Out[31]: 
0   2015-02-04 02:00:00
1   2016-03-05 03:00:00
dtype: datetime64[ns]

You can pass only the columns that you need to assemble.

In [32]: pd.to_datetime(df[['year', 'month', 'day']])
Out[32]: 
0   2015-02-04
1   2016-03-05
dtype: datetime64[ns]

Other Enhancements

  • pd.read_csv() now supports delim_whitespace=True for the Python engine (GH12958)

  • pd.read_csv() now supports opening ZIP files that contains a single CSV, via extension inference or explicit compression='zip' (GH12175)

  • pd.read_csv() now supports opening files using xz compression, via extension inference or explicit compression='xz' is specified; xz compressions is also supported by DataFrame.to_csv in the same way (GH11852)

  • pd.read_msgpack() now always gives writeable ndarrays even when compression is used (GH12359).

  • pd.read_msgpack() now supports serializing and de-serializing categoricals with msgpack (GH12573)

  • .to_json() now supports NDFrames that contain categorical and sparse data (GH10778)

  • interpolate() now supports method='akima' (GH7588).

  • pd.read_excel() now accepts path objects (e.g. pathlib.Path, py.path.local) for the file path, in line with other read_* functions (GH12655)

  • Added .weekday_name property as a component to DatetimeIndex and the .dt accessor. (GH11128)

  • Index.take now handles allow_fill and fill_value consistently (GH12631)

    In [33]: idx = pd.Index([1., 2., 3., 4.], dtype='float')
    
    # default, allow_fill=True, fill_value=None
    In [34]: idx.take([2, -1])
    Out[34]: Float64Index([3.0, 4.0], dtype='float64')
    
    In [35]: idx.take([2, -1], fill_value=True)
    Out[35]: Float64Index([3.0, nan], dtype='float64')
    
  • Index now supports .str.get_dummies() which returns MultiIndex, see Creating Indicator Variables (GH10008, GH10103)

    In [36]: idx = pd.Index(['a|b', 'a|c', 'b|c'])
    
    In [37]: idx.str.get_dummies('|')
    Out[37]: 
    MultiIndex(levels=[[0, 1], [0, 1], [0, 1]],
               labels=[[1, 1, 0], [1, 0, 1], [0, 1, 1]],
               names=['a', 'b', 'c'])
    
  • pd.crosstab() has gained a normalize argument for normalizing frequency tables (GH12569). Examples in the updated docs here.

  • .resample(..).interpolate() is now supported (GH12925)

  • .isin() now accepts passed sets (GH12988)

Sparse changes

These changes conform sparse handling to return the correct types and work to make a smoother experience with indexing.

SparseArray.take now returns a scalar for scalar input, SparseArray for others. Furthermore, it handles a negative indexer with the same rule as Index (GH10560, GH12796)

In [38]: s = pd.SparseArray([np.nan, np.nan, 1, 2, 3, np.nan, 4, 5, np.nan, 6])

In [39]: s.take(0)
Out[39]: nan

In [40]: s.take([1, 2, 3])
Out[40]: 
[nan, 1.0, 2.0]
Fill: nan
IntIndex
Indices: array([1, 2], dtype=int32)
  • Bug in SparseSeries[] indexing with Ellipsis raises KeyError (GH9467)
  • Bug in SparseArray[] indexing with tuples are not handled properly (GH12966)
  • Bug in SparseSeries.loc[] with list-like input raises TypeError (GH10560)
  • Bug in SparseSeries.iloc[] with scalar input may raise IndexError (GH10560)
  • Bug in SparseSeries.loc[], .iloc[] with slice returns SparseArray, rather than SparseSeries (GH10560)
  • Bug in SparseDataFrame.loc[], .iloc[] may results in dense Series, rather than SparseSeries (GH12787)
  • Bug in SparseArray addition ignores fill_value of right hand side (GH12910)
  • Bug in SparseArray mod raises AttributeError (GH12910)
  • Bug in SparseArray pow calculates 1 ** np.nan as np.nan which must be 1 (GH12910)
  • Bug in SparseArray comparison output may incorrect result or raise ValueError (GH12971)
  • Bug in SparseSeries.__repr__ raises TypeError when it is longer than max_rows (GH10560)
  • Bug in SparseSeries.shape ignores fill_value (GH10452)
  • Bug in SparseSeries and SparseArray may have different dtype from its dense values (GH12908)
  • Bug in SparseSeries.reindex incorrectly handle fill_value (GH12797)
  • Bug in SparseArray.to_frame() results in DataFrame, rather than SparseDataFrame (GH9850)
  • Bug in SparseSeries.value_counts() does not count fill_value (GH6749)
  • Bug in SparseArray.to_dense() does not preserve dtype (GH10648)
  • Bug in SparseArray.to_dense() incorrectly handle fill_value (GH12797)
  • Bug in pd.concat() of SparseSeries results in dense (GH10536)
  • Bug in pd.concat() of SparseDataFrame incorrectly handle fill_value (GH9765)
  • Bug in pd.concat() of SparseDataFrame may raise AttributeError (GH12174)
  • Bug in SparseArray.shift() may raise NameError or TypeError (GH12908)

API changes

.groupby(..).nth() changes

The index in .groupby(..).nth() output is now more consistent when the as_index argument is passed (GH11039):

In [41]: df = DataFrame({'A' : ['a', 'b', 'a'],
   ....:                 'B' : [1, 2, 3]})
   ....: 

In [42]: df
Out[42]: 
   A  B
0  a  1
1  b  2
2  a  3

Previous Behavior:

In [3]: df.groupby('A', as_index=True)['B'].nth(0)
Out[3]:
0    1
1    2
Name: B, dtype: int64

In [4]: df.groupby('A', as_index=False)['B'].nth(0)
Out[4]:
0    1
1    2
Name: B, dtype: int64

New Behavior:

In [43]: df.groupby('A', as_index=True)['B'].nth(0)
Out[43]: 
A
a    1
b    2
Name: B, dtype: int64

In [44]: df.groupby('A', as_index=False)['B'].nth(0)
Out[44]: 
0    1
1    2
Name: B, dtype: int64

Furthermore, previously, a .groupby would always sort, regardless if sort=False was passed with .nth().

In [45]: np.random.seed(1234)

In [46]: df = pd.DataFrame(np.random.randn(100, 2), columns=['a', 'b'])

In [47]: df['c'] = np.random.randint(0, 4, 100)

Previous Behavior:

In [4]: df.groupby('c', sort=True).nth(1)
Out[4]:
          a         b
c
0 -0.334077  0.002118
1  0.036142 -2.074978
2 -0.720589  0.887163
3  0.859588 -0.636524

In [5]: df.groupby('c', sort=False).nth(1)
Out[5]:
          a         b
c
0 -0.334077  0.002118
1  0.036142 -2.074978
2 -0.720589  0.887163
3  0.859588 -0.636524

New Behavior:

In [48]: df.groupby('c', sort=True).nth(1)
Out[48]: 
          a         b
c                    
0 -0.334077  0.002118
1  0.036142 -2.074978
2 -0.720589  0.887163
3  0.859588 -0.636524

In [49]: df.groupby('c', sort=False).nth(1)
Out[49]: 
          a         b
c                    
2 -0.720589  0.887163
3  0.859588 -0.636524
0 -0.334077  0.002118
1  0.036142 -2.074978

numpy function compatibility

Compatibility between pandas array-like methods (e.g. sum and take) and their numpy counterparts has been greatly increased by augmenting the signatures of the pandas methods so as to accept arguments that can be passed in from numpy, even if they are not necessarily used in the pandas implementation (GH12644, GH12638, GH12687)

  • .searchsorted() for Index and TimedeltaIndex now accept a sorter argument to maintain compatibility with numpy’s searchsorted function (GH12238)
  • Bug in numpy compatibility of np.round() on a Series (GH12600)

An example of this signature augmentation is illustrated below:

In [50]: sp = pd.SparseDataFrame([1, 2, 3])

In [51]: sp
Out[51]: 
   0
0  1
1  2
2  3

Previous behaviour:

In [2]: np.cumsum(sp, axis=0)
...
TypeError: cumsum() takes at most 2 arguments (4 given)

New behaviour:

In [52]: np.cumsum(sp, axis=0)
Out[52]: 
   0
0  1
1  3
2  6

Using .apply on groupby resampling

Using apply on resampling groupby operations (using a pd.TimeGrouper) now has the same output types as similar apply calls on other groupby operations. (GH11742).

In [53]: df = pd.DataFrame({'date': pd.to_datetime(['10/10/2000', '11/10/2000']),
   ....:                   'value': [10, 13]})
   ....: 

In [54]: df
Out[54]: 
        date  value
0 2000-10-10     10
1 2000-11-10     13

Previous behavior:

In [1]: df.groupby(pd.TimeGrouper(key='date', freq='M')).apply(lambda x: x.value.sum())
Out[1]:
...
TypeError: cannot concatenate a non-NDFrame object

# Output is a Series
In [2]: df.groupby(pd.TimeGrouper(key='date', freq='M')).apply(lambda x: x[['value']].sum())
Out[2]:
date
2000-10-31  value    10
2000-11-30  value    13
dtype: int64

New Behavior:

# Output is a Series
In [55]: df.groupby(pd.TimeGrouper(key='date', freq='M')).apply(lambda x: x.value.sum())
Out[55]:
date
2000-10-31    10
2000-11-30    13
Freq: M, dtype: int64

# Output is a DataFrame
In [56]: df.groupby(pd.TimeGrouper(key='date', freq='M')).apply(lambda x: x[['value']].sum())
Out[56]:
            value
date
2000-10-31     10
2000-11-30     13

Changes in read_csv exceptions

In order to standardize the read_csv API for both the c and python engines, both will now raise an EmptyDataError, a subclass of ValueError, in response to empty columns or header (GH12493, GH12506)

Previous behaviour:

In [1]: df = pd.read_csv(StringIO(''), engine='c')
...
ValueError: No columns to parse from file

In [2]: df = pd.read_csv(StringIO(''), engine='python')
...
StopIteration

New behaviour:

In [1]: df = pd.read_csv(StringIO(''), engine='c')
...
pandas.io.common.EmptyDataError: No columns to parse from file

In [2]: df = pd.read_csv(StringIO(''), engine='python')
...
pandas.io.common.EmptyDataError: No columns to parse from file

In addition to this error change, several others have been made as well:

  • CParserError now sub-classes ValueError instead of just a Exception (GH12551)
  • A CParserError is now raised instead of a generic Exception in read_csv when the c engine cannot parse a column (GH12506)
  • A ValueError is now raised instead of a generic Exception in read_csv when the c engine encounters a NaN value in an integer column (GH12506)
  • A ValueError is now raised instead of a generic Exception in read_csv when true_values is specified, and the c engine encounters an element in a column containing unencodable bytes (GH12506)
  • pandas.parser.OverflowError exception has been removed and has been replaced with Python’s built-in OverflowError exception (GH12506)
  • pd.read_csv() no longer allows a combination of strings and integers for the usecols parameter (GH12678)

to_datetime error changes

Bugs in pd.to_datetime() when passing a unit with convertible entries and errors='coerce' or non-convertible with errors='ignore'. Furthermore, an OutOfBoundsDateime exception will be raised when an out-of-range value is encountered for that unit when errors='raise'. (GH11758, GH13052, GH13059)

Previous behaviour:

In [27]: pd.to_datetime(1420043460, unit='s', errors='coerce')
Out[27]: NaT

In [28]: pd.to_datetime(11111111, unit='D', errors='ignore')
OverflowError: Python int too large to convert to C long

In [29]: pd.to_datetime(11111111, unit='D', errors='raise')
OverflowError: Python int too large to convert to C long

New behaviour:

In [2]: pd.to_datetime(1420043460, unit='s', errors='coerce')
Out[2]: Timestamp('2014-12-31 16:31:00')

In [3]: pd.to_datetime(11111111, unit='D', errors='ignore')
Out[3]: 11111111

In [4]: pd.to_datetime(11111111, unit='D', errors='raise')
OutOfBoundsDatetime: cannot convert input with unit 'D'

Other API changes

  • .swaplevel() for Series, DataFrame, Panel, and MultiIndex now features defaults for its first two parameters i and j that swap the two innermost levels of the index. (GH12934)
  • .searchsorted() for Index and TimedeltaIndex now accept a sorter argument to maintain compatibility with numpy’s searchsorted function (GH12238)
  • Period and PeriodIndex now raises IncompatibleFrequency error which inherits ValueError rather than raw ValueError (GH12615)
  • Series.apply for category dtype now applies the passed function to each of the .categories (and not the .codes), and returns a category dtype if possible (GH12473)
  • read_csv will now raise a TypeError if parse_dates is neither a boolean, list, or dictionary (matches the doc-string) (GH5636)
  • The default for .query()/.eval() is now engine=None, which will use numexpr if it’s installed; otherwise it will fallback to the python engine. This mimics the pre-0.18.1 behavior if numexpr is installed (and which, previously, if numexpr was not installed, .query()/.eval() would raise). (GH12749)
  • pd.show_versions() now includes pandas_datareader version (GH12740)
  • Provide a proper __name__ and __qualname__ attributes for generic functions (GH12021)
  • pd.concat(ignore_index=True) now uses RangeIndex as default (GH12695)
  • pd.merge() and DataFrame.join() will show a UserWarning when merging/joining a single- with a multi-leveled dataframe (GH9455, GH12219)
  • Compat with scipy > 0.17 for deprecated piecewise_polynomial interpolation method; support for the replacement from_derivatives method (GH12887)

Deprecations

  • The method name Index.sym_diff() is deprecated and can be replaced by Index.symmetric_difference() (GH12591)
  • The method name Categorical.sort() is deprecated in favor of Categorical.sort_values() (GH12882)

Performance Improvements

  • Improved speed of SAS reader (GH12656, GH12961)
  • Performance improvements in .groupby(..).cumcount() (GH11039)
  • Improved memory usage in pd.read_csv() when using skiprows=an_integer (GH13005)
  • Improved performance of DataFrame.to_sql when checking case sensitivity for tables. Now only checks if table has been created correctly when table name is not lower case. (GH12876)
  • Improved performance of Period construction and time series plotting (GH12903, GH11831).
  • Improved performance of .str.encode() and .str.decode() methods (GH13008)
  • Improved performance of to_numeric if input is numeric dtype (GH12777)
  • Improved performance of sparse arithmetic with IntIndex (GH13036)

Bug Fixes

  • usecols parameter in pd.read_csv is now respected even when the lines of a CSV file are not even (GH12203)
  • Bug in groupby.transform(..) when axis=1 is specified with a non-monotonic ordered index (GH12713)
  • Bug in Period and PeriodIndex creation raises KeyError if freq="Minute" is specified. Note that “Minute” freq is deprecated in v0.17.0, and recommended to use freq="T" instead (GH11854)
  • Bug in .resample(...).count() with a PeriodIndex always raising a TypeError (GH12774)
  • Bug in .resample(...) with a PeriodIndex casting to a DatetimeIndex when empty (GH12868)
  • Bug in .resample(...) with a PeriodIndex when resampling to an existing frequency (GH12770)
  • Bug in printing data which contains Period with different freq raises ValueError (GH12615)
  • Bug in Series construction with Categorical and dtype='category' is specified (GH12574)
  • Bugs in concatenation with a coercable dtype was too aggressive, resulting in different dtypes in outputformatting when an object was longer than display.max_rows (GH12411, GH12045, GH11594, GH10571, GH12211)
  • Bug in float_format option with option not being validated as a callable. (GH12706)
  • Bug in GroupBy.filter when dropna=False and no groups fulfilled the criteria (GH12768)
  • Bug in __name__ of .cum* functions (GH12021)
  • Bug in .astype() of a Float64Inde/Int64Index to an Int64Index (GH12881)
  • Bug in roundtripping an integer based index in .to_json()/.read_json() when orient='index' (the default) (GH12866)
  • Bug in plotting Categorical dtypes cause error when attempting stacked bar plot (GH13019)
  • Compat with >= numpy 1.11 for NaT comparions (GH12969)
  • Bug in .drop() with a non-unique MultiIndex. (GH12701)
  • Bug in .concat of datetime tz-aware and naive DataFrames (GH12467)
  • Bug in correctly raising a ValueError in .resample(..).fillna(..) when passing a non-string (GH12952)
  • Bug fixes in various encoding and header processing issues in pd.read_sas() (GH12659, GH12654, GH12647, GH12809)
  • Bug in pd.crosstab() where would silently ignore aggfunc if values=None (GH12569).
  • Potential segfault in DataFrame.to_json when serialising datetime.time (GH11473).
  • Potential segfault in DataFrame.to_json when attempting to serialise 0d array (GH11299).
  • Segfault in to_json when attempting to serialise a DataFrame or Series with non-ndarray values; now supports serialization of category, sparse, and datetime64[ns, tz] dtypes (GH10778).
  • Bug in DataFrame.to_json with unsupported dtype not passed to default handler (GH12554).
  • Bug in .align not returning the sub-class (GH12983)
  • Bug in aligning a Series with a DataFrame (GH13037)
  • Bug in ABCPanel in which Panel4D was not being considered as a valid instance of this generic type (GH12810)
  • Bug in consistency of .name on .groupby(..).apply(..) cases (GH12363)
  • Bug in Timestamp.__repr__ that caused pprint to fail in nested structures (GH12622)
  • Bug in Timedelta.min and Timedelta.max, the properties now report the true minimum/maximum timedeltas as recognized by pandas. See the documentation. (GH12727)
  • Bug in .quantile() with interpolation may coerce to float unexpectedly (GH12772)
  • Bug in .quantile() with empty Series may return scalar rather than empty Series (GH12772)
  • Bug in .loc with out-of-bounds in a large indexer would raise IndexError rather than KeyError (GH12527)
  • Bug in resampling when using a TimedeltaIndex and .asfreq(), would previously not include the final fencepost (GH12926)
  • Bug in equality testing with a Categorical in a DataFrame (GH12564)
  • Bug in GroupBy.first(), .last() returns incorrect row when TimeGrouper is used (GH7453)
  • Bug in pd.read_csv() with the c engine when specifying skiprows with newlines in quoted items (GH10911, GH12775)
  • Bug in DataFrame timezone lost when assigning tz-aware datetime Series with alignment (GH12981)
  • Bug in .value_counts() when normalize=True and dropna=True where nulls still contributed to the normalized count (GH12558)
  • Bug in Series.value_counts() loses name if its dtype is category (GH12835)
  • Bug in Series.value_counts() loses timezone info (GH12835)
  • Bug in Series.value_counts(normalize=True) with Categorical raises UnboundLocalError (GH12835)
  • Bug in Panel.fillna() ignoring inplace=True (GH12633)
  • Bug in pd.read_csv() when specifying names, usecols, and parse_dates simultaneously with the c engine (GH9755)
  • Bug in pd.read_csv() when specifying delim_whitespace=True and lineterminator simultaneously with the c engine (GH12912)
  • Bug in Series.rename, DataFrame.rename and DataFrame.rename_axis not treating Series as mappings to relabel (GH12623).
  • Clean in .rolling.min and .rolling.max to enhance dtype handling (GH12373)
  • Bug in groupby where complex types are coerced to float (GH12902)
  • Bug in Series.map raises TypeError if its dtype is category or tz-aware datetime (GH12473)
  • Bugs on 32bit platforms for some test comparisons (GH12972)
  • Bug in index coercion when falling back from RangeIndex construction (GH12893)
  • Better error message in window functions when invalid argument (e.g. a float window) is passed (GH12669)
  • Bug in slicing subclassed DataFrame defined to return subclassed Series may return normal Series (GH11559)
  • Bug in .str accessor methods may raise ValueError if input has name and the result is DataFrame or MultiIndex (GH12617)
  • Bug in DataFrame.last_valid_index() and DataFrame.first_valid_index() on empty frames (GH12800)
  • Bug in CategoricalIndex.get_loc returns different result from regular Index (GH12531)
  • Bug in PeriodIndex.resample where name not propagated (GH12769)
  • Bug in date_range closed keyword and timezones (GH12684).
  • Bug in pd.concat raises AttributeError when input data contains tz-aware datetime and timedelta (GH12620)
  • Bug in pd.concat did not handle empty Series properly (GH11082)
  • Bug in .plot.bar alginment when width is specified with int (GH12979)
  • Bug in fill_value is ignored if the argument to a binary operator is a constant (GH12723)
  • Bug in pd.read_html() when using bs4 flavor and parsing table with a header and only one column (GH9178)
  • Bug in .pivot_table when margins=True and dropna=True where nulls still contributed to margin count (GH12577)
  • Bug in .pivot_table when dropna=False where table index/column names disappear (GH12133)
  • Bug in pd.crosstab() when margins=True and dropna=False which raised (GH12642)
  • Bug in Series.name when name attribute can be a hashable type (GH12610)
  • Bug in .describe() resets categorical columns information (GH11558)
  • Bug where loffset argument was not applied when calling resample().count() on a timeseries (GH12725)
  • pd.read_excel() now accepts column names associated with keyword argument names (GH12870)
  • Bug in pd.to_numeric() with Index returns np.ndarray, rather than Index (GH12777)
  • Bug in pd.to_numeric() with datetime-like may raise TypeError (GH12777)
  • Bug in pd.to_numeric() with scalar raises ValueError (GH12777)

v0.18.0 (March 13, 2016)

This is a major release from 0.17.1 and includes a small number of API changes, several new features, enhancements, and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this version.

Warning

pandas >= 0.18.0 no longer supports compatibility with Python version 2.6 and 3.3 (GH7718, GH11273)

Warning

numexpr version 2.4.4 will now show a warning and not be used as a computation back-end for pandas because of some buggy behavior. This does not affect other versions (>= 2.1 and >= 2.4.6). (GH12489)

Highlights include:

  • Moving and expanding window functions are now methods on Series and DataFrame, similar to .groupby, see here.
  • Adding support for a RangeIndex as a specialized form of the Int64Index for memory savings, see here.
  • API breaking change to the .resample method to make it more .groupby like, see here.
  • Removal of support for positional indexing with floats, which was deprecated since 0.14.0. This will now raise a TypeError, see here.
  • The .to_xarray() function has been added for compatibility with the xarray package, see here.
  • The read_sas function has been enhanced to read sas7bdat files, see here.
  • Addition of the .str.extractall() method, and API changes to the .str.extract() method and .str.cat() method.
  • pd.test() top-level nose test runner is available (GH4327).

Check the API Changes and deprecations before updating.

New features

Window functions are now methods

Window functions have been refactored to be methods on Series/DataFrame objects, rather than top-level functions, which are now deprecated. This allows these window-type functions, to have a similar API to that of .groupby. See the full documentation here (GH11603, GH12373)

In [1]: np.random.seed(1234)

In [2]: df = pd.DataFrame({'A' : range(10), 'B' : np.random.randn(10)})

In [3]: df
Out[3]: 
   A         B
0  0  0.471435
1  1 -1.190976
2  2  1.432707
3  3 -0.312652
4  4 -0.720589
5  5  0.887163
6  6  0.859588
7  7 -0.636524
8  8  0.015696
9  9 -2.242685

Previous Behavior:

In [8]: pd.rolling_mean(df,window=3)
        FutureWarning: pd.rolling_mean is deprecated for DataFrame and will be removed in a future version, replace with
                       DataFrame.rolling(window=3,center=False).mean()
Out[8]:
    A         B
0 NaN       NaN
1 NaN       NaN
2   1  0.237722
3   2 -0.023640
4   3  0.133155
5   4 -0.048693
6   5  0.342054
7   6  0.370076
8   7  0.079587
9   8 -0.954504

New Behavior:

In [4]: r = df.rolling(window=3)

These show a descriptive repr

In [5]: r
Out[5]: Rolling [window=3,center=False,axis=0]

with tab-completion of available methods and properties.

In [9]: r.
r.A           r.agg         r.apply       r.count       r.exclusions  r.max         r.median      r.name        r.skew        r.sum
r.B           r.aggregate   r.corr        r.cov         r.kurt        r.mean        r.min         r.quantile    r.std         r.var

The methods operate on the Rolling object itself

In [6]: r.mean()
Out[6]: 
     A         B
0  NaN       NaN
1  NaN       NaN
2  1.0  0.237722
3  2.0 -0.023640
4  3.0  0.133155
5  4.0 -0.048693
6  5.0  0.342054
7  6.0  0.370076
8  7.0  0.079587
9  8.0 -0.954504

They provide getitem accessors

In [7]: r['A'].mean()
Out[7]: 
0    NaN
1    NaN
2    1.0
3    2.0
4    3.0
5    4.0
6    5.0
7    6.0
8    7.0
9    8.0
Name: A, dtype: float64

And multiple aggregations

In [8]: r.agg({'A' : ['mean','std'],
   ...:        'B' : ['mean','std']})
   ...: 
Out[8]: 
     A              B          
  mean  std      mean       std
0  NaN  NaN       NaN       NaN
1  NaN  NaN       NaN       NaN
2  1.0  1.0  0.237722  1.327364
3  2.0  1.0 -0.023640  1.335505
4  3.0  1.0  0.133155  1.143778
5  4.0  1.0 -0.048693  0.835747
6  5.0  1.0  0.342054  0.920379
7  6.0  1.0  0.370076  0.871850
8  7.0  1.0  0.079587  0.750099
9  8.0  1.0 -0.954504  1.162285

Changes to rename

Series.rename and NDFrame.rename_axis can now take a scalar or list-like argument for altering the Series or axis name, in addition to their old behaviors of altering labels. (GH9494, GH11965)

In [9]: s = pd.Series(np.random.randn(5))

In [10]: s.rename('newname')
Out[10]: 
0    1.150036
1    0.991946
2    0.953324
3   -2.021255
4   -0.334077
Name: newname, dtype: float64
In [11]: df = pd.DataFrame(np.random.randn(5, 2))

In [12]: (df.rename_axis("indexname")
   ....:    .rename_axis("columns_name", axis="columns"))
   ....: 
Out[12]: 
columns_name         0         1
indexname                       
0             0.002118  0.405453
1             0.289092  1.321158
2            -1.546906 -0.202646
3            -0.655969  0.193421
4             0.553439  1.318152

The new functionality works well in method chains. Previously these methods only accepted functions or dicts mapping a label to a new label. This continues to work as before for function or dict-like values.

Range Index

A RangeIndex has been added to the Int64Index sub-classes to support a memory saving alternative for common use cases. This has a similar implementation to the python range object (xrange in python 2), in that it only stores the start, stop, and step values for the index. It will transparently interact with the user API, converting to Int64Index if needed.

This will now be the default constructed index for NDFrame objects, rather than previous an Int64Index. (GH939, GH12070, GH12071, GH12109, GH12888)

Previous Behavior:

In [3]: s = pd.Series(range(1000))

In [4]: s.index
Out[4]:
Int64Index([  0,   1,   2,   3,   4,   5,   6,   7,   8,   9,
            ...
            990, 991, 992, 993, 994, 995, 996, 997, 998, 999], dtype='int64', length=1000)

In [6]: s.index.nbytes
Out[6]: 8000

New Behavior:

In [13]: s = pd.Series(range(1000))

In [14]: s.index
Out[14]: RangeIndex(start=0, stop=1000, step=1)

In [15]: s.index.nbytes
Out[15]: 80

Changes to str.extract

The .str.extract method takes a regular expression with capture groups, finds the first match in each subject string, and returns the contents of the capture groups (GH11386).

In v0.18.0, the expand argument was added to extract.

  • expand=False: it returns a Series, Index, or DataFrame, depending on the subject and regular expression pattern (same behavior as pre-0.18.0).
  • expand=True: it always returns a DataFrame, which is more consistent and less confusing from the perspective of a user.

Currently the default is expand=None which gives a FutureWarning and uses expand=False. To avoid this warning, please explicitly specify expand.

In [1]: pd.Series(['a1', 'b2', 'c3']).str.extract('[ab](\d)', expand=None)
FutureWarning: currently extract(expand=None) means expand=False (return Index/Series/DataFrame)
but in a future version of pandas this will be changed to expand=True (return DataFrame)

Out[1]:
0      1
1      2
2    NaN
dtype: object

Extracting a regular expression with one group returns a Series if expand=False.

In [16]: pd.Series(['a1', 'b2', 'c3']).str.extract('[ab](\d)', expand=False)
Out[16]: 
0      1
1      2
2    NaN
dtype: object

It returns a DataFrame with one column if expand=True.

In [17]: pd.Series(['a1', 'b2', 'c3']).str.extract('[ab](\d)', expand=True)
Out[17]: 
     0
0    1
1    2
2  NaN

Calling on an Index with a regex with exactly one capture group returns an Index if expand=False.

In [18]: s = pd.Series(["a1", "b2", "c3"], ["A11", "B22", "C33"])

In [19]: s.index
Out[19]: Index(['A11', 'B22', 'C33'], dtype='object')

In [20]: s.index.str.extract("(?P<letter>[a-zA-Z])", expand=False)
Out[20]: Index(['A', 'B', 'C'], dtype='object', name='letter')

It returns a DataFrame with one column if expand=True.

In [21]: s.index.str.extract("(?P<letter>[a-zA-Z])", expand=True)
Out[21]: 
  letter
0      A
1      B
2      C

Calling on an Index with a regex with more than one capture group raises ValueError if expand=False.

>>> s.index.str.extract("(?P<letter>[a-zA-Z])([0-9]+)", expand=False)
ValueError: only one regex group is supported with Index

It returns a DataFrame if expand=True.

In [22]: s.index.str.extract("(?P<letter>[a-zA-Z])([0-9]+)", expand=True)
Out[22]: 
  letter   1
0      A  11
1      B  22
2      C  33

In summary, extract(expand=True) always returns a DataFrame with a row for every subject string, and a column for every capture group.

Addition of str.extractall

The .str.extractall method was added (GH11386). Unlike extract, which returns only the first match.

In [23]: s = pd.Series(["a1a2", "b1", "c1"], ["A", "B", "C"])

In [24]: s
Out[24]: 
A    a1a2
B      b1
C      c1
dtype: object

In [25]: s.str.extract("(?P<letter>[ab])(?P<digit>\d)", expand=False)
Out[25]: 
  letter digit
A      a     1
B      b     1
C    NaN   NaN

The extractall method returns all matches.

In [26]: s.str.extractall("(?P<letter>[ab])(?P<digit>\d)")
Out[26]: 
        letter digit
  match             
A 0          a     1
  1          a     2
B 0          b     1

Changes to str.cat

The method .str.cat() concatenates the members of a Series. Before, if NaN values were present in the Series, calling .str.cat() on it would return NaN, unlike the rest of the Series.str.* API. This behavior has been amended to ignore NaN values by default. (GH11435).

A new, friendlier ValueError is added to protect against the mistake of supplying the sep as an arg, rather than as a kwarg. (GH11334).

In [27]: pd.Series(['a','b',np.nan,'c']).str.cat(sep=' ')
Out[27]: 'a b c'

In [28]: pd.Series(['a','b',np.nan,'c']).str.cat(sep=' ', na_rep='?')
Out[28]: 'a b ? c'
In [2]: pd.Series(['a','b',np.nan,'c']).str.cat(' ')
ValueError: Did you mean to supply a `sep` keyword?

Datetimelike rounding

DatetimeIndex, Timestamp, TimedeltaIndex, Timedelta have gained the .round(), .floor() and .ceil() method for datetimelike rounding, flooring and ceiling. (GH4314, GH11963)

Naive datetimes

In [29]: dr = pd.date_range('20130101 09:12:56.1234', periods=3)

In [30]: dr
Out[30]: 
DatetimeIndex(['2013-01-01 09:12:56.123400', '2013-01-02 09:12:56.123400',
               '2013-01-03 09:12:56.123400'],
              dtype='datetime64[ns]', freq='D')

In [31]: dr.round('s')
Out[31]: 
DatetimeIndex(['2013-01-01 09:12:56', '2013-01-02 09:12:56',
               '2013-01-03 09:12:56'],
              dtype='datetime64[ns]', freq=None)

# Timestamp scalar
In [32]: dr[0]
Out[32]: Timestamp('2013-01-01 09:12:56.123400', freq='D')

In [33]: dr[0].round('10s')
Out[33]: Timestamp('2013-01-01 09:13:00')

Tz-aware are rounded, floored and ceiled in local times

In [34]: dr = dr.tz_localize('US/Eastern')

In [35]: dr
Out[35]: 
DatetimeIndex(['2013-01-01 09:12:56.123400-05:00',
               '2013-01-02 09:12:56.123400-05:00',
               '2013-01-03 09:12:56.123400-05:00'],
              dtype='datetime64[ns, US/Eastern]', freq='D')

In [36]: dr.round('s')
Out[36]: 
DatetimeIndex(['2013-01-01 09:12:56-05:00', '2013-01-02 09:12:56-05:00',
               '2013-01-03 09:12:56-05:00'],
              dtype='datetime64[ns, US/Eastern]', freq=None)

Timedeltas

In [37]: t = timedelta_range('1 days 2 hr 13 min 45 us',periods=3,freq='d')

In [38]: t
Out[38]: 
TimedeltaIndex(['1 days 02:13:00.000045', '2 days 02:13:00.000045',
                '3 days 02:13:00.000045'],
               dtype='timedelta64[ns]', freq='D')

In [39]: t.round('10min')
Out[39]: TimedeltaIndex(['1 days 02:10:00', '2 days 02:10:00', '3 days 02:10:00'], dtype='timedelta64[ns]', freq=None)

# Timedelta scalar
In [40]: t[0]
Out[40]: Timedelta('1 days 02:13:00.000045')

In [41]: t[0].round('2h')
Out[41]: Timedelta('1 days 02:00:00')

In addition, .round(), .floor() and .ceil() will be available thru the .dt accessor of Series.

In [42]: s = pd.Series(dr)

In [43]: s
Out[43]: 
0   2013-01-01 09:12:56.123400-05:00
1   2013-01-02 09:12:56.123400-05:00
2   2013-01-03 09:12:56.123400-05:00
dtype: datetime64[ns, US/Eastern]

In [44]: s.dt.round('D')
Out[44]: 
0   2013-01-01 00:00:00-05:00
1   2013-01-02 00:00:00-05:00
2   2013-01-03 00:00:00-05:00
dtype: datetime64[ns, US/Eastern]

Formatting of Integers in FloatIndex

Integers in FloatIndex, e.g. 1., are now formatted with a decimal point and a 0 digit, e.g. 1.0 (GH11713) This change not only affects the display to the console, but also the output of IO methods like .to_csv or .to_html.

Previous Behavior:

In [2]: s = pd.Series([1,2,3], index=np.arange(3.))

In [3]: s
Out[3]:
0    1
1    2
2    3
dtype: int64

In [4]: s.index
Out[4]: Float64Index([0.0, 1.0, 2.0], dtype='float64')

In [5]: print(s.to_csv(path=None))
0,1
1,2
2,3

New Behavior:

In [45]: s = pd.Series([1,2,3], index=np.arange(3.))

In [46]: s
Out[46]: 
0.0    1
1.0    2
2.0    3
dtype: int64

In [47]: s.index
Out[47]: Float64Index([0.0, 1.0, 2.0], dtype='float64')

In [48]: print(s.to_csv(path=None))
0.0,1
1.0,2
2.0,3

Changes to dtype assignment behaviors

When a DataFrame’s slice is updated with a new slice of the same dtype, the dtype of the DataFrame will now remain the same. (GH10503)

Previous Behavior:

In [5]: df = pd.DataFrame({'a': [0, 1, 1],
                           'b': pd.Series([100, 200, 300], dtype='uint32')})

In [7]: df.dtypes
Out[7]:
a     int64
b    uint32
dtype: object

In [8]: ix = df['a'] == 1

In [9]: df.loc[ix, 'b'] = df.loc[ix, 'b']

In [11]: df.dtypes
Out[11]:
a    int64
b    int64
dtype: object

New Behavior:

In [49]: df = pd.DataFrame({'a': [0, 1, 1],
   ....:                    'b': pd.Series([100, 200, 300], dtype='uint32')})
   ....: 

In [50]: df.dtypes
Out[50]: 
a     int64
b    uint32
dtype: object

In [51]: ix = df['a'] == 1

In [52]: df.loc[ix, 'b'] = df.loc[ix, 'b']

In [53]: df.dtypes
Out[53]: 
a     int64
b    uint32
dtype: object

When a DataFrame’s integer slice is partially updated with a new slice of floats that could potentially be downcasted to integer without losing precision, the dtype of the slice will be set to float instead of integer.

Previous Behavior:

In [4]: df = pd.DataFrame(np.array(range(1,10)).reshape(3,3),
                          columns=list('abc'),
                          index=[[4,4,8], [8,10,12]])

In [5]: df
Out[5]:
      a  b  c
4 8   1  2  3
  10  4  5  6
8 12  7  8  9

In [7]: df.ix[4, 'c'] = np.array([0., 1.])

In [8]: df
Out[8]:
      a  b  c
4 8   1  2  0
  10  4  5  1
8 12  7  8  9

New Behavior:

In [54]: df = pd.DataFrame(np.array(range(1,10)).reshape(3,3),
   ....:                   columns=list('abc'),
   ....:                   index=[[4,4,8], [8,10,12]])
   ....: 

In [55]: df
Out[55]: 
      a  b  c
4 8   1  2  3
  10  4  5  6
8 12  7  8  9

In [56]: df.loc[4, 'c'] = np.array([0., 1.])

In [57]: df
Out[57]: 
      a  b    c
4 8   1  2  0.0
  10  4  5  1.0
8 12  7  8  9.0

to_xarray

In a future version of pandas, we will be deprecating Panel and other > 2 ndim objects. In order to provide for continuity, all NDFrame objects have gained the .to_xarray() method in order to convert to xarray objects, which has a pandas-like interface for > 2 ndim. (GH11972)

See the xarray full-documentation here.

In [1]: p = Panel(np.arange(2*3*4).reshape(2,3,4))

In [2]: p.to_xarray()
Out[2]:
<xarray.DataArray (items: 2, major_axis: 3, minor_axis: 4)>
array([[[ 0,  1,  2,  3],
        [ 4,  5,  6,  7],
        [ 8,  9, 10, 11]],

       [[12, 13, 14, 15],
        [16, 17, 18, 19],
        [20, 21, 22, 23]]])
Coordinates:
  * items       (items) int64 0 1
  * major_axis  (major_axis) int64 0 1 2
  * minor_axis  (minor_axis) int64 0 1 2 3

Latex Representation

DataFrame has gained a ._repr_latex_() method in order to allow for conversion to latex in a ipython/jupyter notebook using nbconvert. (GH11778)

Note that this must be activated by setting the option pd.display.latex.repr=True (GH12182)

For example, if you have a jupyter notebook you plan to convert to latex using nbconvert, place the statement pd.display.latex.repr=True in the first cell to have the contained DataFrame output also stored as latex.

The options display.latex.escape and display.latex.longtable have also been added to the configuration and are used automatically by the to_latex method. See the available options docs for more info.

pd.read_sas() changes

read_sas has gained the ability to read SAS7BDAT files, including compressed files. The files can be read in entirety, or incrementally. For full details see here. (GH4052)

Other enhancements

  • Handle truncated floats in SAS xport files (GH11713)
  • Added option to hide index in Series.to_string (GH11729)
  • read_excel now supports s3 urls of the format s3://bucketname/filename (GH11447)
  • add support for AWS_S3_HOST env variable when reading from s3 (GH12198)
  • A simple version of Panel.round() is now implemented (GH11763)
  • For Python 3.x, round(DataFrame), round(Series), round(Panel) will work (GH11763)
  • sys.getsizeof(obj) returns the memory usage of a pandas object, including the values it contains (GH11597)
  • Series gained an is_unique attribute (GH11946)
  • DataFrame.quantile and Series.quantile now accept interpolation keyword (GH10174).
  • Added DataFrame.style.format for more flexible formatting of cell values (GH11692)
  • DataFrame.select_dtypes now allows the np.float16 typecode (GH11990)
  • pivot_table() now accepts most iterables for the values parameter (GH12017)
  • Added Google BigQuery service account authentication support, which enables authentication on remote servers. (GH11881, GH12572). For further details see here
  • HDFStore is now iterable: for k in store is equivalent to for k in store.keys() (GH12221).
  • Add missing methods/fields to .dt for Period (GH8848)
  • The entire codebase has been PEP-ified (GH12096)

Backwards incompatible API changes

  • the leading whitespaces have been removed from the output of .to_string(index=False) method (GH11833)
  • the out parameter has been removed from the Series.round() method. (GH11763)
  • DataFrame.round() leaves non-numeric columns unchanged in its return, rather than raises. (GH11885)
  • DataFrame.head(0) and DataFrame.tail(0) return empty frames, rather than self. (GH11937)
  • Series.head(0) and Series.tail(0) return empty series, rather than self. (GH11937)
  • to_msgpack and read_msgpack encoding now defaults to 'utf-8'. (GH12170)
  • the order of keyword arguments to text file parsing functions (.read_csv(), .read_table(), .read_fwf()) changed to group related arguments. (GH11555)
  • NaTType.isoformat now returns the string 'NaT to allow the result to be passed to the constructor of Timestamp. (GH12300)

NaT and Timedelta operations

NaT and Timedelta have expanded arithmetic operations, which are extended to Series arithmetic where applicable. Operations defined for datetime64[ns] or timedelta64[ns] are now also defined for NaT (GH11564).

NaT now supports arithmetic operations with integers and floats.

In [58]: pd.NaT * 1
Out[58]: NaT

In [59]: pd.NaT * 1.5
Out[59]: NaT

In [60]: pd.NaT / 2
Out[60]: NaT

In [61]: pd.NaT * np.nan
Out[61]: NaT

NaT defines more arithmetic operations with datetime64[ns] and timedelta64[ns].

In [62]: pd.NaT / pd.NaT
Out[62]: nan

In [63]: pd.Timedelta('1s') / pd.NaT
Out[63]: nan

NaT may represent either a datetime64[ns] null or a timedelta64[ns] null. Given the ambiguity, it is treated as a timedelta64[ns], which allows more operations to succeed.

In [64]: pd.NaT + pd.NaT
Out[64]: NaT

# same as
In [65]: pd.Timedelta('1s') + pd.Timedelta('1s')
Out[65]: Timedelta('0 days 00:00:02')

as opposed to

In [3]: pd.Timestamp('19900315') + pd.Timestamp('19900315')
TypeError: unsupported operand type(s) for +: 'Timestamp' and 'Timestamp'

However, when wrapped in a Series whose dtype is datetime64[ns] or timedelta64[ns], the dtype information is respected.

In [1]: pd.Series([pd.NaT], dtype='<M8[ns]') + pd.Series([pd.NaT], dtype='<M8[ns]')
TypeError: can only operate on a datetimes for subtraction,
           but the operator [__add__] was passed
In [66]: pd.Series([pd.NaT], dtype='<m8[ns]') + pd.Series([pd.NaT], dtype='<m8[ns]')
Out[66]: 
0   NaT
dtype: timedelta64[ns]

Timedelta division by floats now works.

In [67]: pd.Timedelta('1s') / 2.0
Out[67]: Timedelta('0 days 00:00:00.500000')

Subtraction by Timedelta in a Series by a Timestamp works (GH11925)

In [68]: ser = pd.Series(pd.timedelta_range('1 day', periods=3))

In [69]: ser
Out[69]: 
0   1 days
1   2 days
2   3 days
dtype: timedelta64[ns]

In [70]: pd.Timestamp('2012-01-01') - ser
Out[70]: 
0   2011-12-31
1   2011-12-30
2   2011-12-29
dtype: datetime64[ns]

NaT.isoformat() now returns 'NaT'. This change allows allows pd.Timestamp to rehydrate any timestamp like object from its isoformat (GH12300).

Changes to msgpack

Forward incompatible changes in msgpack writing format were made over 0.17.0 and 0.18.0; older versions of pandas cannot read files packed by newer versions (GH12129, GH10527)

Bugs in to_msgpack and read_msgpack introduced in 0.17.0 and fixed in 0.18.0, caused files packed in Python 2 unreadable by Python 3 (GH12142). The following table describes the backward and forward compat of msgpacks.

Warning

Packed with Can be unpacked with
pre-0.17 / Python 2 any
pre-0.17 / Python 3 any
0.17 / Python 2
  • ==0.17 / Python 2
  • >=0.18 / any Python
0.17 / Python 3 >=0.18 / any Python
0.18 >= 0.18

0.18.0 is backward-compatible for reading files packed by older versions, except for files packed with 0.17 in Python 2, in which case only they can only be unpacked in Python 2.

Signature change for .rank

Series.rank and DataFrame.rank now have the same signature (GH11759)

Previous signature

In [3]: pd.Series([0,1]).rank(method='average', na_option='keep',
                              ascending=True, pct=False)
Out[3]:
0    1
1    2
dtype: float64

In [4]: pd.DataFrame([0,1]).rank(axis=0, numeric_only=None,
                                 method='average', na_option='keep',
                                 ascending=True, pct=False)
Out[4]:
   0
0  1
1  2

New signature

In [71]: pd.Series([0,1]).rank(axis=0, method='average', numeric_only=None,
   ....:                       na_option='keep', ascending=True, pct=False)
   ....: 
Out[71]: 
0    1.0
1    2.0
dtype: float64

In [72]: pd.DataFrame([0,1]).rank(axis=0, method='average', numeric_only=None,
   ....:                          na_option='keep', ascending=True, pct=False)
   ....: 
Out[72]: 
     0
0  1.0
1  2.0

Bug in QuarterBegin with n=0

In previous versions, the behavior of the QuarterBegin offset was inconsistent depending on the date when the n parameter was 0. (GH11406)

The general semantics of anchored offsets for n=0 is to not move the date when it is an anchor point (e.g., a quarter start date), and otherwise roll forward to the next anchor point.

In [73]: d = pd.Timestamp('2014-02-01')

In [74]: d
Out[74]: Timestamp('2014-02-01 00:00:00')

In [75]: d + pd.offsets.QuarterBegin(n=0, startingMonth=2)
Out[75]: Timestamp('2014-02-01 00:00:00')

In [76]: d + pd.offsets.QuarterBegin(n=0, startingMonth=1)
Out[76]: Timestamp('2014-04-01 00:00:00')

For the QuarterBegin offset in previous versions, the date would be rolled backwards if date was in the same month as the quarter start date.

In [3]: d = pd.Timestamp('2014-02-15')

In [4]: d + pd.offsets.QuarterBegin(n=0, startingMonth=2)
Out[4]: Timestamp('2014-02-01 00:00:00')

This behavior has been corrected in version 0.18.0, which is consistent with other anchored offsets like MonthBegin and YearBegin.

In [77]: d = pd.Timestamp('2014-02-15')

In [78]: d + pd.offsets.QuarterBegin(n=0, startingMonth=2)
Out[78]: Timestamp('2014-05-01 00:00:00')

Resample API

Like the change in the window functions API above, .resample(...) is changing to have a more groupby-like API. (GH11732, GH12702, GH12202, GH12332, GH12334, GH12348, GH12448).

In [79]: np.random.seed(1234)

In [80]: df = pd.DataFrame(np.random.rand(10,4),
   ....:                   columns=list('ABCD'),
   ....:                   index=pd.date_range('2010-01-01 09:00:00', periods=10, freq='s'))
   ....: 

In [81]: df
Out[81]: 
                            A         B         C         D
2010-01-01 09:00:00  0.191519  0.622109  0.437728  0.785359
2010-01-01 09:00:01  0.779976  0.272593  0.276464  0.801872
2010-01-01 09:00:02  0.958139  0.875933  0.357817  0.500995
2010-01-01 09:00:03  0.683463  0.712702  0.370251  0.561196
2010-01-01 09:00:04  0.503083  0.013768  0.772827  0.882641
2010-01-01 09:00:05  0.364886  0.615396  0.075381  0.368824
2010-01-01 09:00:06  0.933140  0.651378  0.397203  0.788730
2010-01-01 09:00:07  0.316836  0.568099  0.869127  0.436173
2010-01-01 09:00:08  0.802148  0.143767  0.704261  0.704581
2010-01-01 09:00:09  0.218792  0.924868  0.442141  0.909316

Previous API:

You would write a resampling operation that immediately evaluates. If a how parameter was not provided, it would default to how='mean'.

In [6]: df.resample('2s')
Out[6]:
                         A         B         C         D
2010-01-01 09:00:00  0.485748  0.447351  0.357096  0.793615
2010-01-01 09:00:02  0.820801  0.794317  0.364034  0.531096
2010-01-01 09:00:04  0.433985  0.314582  0.424104  0.625733
2010-01-01 09:00:06  0.624988  0.609738  0.633165  0.612452
2010-01-01 09:00:08  0.510470  0.534317  0.573201  0.806949

You could also specify a how directly

In [7]: df.resample('2s', how='sum')
Out[7]:
                         A         B         C         D
2010-01-01 09:00:00  0.971495  0.894701  0.714192  1.587231
2010-01-01 09:00:02  1.641602  1.588635  0.728068  1.062191
2010-01-01 09:00:04  0.867969  0.629165  0.848208  1.251465
2010-01-01 09:00:06  1.249976  1.219477  1.266330  1.224904
2010-01-01 09:00:08  1.020940  1.068634  1.146402  1.613897

New API:

Now, you can write .resample(..) as a 2-stage operation like .groupby(...), which yields a Resampler.

In [82]: r = df.resample('2s')

In [83]: r
Out[83]: DatetimeIndexResampler [freq=<2 * Seconds>, axis=0, closed=left, label=left, convention=start, base=0]
Downsampling

You can then use this object to perform operations. These are downsampling operations (going from a higher frequency to a lower one).

In [84]: r.mean()
Out[84]: 
                            A         B         C         D
2010-01-01 09:00:00  0.485748  0.447351  0.357096  0.793615
2010-01-01 09:00:02  0.820801  0.794317  0.364034  0.531096
2010-01-01 09:00:04  0.433985  0.314582  0.424104  0.625733
2010-01-01 09:00:06  0.624988  0.609738  0.633165  0.612452
2010-01-01 09:00:08  0.510470  0.534317  0.573201  0.806949
In [85]: r.sum()
Out[85]: 
                            A         B         C         D
2010-01-01 09:00:00  0.971495  0.894701  0.714192  1.587231
2010-01-01 09:00:02  1.641602  1.588635  0.728068  1.062191
2010-01-01 09:00:04  0.867969  0.629165  0.848208  1.251465
2010-01-01 09:00:06  1.249976  1.219477  1.266330  1.224904
2010-01-01 09:00:08  1.020940  1.068634  1.146402  1.613897

Furthermore, resample now supports getitem operations to perform the resample on specific columns.

In [86]: r[['A','C']].mean()
Out[86]: 
                            A         C
2010-01-01 09:00:00  0.485748  0.357096
2010-01-01 09:00:02  0.820801  0.364034
2010-01-01 09:00:04  0.433985  0.424104
2010-01-01 09:00:06  0.624988  0.633165
2010-01-01 09:00:08  0.510470  0.573201

and .aggregate type operations.

In [87]: r.agg({'A' : 'mean', 'B' : 'sum'})
Out[87]: 
                            A         B
2010-01-01 09:00:00  0.485748  0.894701
2010-01-01 09:00:02  0.820801  1.588635
2010-01-01 09:00:04  0.433985  0.629165
2010-01-01 09:00:06  0.624988  1.219477
2010-01-01 09:00:08  0.510470  1.068634

These accessors can of course, be combined

In [88]: r[['A','B']].agg(['mean','sum'])
Out[88]: 
                            A                   B          
                         mean       sum      mean       sum
2010-01-01 09:00:00  0.485748  0.971495  0.447351  0.894701
2010-01-01 09:00:02  0.820801  1.641602  0.794317  1.588635
2010-01-01 09:00:04  0.433985  0.867969  0.314582  0.629165
2010-01-01 09:00:06  0.624988  1.249976  0.609738  1.219477
2010-01-01 09:00:08  0.510470  1.020940  0.534317  1.068634
Upsampling

Upsampling operations take you from a lower frequency to a higher frequency. These are now performed with the Resampler objects with backfill(), ffill(), fillna() and asfreq() methods.

In [89]: s = pd.Series(np.arange(5,dtype='int64'),
   ....:               index=date_range('2010-01-01', periods=5, freq='Q'))
   ....: 

In [90]: s
Out[90]: 
2010-03-31    0
2010-06-30    1
2010-09-30    2
2010-12-31    3
2011-03-31    4
Freq: Q-DEC, dtype: int64

Previously

In [6]: s.resample('M', fill_method='ffill')
Out[6]:
2010-03-31    0
2010-04-30    0
2010-05-31    0
2010-06-30    1
2010-07-31    1
2010-08-31    1
2010-09-30    2
2010-10-31    2
2010-11-30    2
2010-12-31    3
2011-01-31    3
2011-02-28    3
2011-03-31    4
Freq: M, dtype: int64

New API

In [91]: s.resample('M').ffill()
Out[91]: 
2010-03-31    0
2010-04-30    0
2010-05-31    0
2010-06-30    1
2010-07-31    1
2010-08-31    1
2010-09-30    2
2010-10-31    2
2010-11-30    2
2010-12-31    3
2011-01-31    3
2011-02-28    3
2011-03-31    4
Freq: M, dtype: int64

Note

In the new API, you can either downsample OR upsample. The prior implementation would allow you to pass an aggregator function (like mean) even though you were upsampling, providing a bit of confusion.

Previous API will work but with deprecations

Warning

This new API for resample includes some internal changes for the prior-to-0.18.0 API, to work with a deprecation warning in most cases, as the resample operation returns a deferred object. We can intercept operations and just do what the (pre 0.18.0) API did (with a warning). Here is a typical use case:

In [4]: r = df.resample('2s')

In [6]: r*10
pandas/tseries/resample.py:80: FutureWarning: .resample() is now a deferred operation
use .resample(...).mean() instead of .resample(...)

Out[6]:
                      A         B         C         D
2010-01-01 09:00:00  4.857476  4.473507  3.570960  7.936154
2010-01-01 09:00:02  8.208011  7.943173  3.640340  5.310957
2010-01-01 09:00:04  4.339846  3.145823  4.241039  6.257326
2010-01-01 09:00:06  6.249881  6.097384  6.331650  6.124518
2010-01-01 09:00:08  5.104699  5.343172  5.732009  8.069486

However, getting and assignment operations directly on a Resampler will raise a ValueError:

In [7]: r.iloc[0] = 5
ValueError: .resample() is now a deferred operation
use .resample(...).mean() instead of .resample(...)

There is a situation where the new API can not perform all the operations when using original code. This code is intending to resample every 2s, take the mean AND then take the min of those results.

In [4]: df.resample('2s').min()
Out[4]:
A    0.433985
B    0.314582
C    0.357096
D    0.531096
dtype: float64

The new API will:

In [92]: df.resample('2s').min()
Out[92]: 
                            A         B         C         D
2010-01-01 09:00:00  0.191519  0.272593  0.276464  0.785359
2010-01-01 09:00:02  0.683463  0.712702  0.357817  0.500995
2010-01-01 09:00:04  0.364886  0.013768  0.075381  0.368824
2010-01-01 09:00:06  0.316836  0.568099  0.397203  0.436173
2010-01-01 09:00:08  0.218792  0.143767  0.442141  0.704581

The good news is the return dimensions will differ between the new API and the old API, so this should loudly raise an exception.

To replicate the original operation

In [93]: df.resample('2s').mean().min()
Out[93]: 
A    0.433985
B    0.314582
C    0.357096
D    0.531096
dtype: float64

Changes to eval

In prior versions, new columns assignments in an eval expression resulted in an inplace change to the DataFrame. (GH9297, GH8664, GH10486)

In [94]: df = pd.DataFrame({'a': np.linspace(0, 10, 5), 'b': range(5)})

In [95]: df
Out[95]: 
      a  b
0   0.0  0
1   2.5  1
2   5.0  2
3   7.5  3
4  10.0  4
In [12]: df.eval('c = a + b')
FutureWarning: eval expressions containing an assignment currentlydefault to operating inplace.
This will change in a future version of pandas, use inplace=True to avoid this warning.

In [13]: df
Out[13]:
      a  b     c
0   0.0  0   0.0
1   2.5  1   3.5
2   5.0  2   7.0
3   7.5  3  10.5
4  10.0  4  14.0

In version 0.18.0, a new inplace keyword was added to choose whether the assignment should be done inplace or return a copy.

In [96]: df
Out[96]: 
      a  b     c
0   0.0  0   0.0
1   2.5  1   3.5
2   5.0  2   7.0
3   7.5  3  10.5
4  10.0  4  14.0

In [97]: df.eval('d = c - b', inplace=False)
Out[97]: 
      a  b     c     d
0   0.0  0   0.0   0.0
1   2.5  1   3.5   2.5
2   5.0  2   7.0   5.0
3   7.5  3  10.5   7.5
4  10.0  4  14.0  10.0

In [98]: df
Out[98]: 
      a  b     c
0   0.0  0   0.0
1   2.5  1   3.5
2   5.0  2   7.0
3   7.5  3  10.5
4  10.0  4  14.0

In [99]: df.eval('d = c - b', inplace=True)

In [100]: df
Out[100]: 
      a  b     c     d
0   0.0  0   0.0   0.0
1   2.5  1   3.5   2.5
2   5.0  2   7.0   5.0
3   7.5  3  10.5   7.5
4  10.0  4  14.0  10.0

Warning

For backwards compatibility, inplace defaults to True if not specified. This will change in a future version of pandas. If your code depends on an inplace assignment you should update to explicitly set inplace=True

The inplace keyword parameter was also added the query method.

In [101]: df.query('a > 5')
Out[101]: 
      a  b     c     d
3   7.5  3  10.5   7.5
4  10.0  4  14.0  10.0

In [102]: df.query('a > 5', inplace=True)

In [103]: df
Out[103]: 
      a  b     c     d
3   7.5  3  10.5   7.5
4  10.0  4  14.0  10.0

Warning

Note that the default value for inplace in a query is False, which is consistent with prior versions.

eval has also been updated to allow multi-line expressions for multiple assignments. These expressions will be evaluated one at a time in order. Only assignments are valid for multi-line expressions.

In [104]: df
Out[104]: 
      a  b     c     d
3   7.5  3  10.5   7.5
4  10.0  4  14.0  10.0

In [105]: df.eval("""
   .....: e = d + a
   .....: f = e - 22
   .....: g = f / 2.0""", inplace=True)
   .....: 

In [106]: df
Out[106]: 
      a  b     c     d     e    f    g
3   7.5  3  10.5   7.5  15.0 -7.0 -3.5
4  10.0  4  14.0  10.0  20.0 -2.0 -1.0

Other API Changes

  • DataFrame.between_time and Series.between_time now only parse a fixed set of time strings. Parsing of date strings is no longer supported and raises a ValueError. (GH11818)

    In [107]: s = pd.Series(range(10), pd.date_range('2015-01-01', freq='H', periods=10))
    
    In [108]: s.between_time("7:00am", "9:00am")
    Out[108]: 
    2015-01-01 07:00:00    7
    2015-01-01 08:00:00    8
    2015-01-01 09:00:00    9
    Freq: H, dtype: int64
    

    This will now raise.

    In [2]: s.between_time('20150101 07:00:00','20150101 09:00:00')
    ValueError: Cannot convert arg ['20150101 07:00:00'] to a time.
    
  • .memory_usage() now includes values in the index, as does memory_usage in .info() (GH11597)

  • DataFrame.to_latex() now supports non-ascii encodings (eg utf-8) in Python 2 with the parameter encoding (GH7061)

  • pandas.merge() and DataFrame.merge() will show a specific error message when trying to merge with an object that is not of type DataFrame or a subclass (GH12081)

  • DataFrame.unstack and Series.unstack now take fill_value keyword to allow direct replacement of missing values when an unstack results in missing values in the resulting DataFrame. As an added benefit, specifying fill_value will preserve the data type of the original stacked data. (GH9746)

  • As part of the new API for window functions and resampling, aggregation functions have been clarified, raising more informative error messages on invalid aggregations. (GH9052). A full set of examples are presented in groupby.

  • Statistical functions for NDFrame objects (like sum(), mean(), min()) will now raise if non-numpy-compatible arguments are passed in for **kwargs (GH12301)

  • .to_latex and .to_html gain a decimal parameter like .to_csv; the default is '.' (GH12031)

  • More helpful error message when constructing a DataFrame with empty data but with indices (GH8020)

  • .describe() will now properly handle bool dtype as a categorical (GH6625)

  • More helpful error message with an invalid .transform with user defined input (GH10165)

  • Exponentially weighted functions now allow specifying alpha directly (GH10789) and raise ValueError if parameters violate 0 < alpha <= 1 (GH12492)

Deprecations

  • The functions pd.rolling_*, pd.expanding_*, and pd.ewm* are deprecated and replaced by the corresponding method call. Note that the new suggested syntax includes all of the arguments (even if default) (GH11603)

    In [1]: s = pd.Series(range(3))
    
    In [2]: pd.rolling_mean(s,window=2,min_periods=1)
            FutureWarning: pd.rolling_mean is deprecated for Series and
                 will be removed in a future version, replace with
                 Series.rolling(min_periods=1,window=2,center=False).mean()
    Out[2]:
            0    0.0
            1    0.5
            2    1.5
            dtype: float64
    
    In [3]: pd.rolling_cov(s, s, window=2)
            FutureWarning: pd.rolling_cov is deprecated for Series and
                 will be removed in a future version, replace with
                 Series.rolling(window=2).cov(other=<Series>)
    Out[3]:
            0    NaN
            1    0.5
            2    0.5
            dtype: float64
    
  • The freq and how arguments to the .rolling, .expanding, and .ewm (new) functions are deprecated, and will be removed in a future version. You can simply resample the input prior to creating a window function. (GH11603).

    For example, instead of s.rolling(window=5,freq='D').max() to get the max value on a rolling 5 Day window, one could use s.resample('D').mean().rolling(window=5).max(), which first resamples the data to daily data, then provides a rolling 5 day window.

  • pd.tseries.frequencies.get_offset_name function is deprecated. Use offset’s .freqstr property as alternative (GH11192)

  • pandas.stats.fama_macbeth routines are deprecated and will be removed in a future version (GH6077)

  • pandas.stats.ols, pandas.stats.plm and pandas.stats.var routines are deprecated and will be removed in a future version (GH6077)

  • show a FutureWarning rather than a DeprecationWarning on using long-time deprecated syntax in HDFStore.select, where the where clause is not a string-like (GH12027)

  • The pandas.options.display.mpl_style configuration has been deprecated and will be removed in a future version of pandas. This functionality is better handled by matplotlib’s style sheets (GH11783).

Removal of deprecated float indexers

In GH4892 indexing with floating point numbers on a non-Float64Index was deprecated (in version 0.14.0). In 0.18.0, this deprecation warning is removed and these will now raise a TypeError. (GH12165, GH12333)

In [109]: s = pd.Series([1, 2, 3], index=[4, 5, 6])

In [110]: s
Out[110]: 
4    1
5    2
6    3
dtype: int64

In [111]: s2 = pd.Series([1, 2, 3], index=list('abc'))

In [112]: s2
Out[112]: 
a    1
b    2
c    3
dtype: int64

Previous Behavior:

# this is label indexing
In [2]: s[5.0]
FutureWarning: scalar indexers for index type Int64Index should be integers and not floating point
Out[2]: 2

# this is positional indexing
In [3]: s.iloc[1.0]
FutureWarning: scalar indexers for index type Int64Index should be integers and not floating point
Out[3]: 2

# this is label indexing
In [4]: s.loc[5.0]
FutureWarning: scalar indexers for index type Int64Index should be integers and not floating point
Out[4]: 2

# .ix would coerce 1.0 to the positional 1, and index
In [5]: s2.ix[1.0] = 10
FutureWarning: scalar indexers for index type Index should be integers and not floating point

In [6]: s2
Out[6]:
a     1
b    10
c     3
dtype: int64

New Behavior:

For iloc, getting & setting via a float scalar will always raise.

In [3]: s.iloc[2.0]
TypeError: cannot do label indexing on <class 'pandas.indexes.numeric.Int64Index'> with these indexers [2.0] of <type 'float'>

Other indexers will coerce to a like integer for both getting and setting. The FutureWarning has been dropped for .loc, .ix and [].

In [113]: s[5.0]
Out[113]: 2

In [114]: s.loc[5.0]
Out[114]: 2

and setting

In [115]: s_copy = s.copy()

In [116]: s_copy[5.0] = 10

In [117]: s_copy
Out[117]: 
4     1
5    10
6     3
dtype: int64

In [118]: s_copy = s.copy()

In [119]: s_copy.loc[5.0] = 10

In [120]: s_copy
Out[120]: 
4     1
5    10
6     3
dtype: int64

Positional setting with .ix and a float indexer will ADD this value to the index, rather than previously setting the value by position.

In [3]: s2.ix[1.0] = 10
In [4]: s2
Out[4]:
a       1
b       2
c       3
1.0    10
dtype: int64

Slicing will also coerce integer-like floats to integers for a non-Float64Index.

In [121]: s.loc[5.0:6]
Out[121]: 
5    2
6    3
dtype: int64

Note that for floats that are NOT coercible to ints, the label based bounds will be excluded

In [122]: s.loc[5.1:6]
Out[122]: 
6    3
dtype: int64

Float indexing on a Float64Index is unchanged.

In [123]: s = pd.Series([1, 2, 3], index=np.arange(3.))

In [124]: s[1.0]
Out[124]: 2

In [125]: s[1.0:2.5]
Out[125]: 
1.0    2
2.0    3
dtype: int64

Removal of prior version deprecations/changes

  • Removal of rolling_corr_pairwise in favor of .rolling().corr(pairwise=True) (GH4950)
  • Removal of expanding_corr_pairwise in favor of .expanding().corr(pairwise=True) (GH4950)
  • Removal of DataMatrix module. This was not imported into the pandas namespace in any event (GH12111)
  • Removal of cols keyword in favor of subset in DataFrame.duplicated() and DataFrame.drop_duplicates() (GH6680)
  • Removal of the read_frame and frame_query (both aliases for pd.read_sql) and write_frame (alias of to_sql) functions in the pd.io.sql namespace, deprecated since 0.14.0 (GH6292).
  • Removal of the order keyword from .factorize() (GH6930)

Performance Improvements

  • Improved performance of andrews_curves (GH11534)
  • Improved huge DatetimeIndex, PeriodIndex and TimedeltaIndex’s ops performance including NaT (GH10277)
  • Improved performance of pandas.concat (GH11958)
  • Improved performance of StataReader (GH11591)
  • Improved performance in construction of Categoricals with Series of datetimes containing NaT (GH12077)
  • Improved performance of ISO 8601 date parsing for dates without separators (GH11899), leading zeros (GH11871) and with whitespace preceding the time zone (GH9714)

Bug Fixes

  • Bug in GroupBy.size when data-frame is empty. (GH11699)
  • Bug in Period.end_time when a multiple of time period is requested (GH11738)
  • Regression in .clip with tz-aware datetimes (GH11838)
  • Bug in date_range when the boundaries fell on the frequency (GH11804, GH12409)
  • Bug in consistency of passing nested dicts to .groupby(...).agg(...) (GH9052)
  • Accept unicode in Timedelta constructor (GH11995)
  • Bug in value label reading for StataReader when reading incrementally (GH12014)
  • Bug in vectorized DateOffset when n parameter is 0 (GH11370)
  • Compat for numpy 1.11 w.r.t. NaT comparison changes (GH12049)
  • Bug in read_csv when reading from a StringIO in threads (GH11790)
  • Bug in not treating NaT as a missing value in datetimelikes when factorizing & with Categoricals (GH12077)
  • Bug in getitem when the values of a Series were tz-aware (GH12089)
  • Bug in Series.str.get_dummies when one of the variables was ‘name’ (GH12180)
  • Bug in pd.concat while concatenating tz-aware NaT series. (GH11693, GH11755, GH12217)
  • Bug in pd.read_stata with version <= 108 files (GH12232)
  • Bug in Series.resample using a frequency of Nano when the index is a DatetimeIndex and contains non-zero nanosecond parts (GH12037)
  • Bug in resampling with .nunique and a sparse index (GH12352)
  • Removed some compiler warnings (GH12471)
  • Work around compat issues with boto in python 3.5 (GH11915)
  • Bug in NaT subtraction from Timestamp or DatetimeIndex with timezones (GH11718)
  • Bug in subtraction of Series of a single tz-aware Timestamp (GH12290)
  • Use compat iterators in PY2 to support .next() (GH12299)
  • Bug in Timedelta.round with negative values (GH11690)
  • Bug in .loc against CategoricalIndex may result in normal Index (GH11586)
  • Bug in DataFrame.info when duplicated column names exist (GH11761)
  • Bug in .copy of datetime tz-aware objects (GH11794)
  • Bug in Series.apply and Series.map where timedelta64 was not boxed (GH11349)
  • Bug in DataFrame.set_index() with tz-aware Series (GH12358)
  • Bug in subclasses of DataFrame where AttributeError did not propagate (GH11808)
  • Bug groupby on tz-aware data where selection not returning Timestamp (GH11616)
  • Bug in pd.read_clipboard and pd.to_clipboard functions not supporting Unicode; upgrade included pyperclip to v1.5.15 (GH9263)
  • Bug in DataFrame.query containing an assignment (GH8664)
  • Bug in from_msgpack where __contains__() fails for columns of the unpacked DataFrame, if the DataFrame has object columns. (GH11880)
  • Bug in .resample on categorical data with TimedeltaIndex (GH12169)
  • Bug in timezone info lost when broadcasting scalar datetime to DataFrame (GH11682)
  • Bug in Index creation from Timestamp with mixed tz coerces to UTC (GH11488)
  • Bug in to_numeric where it does not raise if input is more than one dimension (GH11776)
  • Bug in parsing timezone offset strings with non-zero minutes (GH11708)
  • Bug in df.plot using incorrect colors for bar plots under matplotlib 1.5+ (GH11614)
  • Bug in the groupby plot method when using keyword arguments (GH11805).
  • Bug in DataFrame.duplicated and drop_duplicates causing spurious matches when setting keep=False (GH11864)
  • Bug in .loc result with duplicated key may have Index with incorrect dtype (GH11497)
  • Bug in pd.rolling_median where memory allocation failed even with sufficient memory (GH11696)
  • Bug in DataFrame.style with spurious zeros (GH12134)
  • Bug in DataFrame.style with integer columns not starting at 0 (GH12125)
  • Bug in .style.bar may not rendered properly using specific browser (GH11678)
  • Bug in rich comparison of Timedelta with a numpy.array of Timedelta that caused an infinite recursion (GH11835)
  • Bug in DataFrame.round dropping column index name (GH11986)
  • Bug in df.replace while replacing value in mixed dtype Dataframe (GH11698)
  • Bug in Index prevents copying name of passed Index, when a new name is not provided (GH11193)
  • Bug in read_excel failing to read any non-empty sheets when empty sheets exist and sheetname=None (GH11711)
  • Bug in read_excel failing to raise NotImplemented error when keywords parse_dates and date_parser are provided (GH11544)
  • Bug in read_sql with pymysql connections failing to return chunked data (GH11522)
  • Bug in .to_csv ignoring formatting parameters decimal, na_rep, float_format for float indexes (GH11553)
  • Bug in Int64Index and Float64Index preventing the use of the modulo operator (GH9244)
  • Bug in MultiIndex.drop for not lexsorted multi-indexes (GH12078)
  • Bug in DataFrame when masking an empty DataFrame (GH11859)
  • Bug in .plot potentially modifying the colors input when the number of columns didn’t match the number of series provided (GH12039).
  • Bug in Series.plot failing when index has a CustomBusinessDay frequency (GH7222).
  • Bug in .to_sql for datetime.time values with sqlite fallback (GH8341)
  • Bug in read_excel failing to read data with one column when squeeze=True (GH12157)
  • Bug in read_excel failing to read one empty column (GH12292, GH9002)
  • Bug in .groupby where a KeyError was not raised for a wrong column if there was only one row in the dataframe (GH11741)
  • Bug in .read_csv with dtype specified on empty data producing an error (GH12048)
  • Bug in .read_csv where strings like '2E' are treated as valid floats (GH12237)
  • Bug in building pandas with debugging symbols (GH12123)
  • Removed millisecond property of DatetimeIndex. This would always raise a ValueError (GH12019).
  • Bug in Series constructor with read-only data (GH11502)
  • Removed pandas.util.testing.choice(). Should use np.random.choice(), instead. (GH12386)
  • Bug in .loc setitem indexer preventing the use of a TZ-aware DatetimeIndex (GH12050)
  • Bug in .style indexes and multi-indexes not appearing (GH11655)
  • Bug in to_msgpack and from_msgpack which did not correctly serialize or deserialize NaT (GH12307).
  • Bug in .skew and .kurt due to roundoff error for highly similar values (GH11974)
  • Bug in Timestamp constructor where microsecond resolution was lost if HHMMSS were not separated with ‘:’ (GH10041)
  • Bug in buffer_rd_bytes src->buffer could be freed more than once if reading failed, causing a segfault (GH12098)
  • Bug in crosstab where arguments with non-overlapping indexes would return a KeyError (GH10291)
  • Bug in DataFrame.apply in which reduction was not being prevented for cases in which dtype was not a numpy dtype (GH12244)
  • Bug when initializing categorical series with a scalar value. (GH12336)
  • Bug when specifying a UTC DatetimeIndex by setting utc=True in .to_datetime (GH11934)
  • Bug when increasing the buffer size of CSV reader in read_csv (GH12494)
  • Bug when setting columns of a DataFrame with duplicate column names (GH12344)

v0.17.1 (November 21, 2015)

Note

We are proud to announce that pandas has become a sponsored project of the (NumFOCUS organization). This will help ensure the success of development of pandas as a world-class open-source project.

This is a minor bug-fix release from 0.17.0 and includes a large number of bug fixes along several new features, enhancements, and performance improvements. We recommend that all users upgrade to this version.

Highlights include:

  • Support for Conditional HTML Formatting, see here
  • Releasing the GIL on the csv reader & other ops, see here
  • Fixed regression in DataFrame.drop_duplicates from 0.16.2, causing incorrect results on integer values (GH11376)

New features

Conditional HTML Formatting

Warning

This is a new feature and is under active development. We’ll be adding features an possibly making breaking changes in future releases. Feedback is welcome.

We’ve added experimental support for conditional HTML formatting: the visual styling of a DataFrame based on the data. The styling is accomplished with HTML and CSS. Acesses the styler class with the pandas.DataFrame.style, attribute, an instance of Styler with your data attached.

Here’s a quick example:

In [1]: np.random.seed(123)

In [2]: df = DataFrame(np.random.randn(10, 5), columns=list('abcde'))

In [3]: html = df.style.background_gradient(cmap='viridis', low=.5)

We can render the HTML to get the following table.

a b c d e
0 -1.085631 0.997345 0.282978 -1.506295 -0.5786
1 1.651437 -2.426679 -0.428913 1.265936 -0.86674
2 -0.678886 -0.094709 1.49139 -0.638902 -0.443982
3 -0.434351 2.20593 2.186786 1.004054 0.386186
4 0.737369 1.490732 -0.935834 1.175829 -1.253881
5 -0.637752 0.907105 -1.428681 -0.140069 -0.861755
6 -0.255619 -2.798589 -1.771533 -0.699877 0.927462
7 -0.173636 0.002846 0.688223 -0.879536 0.283627
8 -0.805367 -1.727669 -0.3909 0.573806 0.338589
9 -0.01183 2.392365 0.412912 0.978736 2.238143

Styler interacts nicely with the Jupyter Notebook. See the documentation for more.

Enhancements

  • DatetimeIndex now supports conversion to strings with astype(str) (GH10442)

  • Support for compression (gzip/bz2) in pandas.DataFrame.to_csv() (GH7615)

  • pd.read_* functions can now also accept pathlib.Path, or py._path.local.LocalPath objects for the filepath_or_buffer argument. (GH11033) - The DataFrame and Series functions .to_csv(), .to_html() and .to_latex() can now handle paths beginning with tildes (e.g. ~/Documents/) (GH11438)

  • DataFrame now uses the fields of a namedtuple as columns, if columns are not supplied (GH11181)

  • DataFrame.itertuples() now returns namedtuple objects, when possible. (GH11269, GH11625)

  • Added axvlines_kwds to parallel coordinates plot (GH10709)

  • Option to .info() and .memory_usage() to provide for deep introspection of memory consumption. Note that this can be expensive to compute and therefore is an optional parameter. (GH11595)

    In [4]: df = DataFrame({'A' : ['foo']*1000})
    
    In [5]: df['B'] = df['A'].astype('category')
    
    # shows the '+' as we have object dtypes
    In [6]: df.info()
    <class 'pandas.core.frame.DataFrame'>
    RangeIndex: 1000 entries, 0 to 999
    Data columns (total 2 columns):
    A    1000 non-null object
    B    1000 non-null category
    dtypes: category(1), object(1)
    memory usage: 9.0+ KB
    
    # we have an accurate memory assessment (but can be expensive to compute this)
    In [7]: df.info(memory_usage='deep')
    <class 'pandas.core.frame.DataFrame'>
    RangeIndex: 1000 entries, 0 to 999
    Data columns (total 2 columns):
    A    1000 non-null object
    B    1000 non-null category
    dtypes: category(1), object(1)
    memory usage: 75.4 KB
    
  • Index now has a fillna method (GH10089)

    In [8]: pd.Index([1, np.nan, 3]).fillna(2)
    Out[8]: Float64Index([1.0, 2.0, 3.0], dtype='float64')
    
  • Series of type category now make .str.<...> and .dt.<...> accessor methods / properties available, if the categories are of that type. (GH10661)

    In [9]: s = pd.Series(list('aabb')).astype('category')
    
    In [10]: s
    Out[10]: 
    0    a
    1    a
    2    b
    3    b
    dtype: category
    Categories (2, object): [a, b]
    
    In [11]: s.str.contains("a")
    Out[11]: 
    0     True
    1     True
    2    False
    3    False
    dtype: bool
    
    In [12]: date = pd.Series(pd.date_range('1/1/2015', periods=5)).astype('category')
    
    In [13]: date
    Out[13]: 
    0   2015-01-01
    1   2015-01-02
    2   2015-01-03
    3   2015-01-04
    4   2015-01-05
    dtype: category
    Categories (5, datetime64[ns]): [2015-01-01, 2015-01-02, 2015-01-03, 2015-01-04, 2015-01-05]
    
    In [14]: date.dt.day
    Out[14]: 
    0    1
    1    2
    2    3
    3    4
    4    5
    dtype: int64
    
  • pivot_table now has a margins_name argument so you can use something other than the default of ‘All’ (GH3335)

  • Implement export of datetime64[ns, tz] dtypes with a fixed HDF5 store (GH11411)

  • Pretty printing sets (e.g. in DataFrame cells) now uses set literal syntax ({x, y}) instead of Legacy Python syntax (set([x, y])) (GH11215)

  • Improve the error message in pandas.io.gbq.to_gbq() when a streaming insert fails (GH11285) and when the DataFrame does not match the schema of the destination table (GH11359)

API changes

  • raise NotImplementedError in Index.shift for non-supported index types (GH8038)
  • min and max reductions on datetime64 and timedelta64 dtyped series now result in NaT and not nan (GH11245).
  • Indexing with a null key will raise a TypeError, instead of a ValueError (GH11356)
  • Series.ptp will now ignore missing values by default (GH11163)

Deprecations

  • The pandas.io.ga module which implements google-analytics support is deprecated and will be removed in a future version (GH11308)
  • Deprecate the engine keyword in .to_csv(), which will be removed in a future version (GH11274)

Performance Improvements

  • Checking monotonic-ness before sorting on an index (GH11080)
  • Series.dropna performance improvement when its dtype can’t contain NaN (GH11159)
  • Release the GIL on most datetime field operations (e.g. DatetimeIndex.year, Series.dt.year), normalization, and conversion to and from Period, DatetimeIndex.to_period and PeriodIndex.to_timestamp (GH11263)
  • Release the GIL on some rolling algos: rolling_median, rolling_mean, rolling_max, rolling_min, rolling_var, rolling_kurt, rolling_skew (GH11450)
  • Release the GIL when reading and parsing text files in read_csv, read_table (GH11272)
  • Improved performance of rolling_median (GH11450)
  • Improved performance of to_excel (GH11352)
  • Performance bug in repr of Categorical categories, which was rendering the strings before chopping them for display (GH11305)
  • Performance improvement in Categorical.remove_unused_categories, (GH11643).
  • Improved performance of Series constructor with no data and DatetimeIndex (GH11433)
  • Improved performance of shift, cumprod, and cumsum with groupby (GH4095)

Bug Fixes

  • SparseArray.__iter__() now does not cause PendingDeprecationWarning in Python 3.5 (GH11622)
  • Regression from 0.16.2 for output formatting of long floats/nan, restored in (GH11302)
  • Series.sort_index() now correctly handles the inplace option (GH11402)
  • Incorrectly distributed .c file in the build on PyPi when reading a csv of floats and passing na_values=<a scalar> would show an exception (GH11374)
  • Bug in .to_latex() output broken when the index has a name (GH10660)
  • Bug in HDFStore.append with strings whose encoded length exceeded the max unencoded length (GH11234)
  • Bug in merging datetime64[ns, tz] dtypes (GH11405)
  • Bug in HDFStore.select when comparing with a numpy scalar in a where clause (GH11283)
  • Bug in using DataFrame.ix with a multi-index indexer (GH11372)
  • Bug in date_range with ambiguous endpoints (GH11626)
  • Prevent adding new attributes to the accessors .str, .dt and .cat. Retrieving such a value was not possible, so error out on setting it. (GH10673)
  • Bug in tz-conversions with an ambiguous time and .dt accessors (GH11295)
  • Bug in output formatting when using an index of ambiguous times (GH11619)
  • Bug in comparisons of Series vs list-likes (GH11339)
  • Bug in DataFrame.replace with a datetime64[ns, tz] and a non-compat to_replace (GH11326, GH11153)
  • Bug in isnull where numpy.datetime64('NaT') in a numpy.array was not determined to be null(GH11206)
  • Bug in list-like indexing with a mixed-integer Index (GH11320)
  • Bug in pivot_table with margins=True when indexes are of Categorical dtype (GH10993)
  • Bug in DataFrame.plot cannot use hex strings colors (GH10299)
  • Regression in DataFrame.drop_duplicates from 0.16.2, causing incorrect results on integer values (GH11376)
  • Bug in pd.eval where unary ops in a list error (GH11235)
  • Bug in squeeze() with zero length arrays (GH11230, GH8999)
  • Bug in describe() dropping column names for hierarchical indexes (GH11517)
  • Bug in DataFrame.pct_change() not propagating axis keyword on .fillna method (GH11150)
  • Bug in .to_csv() when a mix of integer and string column names are passed as the columns parameter (GH11637)
  • Bug in indexing with a range, (GH11652)
  • Bug in inference of numpy scalars and preserving dtype when setting columns (GH11638)
  • Bug in to_sql using unicode column names giving UnicodeEncodeError with (GH11431).
  • Fix regression in setting of xticks in plot (GH11529).
  • Bug in holiday.dates where observance rules could not be applied to holiday and doc enhancement (GH11477, GH11533)
  • Fix plotting issues when having plain Axes instances instead of SubplotAxes (GH11520, GH11556).
  • Bug in DataFrame.to_latex() produces an extra rule when header=False (GH7124)
  • Bug in df.groupby(...).apply(func) when a func returns a Series containing a new datetimelike column (GH11324)
  • Bug in pandas.json when file to load is big (GH11344)
  • Bugs in to_excel with duplicate columns (GH11007, GH10982, GH10970)
  • Fixed a bug that prevented the construction of an empty series of dtype datetime64[ns, tz] (GH11245).
  • Bug in read_excel with multi-index containing integers (GH11317)
  • Bug in to_excel with openpyxl 2.2+ and merging (GH11408)
  • Bug in DataFrame.to_dict() produces a np.datetime64 object instead of Timestamp when only datetime is present in data (GH11327)
  • Bug in DataFrame.corr() raises exception when computes Kendall correlation for DataFrames with boolean and not boolean columns (GH11560)
  • Bug in the link-time error caused by C inline functions on FreeBSD 10+ (with clang) (GH10510)
  • Bug in DataFrame.to_csv in passing through arguments for formatting MultiIndexes, including date_format (GH7791)
  • Bug in DataFrame.join() with how='right' producing a TypeError (GH11519)
  • Bug in Series.quantile with empty list results has Index with object dtype (GH11588)
  • Bug in pd.merge results in empty Int64Index rather than Index(dtype=object) when the merge result is empty (GH11588)
  • Bug in Categorical.remove_unused_categories when having NaN values (GH11599)
  • Bug in DataFrame.to_sparse() loses column names for MultiIndexes (GH11600)
  • Bug in DataFrame.round() with non-unique column index producing a Fatal Python error (GH11611)
  • Bug in DataFrame.round() with decimals being a non-unique indexed Series producing extra columns (GH11618)

v0.17.0 (October 9, 2015)

This is a major release from 0.16.2 and includes a small number of API changes, several new features, enhancements, and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this version.

Warning

pandas >= 0.17.0 will no longer support compatibility with Python version 3.2 (GH9118)

Warning

The pandas.io.data package is deprecated and will be replaced by the pandas-datareader package. This will allow the data modules to be independently updated to your pandas installation. The API for pandas-datareader v0.1.1 is exactly the same as in pandas v0.17.0 (GH8961, GH10861).

After installing pandas-datareader, you can easily change your imports:

from pandas.io import data, wb

becomes

from pandas_datareader import data, wb

Highlights include:

  • Release the Global Interpreter Lock (GIL) on some cython operations, see here
  • Plotting methods are now available as attributes of the .plot accessor, see here
  • The sorting API has been revamped to remove some long-time inconsistencies, see here
  • Support for a datetime64[ns] with timezones as a first-class dtype, see here
  • The default for to_datetime will now be to raise when presented with unparseable formats, previously this would return the original input. Also, date parse functions now return consistent results. See here
  • The default for dropna in HDFStore has changed to False, to store by default all rows even if they are all NaN, see here
  • Datetime accessor (dt) now supports Series.dt.strftime to generate formatted strings for datetime-likes, and Series.dt.total_seconds to generate each duration of the timedelta in seconds. See here
  • Period and PeriodIndex can handle multiplied freq like 3D, which corresponding to 3 days span. See here
  • Development installed versions of pandas will now have PEP440 compliant version strings (GH9518)
  • Development support for benchmarking with the Air Speed Velocity library (GH8361)
  • Support for reading SAS xport files, see here
  • Documentation comparing SAS to pandas, see here
  • Removal of the automatic TimeSeries broadcasting, deprecated since 0.8.0, see here
  • Display format with plain text can optionally align with Unicode East Asian Width, see here
  • Compatibility with Python 3.5 (GH11097)
  • Compatibility with matplotlib 1.5.0 (GH11111)

Check the API Changes and deprecations before updating.

New features

Datetime with TZ

We are adding an implementation that natively supports datetime with timezones. A Series or a DataFrame column previously could be assigned a datetime with timezones, and would work as an object dtype. This had performance issues with a large number rows. See the docs for more details. (GH8260, GH10763, GH11034).

The new implementation allows for having a single-timezone across all rows, with operations in a performant manner.

In [1]: df = DataFrame({'A' : date_range('20130101',periods=3),
   ...:                 'B' : date_range('20130101',periods=3,tz='US/Eastern'),
   ...:                 'C' : date_range('20130101',periods=3,tz='CET')})
   ...: 

In [2]: df
Out[2]: 
           A                         B                         C
0 2013-01-01 2013-01-01 00:00:00-05:00 2013-01-01 00:00:00+01:00
1 2013-01-02 2013-01-02 00:00:00-05:00 2013-01-02 00:00:00+01:00
2 2013-01-03 2013-01-03 00:00:00-05:00 2013-01-03 00:00:00+01:00

In [3]: df.dtypes
Out[3]: 
A                datetime64[ns]
B    datetime64[ns, US/Eastern]
C           datetime64[ns, CET]
dtype: object
In [4]: df.B
Out[4]: 
0   2013-01-01 00:00:00-05:00
1   2013-01-02 00:00:00-05:00
2   2013-01-03 00:00:00-05:00
Name: B, dtype: datetime64[ns, US/Eastern]

In [5]: df.B.dt.tz_localize(None)
Out[5]: 
0   2013-01-01
1   2013-01-02
2   2013-01-03
Name: B, dtype: datetime64[ns]

This uses a new-dtype representation as well, that is very similar in look-and-feel to its numpy cousin datetime64[ns]

In [6]: df['B'].dtype
Out[6]: datetime64[ns, US/Eastern]

In [7]: type(df['B'].dtype)
Out[7]: pandas.core.dtypes.dtypes.DatetimeTZDtype

Note

There is a slightly different string repr for the underlying DatetimeIndex as a result of the dtype changes, but functionally these are the same.

Previous Behavior:

In [1]: pd.date_range('20130101',periods=3,tz='US/Eastern')
Out[1]: DatetimeIndex(['2013-01-01 00:00:00-05:00', '2013-01-02 00:00:00-05:00',
                       '2013-01-03 00:00:00-05:00'],
                      dtype='datetime64[ns]', freq='D', tz='US/Eastern')

In [2]: pd.date_range('20130101',periods=3,tz='US/Eastern').dtype
Out[2]: dtype('<M8[ns]')

New Behavior:

In [8]: pd.date_range('20130101',periods=3,tz='US/Eastern')
Out[8]: 
DatetimeIndex(['2013-01-01 00:00:00-05:00', '2013-01-02 00:00:00-05:00',
               '2013-01-03 00:00:00-05:00'],
              dtype='datetime64[ns, US/Eastern]', freq='D')

In [9]: pd.date_range('20130101',periods=3,tz='US/Eastern').dtype
Out[9]: datetime64[ns, US/Eastern]

Releasing the GIL

We are releasing the global-interpreter-lock (GIL) on some cython operations. This will allow other threads to run simultaneously during computation, potentially allowing performance improvements from multi-threading. Notably groupby, nsmallest, value_counts and some indexing operations benefit from this. (GH8882)

For example the groupby expression in the following code will have the GIL released during the factorization step, e.g. df.groupby('key') as well as the .sum() operation.

N = 1000000
ngroups = 10
df = DataFrame({'key' : np.random.randint(0,ngroups,size=N),
                'data' : np.random.randn(N) })
df.groupby('key')['data'].sum()

Releasing of the GIL could benefit an application that uses threads for user interactions (e.g. QT), or performing multi-threaded computations. A nice example of a library that can handle these types of computation-in-parallel is the dask library.

Plot submethods

The Series and DataFrame .plot() method allows for customizing plot types by supplying the kind keyword arguments. Unfortunately, many of these kinds of plots use different required and optional keyword arguments, which makes it difficult to discover what any given plot kind uses out of the dozens of possible arguments.

To alleviate this issue, we have added a new, optional plotting interface, which exposes each kind of plot as a method of the .plot attribute. Instead of writing series.plot(kind=<kind>, ...), you can now also use series.plot.<kind>(...):

In [10]: df = pd.DataFrame(np.random.rand(10, 2), columns=['a', 'b'])

In [11]: df.plot.bar()
_images/whatsnew_plot_submethods.png

As a result of this change, these methods are now all discoverable via tab-completion:

In [12]: df.plot.<TAB>
df.plot.area     df.plot.barh     df.plot.density  df.plot.hist     df.plot.line     df.plot.scatter
df.plot.bar      df.plot.box      df.plot.hexbin   df.plot.kde      df.plot.pie

Each method signature only includes relevant arguments. Currently, these are limited to required arguments, but in the future these will include optional arguments, as well. For an overview, see the new Plotting API documentation.

Additional methods for dt accessor

strftime

We are now supporting a Series.dt.strftime method for datetime-likes to generate a formatted string (GH10110). Examples:

# DatetimeIndex
In [13]: s = pd.Series(pd.date_range('20130101', periods=4))

In [14]: s
Out[14]: 
0   2013-01-01
1   2013-01-02
2   2013-01-03
3   2013-01-04
dtype: datetime64[ns]

In [15]: s.dt.strftime('%Y/%m/%d')
Out[15]: 
0    2013/01/01
1    2013/01/02
2    2013/01/03
3    2013/01/04
dtype: object
# PeriodIndex
In [16]: s = pd.Series(pd.period_range('20130101', periods=4))

In [17]: s
Out[17]: 
0   2013-01-01
1   2013-01-02
2   2013-01-03
3   2013-01-04
dtype: object

In [18]: s.dt.strftime('%Y/%m/%d')
Out[18]: 
0    2013/01/01
1    2013/01/02
2    2013/01/03
3    2013/01/04
dtype: object

The string format is as the python standard library and details can be found here

total_seconds

pd.Series of type timedelta64 has new method .dt.total_seconds() returning the duration of the timedelta in seconds (GH10817)

# TimedeltaIndex
In [19]: s = pd.Series(pd.timedelta_range('1 minutes', periods=4))

In [20]: s
Out[20]: 
0   0 days 00:01:00
1   1 days 00:01:00
2   2 days 00:01:00
3   3 days 00:01:00
dtype: timedelta64[ns]

In [21]: s.dt.total_seconds()
Out[21]: 
0        60.0
1     86460.0
2    172860.0
3    259260.0
dtype: float64

Period Frequency Enhancement

Period, PeriodIndex and period_range can now accept multiplied freq. Also, Period.freq and PeriodIndex.freq are now stored as a DateOffset instance like DatetimeIndex, and not as str (GH7811)

A multiplied freq represents a span of corresponding length. The example below creates a period of 3 days. Addition and subtraction will shift the period by its span.

In [22]: p = pd.Period('2015-08-01', freq='3D')

In [23]: p
Out[23]: Period('2015-08-01', '3D')

In [24]: p + 1
Out[24]: Period('2015-08-04', '3D')

In [25]: p - 2
Out[25]: Period('2015-07-26', '3D')

In [26]: p.to_timestamp()
Out[26]: Timestamp('2015-08-01 00:00:00')

In [27]: p.to_timestamp(how='E')
Out[27]: Timestamp('2015-08-03 00:00:00')

You can use the multiplied freq in PeriodIndex and period_range.

In [28]: idx = pd.period_range('2015-08-01', periods=4, freq='2D')

In [29]: idx
Out[29]: PeriodIndex(['2015-08-01', '2015-08-03', '2015-08-05', '2015-08-07'], dtype='period[2D]', freq='2D')

In [30]: idx + 1
Out[30]: PeriodIndex(['2015-08-03', '2015-08-05', '2015-08-07', '2015-08-09'], dtype='period[2D]', freq='2D')

Support for SAS XPORT files

read_sas() provides support for reading SAS XPORT format files. (GH4052).

df = pd.read_sas('sas_xport.xpt')

It is also possible to obtain an iterator and read an XPORT file incrementally.

for df in pd.read_sas('sas_xport.xpt', chunksize=10000)
    do_something(df)

See the docs for more details.

Support for Math Functions in .eval()

eval() now supports calling math functions (GH4893)

df = pd.DataFrame({'a': np.random.randn(10)})
df.eval("b = sin(a)")

The support math functions are sin, cos, exp, log, expm1, log1p, sqrt, sinh, cosh, tanh, arcsin, arccos, arctan, arccosh, arcsinh, arctanh, abs and arctan2.

These functions map to the intrinsics for the NumExpr engine. For the Python engine, they are mapped to NumPy calls.

Changes to Excel with MultiIndex

In version 0.16.2 a DataFrame with MultiIndex columns could not be written to Excel via to_excel. That functionality has been added (GH10564), along with updating read_excel so that the data can be read back with, no loss of information, by specifying which columns/rows make up the MultiIndex in the header and index_col parameters (GH4679)

See the documentation for more details.

In [31]: df = pd.DataFrame([[1,2,3,4], [5,6,7,8]],
   ....:                   columns = pd.MultiIndex.from_product([['foo','bar'],['a','b']],
   ....:                                                        names = ['col1', 'col2']),
   ....:                   index = pd.MultiIndex.from_product([['j'], ['l', 'k']],
   ....:                                                      names = ['i1', 'i2']))
   ....: 

In [32]: df
Out[32]: 
col1  foo    bar   
col2    a  b   a  b
i1 i2              
j  l    1  2   3  4
   k    5  6   7  8

In [33]: df.to_excel('test.xlsx')

In [34]: df = pd.read_excel('test.xlsx', header=[0,1], index_col=[0,1])

In [35]: df
Out[35]: 
col1  foo    bar   
col2    a  b   a  b
i1 i2              
j  l    1  2   3  4
   k    5  6   7  8

Previously, it was necessary to specify the has_index_names argument in read_excel, if the serialized data had index names. For version 0.17.0 the ouptput format of to_excel has been changed to make this keyword unnecessary - the change is shown below.

Old

_images/old-excel-index.png

New

_images/new-excel-index.png

Warning

Excel files saved in version 0.16.2 or prior that had index names will still able to be read in, but the has_index_names argument must specified to True.

Google BigQuery Enhancements

  • Added ability to automatically create a table/dataset using the pandas.io.gbq.to_gbq() function if the destination table/dataset does not exist. (GH8325, GH11121).
  • Added ability to replace an existing table and schema when calling the pandas.io.gbq.to_gbq() function via the if_exists argument. See the docs for more details (GH8325).
  • InvalidColumnOrder and InvalidPageToken in the gbq module will raise ValueError instead of IOError.
  • The generate_bq_schema() function is now deprecated and will be removed in a future version (GH11121)
  • The gbq module will now support Python 3 (GH11094).

Display Alignment with Unicode East Asian Width

Warning

Enabling this option will affect the performance for printing of DataFrame and Series (about 2 times slower). Use only when it is actually required.

Some East Asian countries use Unicode characters its width is corresponding to 2 alphabets. If a DataFrame or Series contains these characters, the default output cannot be aligned properly. The following options are added to enable precise handling for these characters.

  • display.unicode.east_asian_width: Whether to use the Unicode East Asian Width to calculate the display text width. (GH2612)
  • display.unicode.ambiguous_as_wide: Whether to handle Unicode characters belong to Ambiguous as Wide. (GH11102)
In [36]: df = pd.DataFrame({u'国籍': ['UK', u'日本'], u'名前': ['Alice', u'しのぶ']})

In [37]: df;
_images/option_unicode01.png
In [38]: pd.set_option('display.unicode.east_asian_width', True)

In [39]: df;
_images/option_unicode02.png

For further details, see here

Other enhancements

  • Support for openpyxl >= 2.2. The API for style support is now stable (GH10125)

  • merge now accepts the argument indicator which adds a Categorical-type column (by default called _merge) to the output object that takes on the values (GH8790)

    Observation Origin _merge value
    Merge key only in 'left' frame left_only
    Merge key only in 'right' frame right_only
    Merge key in both frames both
    In [40]: df1 = pd.DataFrame({'col1':[0,1], 'col_left':['a','b']})
    
    In [41]: df2 = pd.DataFrame({'col1':[1,2,2],'col_right':[2,2,2]})
    
    In [42]: pd.merge(df1, df2, on='col1', how='outer', indicator=True)
    Out[42]: 
       col1 col_left  col_right      _merge
    0     0        a        NaN   left_only
    1     1        b        2.0        both
    2     2      NaN        2.0  right_only
    3     2      NaN        2.0  right_only
    

    For more, see the updated docs

  • pd.to_numeric is a new function to coerce strings to numbers (possibly with coercion) (GH11133)

  • pd.merge will now allow duplicate column names if they are not merged upon (GH10639).

  • pd.pivot will now allow passing index as None (GH3962).

  • pd.concat will now use existing Series names if provided (GH10698).

    In [43]: foo = pd.Series([1,2], name='foo')
    
    In [44]: bar = pd.Series([1,2])
    
    In [45]: baz = pd.Series([4,5])
    

    Previous Behavior:

    In [1] pd.concat([foo, bar, baz], 1)
    Out[1]:
          0  1  2
       0  1  1  4
       1  2  2  5
    

    New Behavior:

    In [46]: pd.concat([foo, bar, baz], 1)
    Out[46]: 
       foo  0  1
    0    1  1  4
    1    2  2  5
    
  • DataFrame has gained the nlargest and nsmallest methods (GH10393)

  • Add a limit_direction keyword argument that works with limit to enable interpolate to fill NaN values forward, backward, or both (GH9218, GH10420, GH11115)

    In [47]: ser = pd.Series([np.nan, np.nan, 5, np.nan, np.nan, np.nan, 13])
    
    In [48]: ser.interpolate(limit=1, limit_direction='both')
    Out[48]: 
    0     NaN
    1     5.0
    2     5.0
    3     7.0
    4     NaN
    5    11.0
    6    13.0
    dtype: float64
    
  • Added a DataFrame.round method to round the values to a variable number of decimal places (GH10568).

    In [49]: df = pd.DataFrame(np.random.random([3, 3]), columns=['A', 'B', 'C'],
       ....: index=['first', 'second', 'third'])
       ....: 
    
    In [50]: df
    Out[50]: 
                   A         B         C
    first   0.342764  0.304121  0.417022
    second  0.681301  0.875457  0.510422
    third   0.669314  0.585937  0.624904
    
    In [51]: df.round(2)
    Out[51]: 
               A     B     C
    first   0.34  0.30  0.42
    second  0.68  0.88  0.51
    third   0.67  0.59  0.62
    
    In [52]: df.round({'A': 0, 'C': 2})
    Out[52]: 
              A         B     C
    first   0.0  0.304121  0.42
    second  1.0  0.875457  0.51
    third   1.0  0.585937  0.62
    
  • drop_duplicates and duplicated now accept a keep keyword to target first, last, and all duplicates. The take_last keyword is deprecated, see here (GH6511, GH8505)

    In [53]: s = pd.Series(['A', 'B', 'C', 'A', 'B', 'D'])
    
    In [54]: s.drop_duplicates()
    Out[54]: 
    0    A
    1    B
    2    C
    5    D
    dtype: object
    
    In [55]: s.drop_duplicates(keep='last')
    Out[55]: 
    2    C
    3    A
    4    B
    5    D
    dtype: object
    
    In [56]: s.drop_duplicates(keep=False)
    Out[56]: 
    2    C
    5    D
    dtype: object
    
  • Reindex now has a tolerance argument that allows for finer control of Limits on filling while reindexing (GH10411):

    In [57]: df = pd.DataFrame({'x': range(5),
       ....:                    't': pd.date_range('2000-01-01', periods=5)})
       ....: 
    
    In [58]: df.reindex([0.1, 1.9, 3.5],
       ....:            method='nearest',
       ....:            tolerance=0.2)
       ....: 
    Out[58]: 
           x          t
    0.1  0.0 2000-01-01
    1.9  2.0 2000-01-03
    3.5  NaN        NaT
    

    When used on a DatetimeIndex, TimedeltaIndex or PeriodIndex, tolerance will coerced into a Timedelta if possible. This allows you to specify tolerance with a string:

    In [59]: df = df.set_index('t')
    
    In [60]: df.reindex(pd.to_datetime(['1999-12-31']),
       ....:            method='nearest',
       ....:            tolerance='1 day')
       ....: 
    Out[60]: 
                x
    1999-12-31  0
    

    tolerance is also exposed by the lower level Index.get_indexer and Index.get_loc methods.

  • Added functionality to use the base argument when resampling a TimeDeltaIndex (GH10530)

  • DatetimeIndex can be instantiated using strings contains NaT (GH7599)

  • to_datetime can now accept the yearfirst keyword (GH7599)

  • pandas.tseries.offsets larger than the Day offset can now be used with a Series for addition/subtraction (GH10699). See the docs for more details.

  • pd.Timedelta.total_seconds() now returns Timedelta duration to ns precision (previously microsecond precision) (GH10939)

  • PeriodIndex now supports arithmetic with np.ndarray (GH10638)

  • Support pickling of Period objects (GH10439)

  • .as_blocks will now take a copy optional argument to return a copy of the data, default is to copy (no change in behavior from prior versions), (GH9607)

  • regex argument to DataFrame.filter now handles numeric column names instead of raising ValueError (GH10384).

  • Enable reading gzip compressed files via URL, either by explicitly setting the compression parameter or by inferring from the presence of the HTTP Content-Encoding header in the response (GH8685)

  • Enable writing Excel files in memory using StringIO/BytesIO (GH7074)

  • Enable serialization of lists and dicts to strings in ExcelWriter (GH8188)

  • SQL io functions now accept a SQLAlchemy connectable. (GH7877)

  • pd.read_sql and to_sql can accept database URI as con parameter (GH10214)

  • read_sql_table will now allow reading from views (GH10750).

  • Enable writing complex values to HDFStores when using the table format (GH10447)

  • Enable pd.read_hdf to be used without specifying a key when the HDF file contains a single dataset (GH10443)

  • pd.read_stata will now read Stata 118 type files. (GH9882)

  • msgpack submodule has been updated to 0.4.6 with backward compatibility (GH10581)

  • DataFrame.to_dict now accepts orient='index' keyword argument (GH10844).

  • DataFrame.apply will return a Series of dicts if the passed function returns a dict and reduce=True (GH8735).

  • Allow passing kwargs to the interpolation methods (GH10378).

  • Improved error message when concatenating an empty iterable of Dataframe objects (GH9157)

  • pd.read_csv can now read bz2-compressed files incrementally, and the C parser can read bz2-compressed files from AWS S3 (GH11070, GH11072).

  • In pd.read_csv, recognize s3n:// and s3a:// URLs as designating S3 file storage (GH11070, GH11071).

  • Read CSV files from AWS S3 incrementally, instead of first downloading the entire file. (Full file download still required for compressed files in Python 2.) (GH11070, GH11073)

  • pd.read_csv is now able to infer compression type for files read from AWS S3 storage (GH11070, GH11074).

Backwards incompatible API changes

Changes to sorting API

The sorting API has had some longtime inconsistencies. (GH9816, GH8239).

Here is a summary of the API PRIOR to 0.17.0:

  • Series.sort is INPLACE while DataFrame.sort returns a new object.
  • Series.order returns a new object
  • It was possible to use Series/DataFrame.sort_index to sort by values by passing the by keyword.
  • Series/DataFrame.sortlevel worked only on a MultiIndex for sorting by index.

To address these issues, we have revamped the API:

  • We have introduced a new method, DataFrame.sort_values(), which is the merger of DataFrame.sort(), Series.sort(), and Series.order(), to handle sorting of values.
  • The existing methods Series.sort(), Series.order(), and DataFrame.sort() have been deprecated and will be removed in a future version.
  • The by argument of DataFrame.sort_index() has been deprecated and will be removed in a future version.
  • The existing method .sort_index() will gain the level keyword to enable level sorting.

We now have two distinct and non-overlapping methods of sorting. A * marks items that will show a FutureWarning.

To sort by the values:

Previous Replacement
* Series.order() Series.sort_values()
* Series.sort() Series.sort_values(inplace=True)
* DataFrame.sort(columns=...) DataFrame.sort_values(by=...)

To sort by the index:

Previous Replacement
Series.sort_index() Series.sort_index()
Series.sortlevel(level=...) Series.sort_index(level=...)
DataFrame.sort_index() DataFrame.sort_index()
DataFrame.sortlevel(level=...) DataFrame.sort_index(level=...)
* DataFrame.sort() DataFrame.sort_index()

We have also deprecated and changed similar methods in two Series-like classes, Index and Categorical.

Previous Replacement
* Index.order() Index.sort_values()
* Categorical.order() Categorical.sort_values()

Changes to to_datetime and to_timedelta

Error handling

The default for pd.to_datetime error handling has changed to errors='raise'. In prior versions it was errors='ignore'. Furthermore, the coerce argument has been deprecated in favor of errors='coerce'. This means that invalid parsing will raise rather that return the original input as in previous versions. (GH10636)

Previous Behavior:

In [2]: pd.to_datetime(['2009-07-31', 'asd'])
Out[2]: array(['2009-07-31', 'asd'], dtype=object)

New Behavior:

In [3]: pd.to_datetime(['2009-07-31', 'asd'])
ValueError: Unknown string format

Of course you can coerce this as well.

In [61]: to_datetime(['2009-07-31', 'asd'], errors='coerce')
Out[61]: DatetimeIndex(['2009-07-31', 'NaT'], dtype='datetime64[ns]', freq=None)

To keep the previous behavior, you can use errors='ignore':

In [62]: to_datetime(['2009-07-31', 'asd'], errors='ignore')
Out[62]: array(['2009-07-31', 'asd'], dtype=object)

Furthermore, pd.to_timedelta has gained a similar API, of errors='raise'|'ignore'|'coerce', and the coerce keyword has been deprecated in favor of errors='coerce'.

Consistent Parsing

The string parsing of to_datetime, Timestamp and DatetimeIndex has been made consistent. (GH7599)

Prior to v0.17.0, Timestamp and to_datetime may parse year-only datetime-string incorrectly using today’s date, otherwise DatetimeIndex uses the beginning of the year. Timestamp and to_datetime may raise ValueError in some types of datetime-string which DatetimeIndex can parse, such as a quarterly string.

Previous Behavior:

In [1]: Timestamp('2012Q2')
Traceback
   ...
ValueError: Unable to parse 2012Q2

# Results in today's date.
In [2]: Timestamp('2014')
Out [2]: 2014-08-12 00:00:00

v0.17.0 can parse them as below. It works on DatetimeIndex also.

New Behavior:

In [63]: Timestamp('2012Q2')
Out[63]: Timestamp('2012-04-01 00:00:00')

In [64]: Timestamp('2014')
Out[64]: Timestamp('2014-01-01 00:00:00')

In [65]: DatetimeIndex(['2012Q2', '2014'])
Out[65]: DatetimeIndex(['2012-04-01', '2014-01-01'], dtype='datetime64[ns]', freq=None)

Note

If you want to perform calculations based on today’s date, use Timestamp.now() and pandas.tseries.offsets.

In [66]: import pandas.tseries.offsets as offsets

In [67]: Timestamp.now()
Out[67]: Timestamp('2018-07-05 17:10:40.227744')

In [68]: Timestamp.now() + offsets.DateOffset(years=1)
Out[68]: Timestamp('2019-07-05 17:10:40.228803')

Changes to Index Comparisons

Operator equal on Index should behavior similarly to Series (GH9947, GH10637)

Starting in v0.17.0, comparing Index objects of different lengths will raise a ValueError. This is to be consistent with the behavior of Series.

Previous Behavior:

In [2]: pd.Index([1, 2, 3]) == pd.Index([1, 4, 5])
Out[2]: array([ True, False, False], dtype=bool)

In [3]: pd.Index([1, 2, 3]) == pd.Index([2])
Out[3]: array([False,  True, False], dtype=bool)

In [4]: pd.Index([1, 2, 3]) == pd.Index([1, 2])
Out[4]: False

New Behavior:

In [8]: pd.Index([1, 2, 3]) == pd.Index([1, 4, 5])
Out[8]: array([ True, False, False], dtype=bool)

In [9]: pd.Index([1, 2, 3]) == pd.Index([2])
ValueError: Lengths must match to compare

In [10]: pd.Index([1, 2, 3]) == pd.Index([1, 2])
ValueError: Lengths must match to compare

Note that this is different from the numpy behavior where a comparison can be broadcast:

In [69]: np.array([1, 2, 3]) == np.array([1])
Out[69]: array([ True, False, False], dtype=bool)

or it can return False if broadcasting can not be done:

In [70]: np.array([1, 2, 3]) == np.array([1, 2])
Out[70]: False

Changes to Boolean Comparisons vs. None

Boolean comparisons of a Series vs None will now be equivalent to comparing with np.nan, rather than raise TypeError. (GH1079).

In [71]: s = Series(range(3))

In [72]: s.iloc[1] = None

In [73]: s
Out[73]: 
0    0.0
1    NaN
2    2.0
dtype: float64

Previous Behavior:

In [5]: s==None
TypeError: Could not compare <type 'NoneType'> type with Series

New Behavior:

In [74]: s==None
Out[74]: 
0    False
1    False
2    False
dtype: bool

Usually you simply want to know which values are null.

In [75]: s.isnull()
Out[75]: 
0    False
1     True
2    False
dtype: bool

Warning

You generally will want to use isnull/notnull for these types of comparisons, as isnull/notnull tells you which elements are null. One has to be mindful that nan's don’t compare equal, but None's do. Note that Pandas/numpy uses the fact that np.nan != np.nan, and treats None like np.nan.

In [76]: None == None
Out[76]: True

In [77]: np.nan == np.nan
Out[77]: False

HDFStore dropna behavior

The default behavior for HDFStore write functions with format='table' is now to keep rows that are all missing. Previously, the behavior was to drop rows that were all missing save the index. The previous behavior can be replicated using the dropna=True option. (GH9382)

Previous Behavior:

In [78]: df_with_missing = pd.DataFrame({'col1':[0, np.nan, 2],
   ....:                                 'col2':[1, np.nan, np.nan]})
   ....: 

In [79]: df_with_missing
Out[79]: 
   col1  col2
0   0.0   1.0
1   NaN   NaN
2   2.0   NaN
In [27]:
df_with_missing.to_hdf('file.h5',
                       'df_with_missing',
                       format='table',
                       mode='w')

In [28]: pd.read_hdf('file.h5', 'df_with_missing')

Out [28]:
      col1  col2
  0     0     1
  2     2   NaN

New Behavior:

In [80]: df_with_missing.to_hdf('file.h5',
   ....:                        'df_with_missing',
   ....:                         format='table',
   ....:                         mode='w')
   ....: 

In [81]: pd.read_hdf('file.h5', 'df_with_missing')
Out[81]: 
   col1  col2
0   0.0   1.0
1   NaN   NaN
2   2.0   NaN

See the docs for more details.

Changes to display.precision option

The display.precision option has been clarified to refer to decimal places (GH10451).

Earlier versions of pandas would format floating point numbers to have one less decimal place than the value in display.precision.

In [1]: pd.set_option('display.precision', 2)

In [2]: pd.DataFrame({'x': [123.456789]})
Out[2]:
       x
0  123.5

If interpreting precision as “significant figures” this did work for scientific notation but that same interpretation did not work for values with standard formatting. It was also out of step with how numpy handles formatting.

Going forward the value of display.precision will directly control the number of places after the decimal, for regular formatting as well as scientific notation, similar to how numpy’s precision print option works.

In [82]: pd.set_option('display.precision', 2)

In [83]: pd.DataFrame({'x': [123.456789]})
Out[83]: 
        x
0  123.46

To preserve output behavior with prior versions the default value of display.precision has been reduced to 6 from 7.

Changes to Categorical.unique

Categorical.unique now returns new Categoricals with categories and codes that are unique, rather than returning np.array (GH10508)

  • unordered category: values and categories are sorted by appearance order.
  • ordered category: values are sorted by appearance order, categories keep existing order.
In [84]: cat = pd.Categorical(['C', 'A', 'B', 'C'],
   ....:                      categories=['A', 'B', 'C'],
   ....:                      ordered=True)
   ....: 

In [85]: cat
Out[85]: 
[C, A, B, C]
Categories (3, object): [A < B < C]

In [86]: cat.unique()
Out[86]: 
[C, A, B]
Categories (3, object): [A < B < C]

In [87]: cat = pd.Categorical(['C', 'A', 'B', 'C'],
   ....:                      categories=['A', 'B', 'C'])
   ....: 

In [88]: cat
Out[88]: 
[C, A, B, C]
Categories (3, object): [A, B, C]

In [89]: cat.unique()
Out[89]: 
[C, A, B]
Categories (3, object): [C, A, B]

Changes to bool passed as header in Parsers

In earlier versions of pandas, if a bool was passed the header argument of read_csv, read_excel, or read_html it was implicitly converted to an integer, resulting in header=0 for False and header=1 for True (GH6113)

A bool input to header will now raise a TypeError

In [29]: df = pd.read_csv('data.csv', header=False)
TypeError: Passing a bool to header is invalid. Use header=None for no header or
header=int or list-like of ints to specify the row(s) making up the column names

Other API Changes

  • Line and kde plot with subplots=True now uses default colors, not all black. Specify color='k' to draw all lines in black (GH9894)

  • Calling the .value_counts() method on a Series with a categorical dtype now returns a Series with a CategoricalIndex (GH10704)

  • The metadata properties of subclasses of pandas objects will now be serialized (GH10553).

  • groupby using Categorical follows the same rule as Categorical.unique described above (GH10508)

  • When constructing DataFrame with an array of complex64 dtype previously meant the corresponding column was automatically promoted to the complex128 dtype. Pandas will now preserve the itemsize of the input for complex data (GH10952)

  • some numeric reduction operators would return ValueError, rather than TypeError on object types that includes strings and numbers (GH11131)

  • Passing currently unsupported chunksize argument to read_excel or ExcelFile.parse will now raise NotImplementedError (GH8011)

  • Allow an ExcelFile object to be passed into read_excel (GH11198)

  • DatetimeIndex.union does not infer freq if self and the input have None as freq (GH11086)

  • NaT’s methods now either raise ValueError, or return np.nan or NaT (GH9513)

    Behavior Methods
    return np.nan weekday, isoweekday
    return NaT date, now, replace, to_datetime, today
    return np.datetime64('NaT') to_datetime64 (unchanged)
    raise ValueError All other public methods (names not beginning with underscores)

Deprecations

  • For Series the following indexing functions are deprecated (GH10177).

    Deprecated Function Replacement
    .irow(i) .iloc[i] or .iat[i]
    .iget(i) .iloc[i] or .iat[i]
    .iget_value(i) .iloc[i] or .iat[i]
  • For DataFrame the following indexing functions are deprecated (GH10177).

    Deprecated Function Replacement
    .irow(i) .iloc[i]
    .iget_value(i, j) .iloc[i, j] or .iat[i, j]
    .icol(j) .iloc[:, j]

Note

These indexing function have been deprecated in the documentation since 0.11.0.

  • Categorical.name was deprecated to make Categorical more numpy.ndarray like. Use Series(cat, name="whatever") instead (GH10482).
  • Setting missing values (NaN) in a Categorical’s categories will issue a warning (GH10748). You can still have missing values in the values.
  • drop_duplicates and duplicated’s take_last keyword was deprecated in favor of keep. (GH6511, GH8505)
  • Series.nsmallest and nlargest’s take_last keyword was deprecated in favor of keep. (GH10792)
  • DataFrame.combineAdd and DataFrame.combineMult are deprecated. They can easily be replaced by using the add and mul methods: DataFrame.add(other, fill_value=0) and DataFrame.mul(other, fill_value=1.) (GH10735).
  • TimeSeries deprecated in favor of Series (note that this has been an alias since 0.13.0), (GH10890)
  • SparsePanel deprecated and will be removed in a future version (GH11157).
  • Series.is_time_series deprecated in favor of Series.index.is_all_dates (GH11135)
  • Legacy offsets (like 'A@JAN') are deprecated (note that this has been alias since 0.8.0) (GH10878)
  • WidePanel deprecated in favor of Panel, LongPanel in favor of DataFrame (note these have been aliases since < 0.11.0), (GH10892)
  • DataFrame.convert_objects has been deprecated in favor of type-specific functions pd.to_datetime, pd.to_timestamp and pd.to_numeric (new in 0.17.0) (GH11133).

Removal of prior version deprecations/changes

  • Removal of na_last parameters from Series.order() and Series.sort(), in favor of na_position. (GH5231)

  • Remove of percentile_width from .describe(), in favor of percentiles. (GH7088)

  • Removal of colSpace parameter from DataFrame.to_string(), in favor of col_space, circa 0.8.0 version.

  • Removal of automatic time-series broadcasting (GH2304)

    In [90]: np.random.seed(1234)
    
    In [91]: df = DataFrame(np.random.randn(5,2),columns=list('AB'),index=date_range('20130101',periods=5))
    
    In [92]: df
    Out[92]: 
                       A         B
    2013-01-01  0.471435 -1.190976
    2013-01-02  1.432707 -0.312652
    2013-01-03 -0.720589  0.887163
    2013-01-04  0.859588 -0.636524
    2013-01-05  0.015696 -2.242685
    

    Previously

    In [3]: df + df.A
    FutureWarning: TimeSeries broadcasting along DataFrame index by default is deprecated.
    Please use DataFrame.<op> to explicitly broadcast arithmetic operations along the index
    
    Out[3]:
                        A         B
    2013-01-01  0.942870 -0.719541
    2013-01-02  2.865414  1.120055
    2013-01-03 -1.441177  0.166574
    2013-01-04  1.719177  0.223065
    2013-01-05  0.031393 -2.226989
    

    Current

    In [93]: df.add(df.A,axis='index')
    Out[93]: 
                       A         B
    2013-01-01  0.942870 -0.719541
    2013-01-02  2.865414  1.120055
    2013-01-03 -1.441177  0.166574
    2013-01-04  1.719177  0.223065
    2013-01-05  0.031393 -2.226989
    
  • Remove table keyword in HDFStore.put/append, in favor of using format= (GH4645)

  • Remove kind in read_excel/ExcelFile as its unused (GH4712)

  • Remove infer_type keyword from pd.read_html as its unused (GH4770, GH7032)

  • Remove offset and timeRule keywords from Series.tshift/shift, in favor of freq (GH4853, GH4864)

  • Remove pd.load/pd.save aliases in favor of pd.to_pickle/pd.read_pickle (GH3787)

Performance Improvements

  • Development support for benchmarking with the Air Speed Velocity library (GH8361)
  • Added vbench benchmarks for alternative ExcelWriter engines and reading Excel files (GH7171)
  • Performance improvements in Categorical.value_counts (GH10804)
  • Performance improvements in SeriesGroupBy.nunique and SeriesGroupBy.value_counts and SeriesGroupby.transform (GH10820, GH11077)
  • Performance improvements in DataFrame.drop_duplicates with integer dtypes (GH10917)
  • Performance improvements in DataFrame.duplicated with wide frames. (GH10161, GH11180)
  • 4x improvement in timedelta string parsing (GH6755, GH10426)
  • 8x improvement in timedelta64 and datetime64 ops (GH6755)
  • Significantly improved performance of indexing MultiIndex with slicers (GH10287)
  • 8x improvement in iloc using list-like input (GH10791)
  • Improved performance of Series.isin for datetimelike/integer Series (GH10287)
  • 20x improvement in concat of Categoricals when categories are identical (GH10587)
  • Improved performance of to_datetime when specified format string is ISO8601 (GH10178)
  • 2x improvement of Series.value_counts for float dtype (GH10821)
  • Enable infer_datetime_format in to_datetime when date components do not have 0 padding (GH11142)
  • Regression from 0.16.1 in constructing DataFrame from nested dictionary (GH11084)
  • Performance improvements in addition/subtraction operations for DateOffset with Series or DatetimeIndex (GH10744, GH11205)

Bug Fixes

  • Bug in incorrection computation of .mean() on timedelta64[ns] because of overflow (GH9442)
  • Bug in .isin on older numpies (GH11232)
  • Bug in DataFrame.to_html(index=False) renders unnecessary name row (GH10344)
  • Bug in DataFrame.to_latex() the column_format argument could not be passed (GH9402)
  • Bug in DatetimeIndex when localizing with NaT (GH10477)
  • Bug in Series.dt ops in preserving meta-data (GH10477)
  • Bug in preserving NaT when passed in an otherwise invalid to_datetime construction (GH10477)
  • Bug in DataFrame.apply when function returns categorical series. (GH9573)
  • Bug in to_datetime with invalid dates and formats supplied (GH10154)
  • Bug in Index.drop_duplicates dropping name(s) (GH10115)
  • Bug in Series.quantile dropping name (GH10881)
  • Bug in pd.Series when setting a value on an empty Series whose index has a frequency. (GH10193)
  • Bug in pd.Series.interpolate with invalid order keyword values. (GH10633)
  • Bug in DataFrame.plot raises ValueError when color name is specified by multiple characters (GH10387)
  • Bug in Index construction with a mixed list of tuples (GH10697)
  • Bug in DataFrame.reset_index when index contains NaT. (GH10388)
  • Bug in ExcelReader when worksheet is empty (GH6403)
  • Bug in BinGrouper.group_info where returned values are not compatible with base class (GH10914)
  • Bug in clearing the cache on DataFrame.pop and a subsequent inplace op (GH10912)
  • Bug in indexing with a mixed-integer Index causing an ImportError (GH10610)
  • Bug in Series.count when index has nulls (GH10946)
  • Bug in pickling of a non-regular freq DatetimeIndex (GH11002)
  • Bug causing DataFrame.where to not respect the axis parameter when the frame has a symmetric shape. (GH9736)
  • Bug in Table.select_column where name is not preserved (GH10392)
  • Bug in offsets.generate_range where start and end have finer precision than offset (GH9907)
  • Bug in pd.rolling_* where Series.name would be lost in the output (GH10565)
  • Bug in stack when index or columns are not unique. (GH10417)
  • Bug in setting a Panel when an axis has a multi-index (GH10360)
  • Bug in USFederalHolidayCalendar where USMemorialDay and USMartinLutherKingJr were incorrect (GH10278 and GH9760 )
  • Bug in .sample() where returned object, if set, gives unnecessary SettingWithCopyWarning (GH10738)
  • Bug in .sample() where weights passed as Series were not aligned along axis before being treated positionally, potentially causing problems if weight indices were not aligned with sampled object. (GH10738)
  • Regression fixed in (GH9311, GH6620, GH9345), where groupby with a datetime-like converting to float with certain aggregators (GH10979)
  • Bug in DataFrame.interpolate with axis=1 and inplace=True (GH10395)
  • Bug in io.sql.get_schema when specifying multiple columns as primary key (GH10385).
  • Bug in groupby(sort=False) with datetime-like Categorical raises ValueError (GH10505)
  • Bug in groupby(axis=1) with filter() throws IndexError (GH11041)
  • Bug in test_categorical on big-endian builds (GH10425)
  • Bug in Series.shift and DataFrame.shift not supporting categorical data (GH9416)
  • Bug in Series.map using categorical Series raises AttributeError (GH10324)
  • Bug in MultiIndex.get_level_values including Categorical raises AttributeError (GH10460)
  • Bug in pd.get_dummies with sparse=True not returning SparseDataFrame (GH10531)
  • Bug in Index subtypes (such as PeriodIndex) not returning their own type for .drop and .insert methods (GH10620)
  • Bug in algos.outer_join_indexer when right array is empty (GH10618)
  • Bug in filter (regression from 0.16.0) and transform when grouping on multiple keys, one of which is datetime-like (GH10114)
  • Bug in to_datetime and to_timedelta causing Index name to be lost (GH10875)
  • Bug in len(DataFrame.groupby) causing IndexError when there’s a column containing only NaNs (GH11016)
  • Bug that caused segfault when resampling an empty Series (GH10228)
  • Bug in DatetimeIndex and PeriodIndex.value_counts resets name from its result, but retains in result’s Index. (GH10150)
  • Bug in pd.eval using numexpr engine coerces 1 element numpy array to scalar (GH10546)
  • Bug in pd.concat with axis=0 when column is of dtype category (GH10177)
  • Bug in read_msgpack where input type is not always checked (GH10369, GH10630)
  • Bug in pd.read_csv with kwargs index_col=False, index_col=['a', 'b'] or dtype (GH10413, GH10467, GH10577)
  • Bug in Series.from_csv with header kwarg not setting the Series.name or the Series.index.name (GH10483)
  • Bug in groupby.var which caused variance to be inaccurate for small float values (GH10448)
  • Bug in Series.plot(kind='hist') Y Label not informative (GH10485)
  • Bug in read_csv when using a converter which generates a uint8 type (GH9266)
  • Bug causes memory leak in time-series line and area plot (GH9003)
  • Bug when setting a Panel sliced along the major or minor axes when the right-hand side is a DataFrame (GH11014)
  • Bug that returns None and does not raise NotImplementedError when operator functions (e.g. .add) of Panel are not implemented (GH7692)
  • Bug in line and kde plot cannot accept multiple colors when subplots=True (GH9894)
  • Bug in DataFrame.plot raises ValueError when color name is specified by multiple characters (GH10387)
  • Bug in left and right align of Series with MultiIndex may be inverted (GH10665)
  • Bug in left and right join of with MultiIndex may be inverted (GH10741)
  • Bug in read_stata when reading a file with a different order set in columns (GH10757)
  • Bug in Categorical may not representing properly when category contains tz or Period (GH10713)
  • Bug in Categorical.__iter__ may not returning correct datetime and Period (GH10713)
  • Bug in indexing with a PeriodIndex on an object with a PeriodIndex (GH4125)
  • Bug in read_csv with engine='c': EOF preceded by a comment, blank line, etc. was not handled correctly (GH10728, GH10548)
  • Reading “famafrench” data via DataReader results in HTTP 404 error because of the website url is changed (GH10591).
  • Bug in read_msgpack where DataFrame to decode has duplicate column names (GH9618)
  • Bug in io.common.get_filepath_or_buffer which caused reading of valid S3 files to fail if the bucket also contained keys for which the user does not have read permission (GH10604)
  • Bug in vectorised setting of timestamp columns with python datetime.date and numpy datetime64 (GH10408, GH10412)
  • Bug in Index.take may add unnecessary freq attribute (GH10791)
  • Bug in merge with empty DataFrame may raise IndexError (GH10824)
  • Bug in to_latex where unexpected keyword argument for some documented arguments (GH10888)
  • Bug in indexing of large DataFrame where IndexError is uncaught (GH10645 and GH10692)
  • Bug in read_csv when using the nrows or chunksize parameters if file contains only a header line (GH9535)
  • Bug in serialization of category types in HDF5 in presence of alternate encodings. (GH10366)
  • Bug in pd.DataFrame when constructing an empty DataFrame with a string dtype (GH9428)
  • Bug in pd.DataFrame.diff when DataFrame is not consolidated (GH10907)
  • Bug in pd.unique for arrays with the datetime64 or timedelta64 dtype that meant an array with object dtype was returned instead the original dtype (GH9431)
  • Bug in Timedelta raising error when slicing from 0s (GH10583)
  • Bug in DatetimeIndex.take and TimedeltaIndex.take may not raise IndexError against invalid index (GH10295)
  • Bug in Series([np.nan]).astype('M8[ms]'), which now returns Series([pd.NaT]) (GH10747)
  • Bug in PeriodIndex.order reset freq (GH10295)
  • Bug in date_range when freq divides end as nanos (GH10885)
  • Bug in iloc allowing memory outside bounds of a Series to be accessed with negative integers (GH10779)
  • Bug in read_msgpack where encoding is not respected (GH10581)
  • Bug preventing access to the first index when using iloc with a list containing the appropriate negative integer (GH10547, GH10779)
  • Bug in TimedeltaIndex formatter causing error while trying to save DataFrame with TimedeltaIndex using to_csv (GH10833)
  • Bug in DataFrame.where when handling Series slicing (GH10218, GH9558)
  • Bug where pd.read_gbq throws ValueError when Bigquery returns zero rows (GH10273)
  • Bug in to_json which was causing segmentation fault when serializing 0-rank ndarray (GH9576)
  • Bug in plotting functions may raise IndexError when plotted on GridSpec (GH10819)
  • Bug in plot result may show unnecessary minor ticklabels (GH10657)
  • Bug in groupby incorrect computation for aggregation on DataFrame with NaT (E.g first, last, min). (GH10590, GH11010)
  • Bug when constructing DataFrame where passing a dictionary with only scalar values and specifying columns did not raise an error (GH10856)
  • Bug in .var() causing roundoff errors for highly similar values (GH10242)
  • Bug in DataFrame.plot(subplots=True) with duplicated columns outputs incorrect result (GH10962)
  • Bug in Index arithmetic may result in incorrect class (GH10638)
  • Bug in date_range results in empty if freq is negative annually, quarterly and monthly (GH11018)
  • Bug in DatetimeIndex cannot infer negative freq (GH11018)
  • Remove use of some deprecated numpy comparison operations, mainly in tests. (GH10569)
  • Bug in Index dtype may not applied properly (GH11017)
  • Bug in io.gbq when testing for minimum google api client version (GH10652)
  • Bug in DataFrame construction from nested dict with timedelta keys (GH11129)
  • Bug in .fillna against may raise TypeError when data contains datetime dtype (GH7095, GH11153)
  • Bug in .groupby when number of keys to group by is same as length of index (GH11185)
  • Bug in convert_objects where converted values might not be returned if all null and coerce (GH9589)
  • Bug in convert_objects where copy keyword was not respected (GH9589)

v0.16.2 (June 12, 2015)

This is a minor bug-fix release from 0.16.1 and includes a a large number of bug fixes along some new features (pipe() method), enhancements, and performance improvements.

We recommend that all users upgrade to this version.

Highlights include:

  • A new pipe method, see here
  • Documentation on how to use numba with pandas, see here

New features

Pipe

We’ve introduced a new method DataFrame.pipe(). As suggested by the name, pipe should be used to pipe data through a chain of function calls. The goal is to avoid confusing nested function calls like

# df is a DataFrame
# f, g, and h are functions that take and return DataFrames
f(g(h(df), arg1=1), arg2=2, arg3=3)

The logic flows from inside out, and function names are separated from their keyword arguments. This can be rewritten as

(df.pipe(h)
   .pipe(g, arg1=1)
   .pipe(f, arg2=2, arg3=3)
)

Now both the code and the logic flow from top to bottom. Keyword arguments are next to their functions. Overall the code is much more readable.

In the example above, the functions f, g, and h each expected the DataFrame as the first positional argument. When the function you wish to apply takes its data anywhere other than the first argument, pass a tuple of (function, keyword) indicating where the DataFrame should flow. For example:

In [1]: import statsmodels.formula.api as sm

In [2]: bb = pd.read_csv('data/baseball.csv', index_col='id')

# sm.ols takes (formula, data)
In [3]: (bb.query('h > 0')
   ...:    .assign(ln_h = lambda df: np.log(df.h))
   ...:    .pipe((sm.ols, 'data'), 'hr ~ ln_h + year + g + C(lg)')
   ...:    .fit()
   ...:    .summary()
   ...: )
   ...: 
Out[3]: 
<class 'statsmodels.iolib.summary.Summary'>
"""
                            OLS Regression Results                            
==============================================================================
Dep. Variable:                     hr   R-squared:                       0.685
Model:                            OLS   Adj. R-squared:                  0.665
Method:                 Least Squares   F-statistic:                     34.28
Date:                Thu, 05 Jul 2018   Prob (F-statistic):           3.48e-15
Time:                        17:10:40   Log-Likelihood:                -205.92
No. Observations:                  68   AIC:                             421.8
Df Residuals:                      63   BIC:                             432.9
Df Model:                           4                                         
Covariance Type:            nonrobust                                         
===============================================================================
                  coef    std err          t      P>|t|      [0.025      0.975]
-------------------------------------------------------------------------------
Intercept   -8484.7720   4664.146     -1.819      0.074   -1.78e+04     835.780
C(lg)[T.NL]    -2.2736      1.325     -1.716      0.091      -4.922       0.375
ln_h           -1.3542      0.875     -1.547      0.127      -3.103       0.395
year            4.2277      2.324      1.819      0.074      -0.417       8.872
g               0.1841      0.029      6.258      0.000       0.125       0.243
==============================================================================
Omnibus:                       10.875   Durbin-Watson:                   1.999
Prob(Omnibus):                  0.004   Jarque-Bera (JB):               17.298
Skew:                           0.537   Prob(JB):                     0.000175
Kurtosis:                       5.225   Cond. No.                     1.49e+07
==============================================================================

Warnings:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
[2] The condition number is large, 1.49e+07. This might indicate that there are
strong multicollinearity or other numerical problems.
"""

The pipe method is inspired by unix pipes, which stream text through processes. More recently dplyr and magrittr have introduced the popular (%>%) pipe operator for R.

See the documentation for more. (GH10129)

Other Enhancements

  • Added rsplit to Index/Series StringMethods (GH10303)

  • Removed the hard-coded size limits on the DataFrame HTML representation in the IPython notebook, and leave this to IPython itself (only for IPython v3.0 or greater). This eliminates the duplicate scroll bars that appeared in the notebook with large frames (GH10231).

    Note that the notebook has a toggle output scrolling feature to limit the display of very large frames (by clicking left of the output). You can also configure the way DataFrames are displayed using the pandas options, see here here.

  • axis parameter of DataFrame.quantile now accepts also index and column. (GH9543)

API Changes

  • Holiday now raises NotImplementedError if both offset and observance are used in the constructor instead of returning an incorrect result (GH10217).

Performance Improvements

  • Improved Series.resample performance with dtype=datetime64[ns] (GH7754)
  • Increase performance of str.split when expand=True (GH10081)

Bug Fixes

  • Bug in Series.hist raises an error when a one row Series was given (GH10214)
  • Bug where HDFStore.select modifies the passed columns list (GH7212)
  • Bug in Categorical repr with display.width of None in Python 3 (GH10087)
  • Bug in to_json with certain orients and a CategoricalIndex would segfault (GH10317)
  • Bug where some of the nan funcs do not have consistent return dtypes (GH10251)
  • Bug in DataFrame.quantile on checking that a valid axis was passed (GH9543)
  • Bug in groupby.apply aggregation for Categorical not preserving categories (GH10138)
  • Bug in to_csv where date_format is ignored if the datetime is fractional (GH10209)
  • Bug in DataFrame.to_json with mixed data types (GH10289)
  • Bug in cache updating when consolidating (GH10264)
  • Bug in mean() where integer dtypes can overflow (GH10172)
  • Bug where Panel.from_dict does not set dtype when specified (GH10058)
  • Bug in Index.union raises AttributeError when passing array-likes. (GH10149)
  • Bug in Timestamp’s‘ microsecond, quarter, dayofyear, week and daysinmonth properties return np.int type, not built-in int. (GH10050)
  • Bug in NaT raises AttributeError when accessing to daysinmonth, dayofweek properties. (GH10096)
  • Bug in Index repr when using the max_seq_items=None setting (GH10182).
  • Bug in getting timezone data with dateutil on various platforms ( GH9059, GH8639, GH9663, GH10121)
  • Bug in displaying datetimes with mixed frequencies; display ‘ms’ datetimes to the proper precision. (GH10170)
  • Bug in setitem where type promotion is applied to the entire block (GH10280)
  • Bug in Series arithmetic methods may incorrectly hold names (GH10068)
  • Bug in GroupBy.get_group when grouping on multiple keys, one of which is categorical. (GH10132)
  • Bug in DatetimeIndex and TimedeltaIndex names are lost after timedelta arithmetics ( GH9926)
  • Bug in DataFrame construction from nested dict with datetime64 (GH10160)
  • Bug in Series construction from dict with datetime64 keys (GH9456)
  • Bug in Series.plot(label="LABEL") not correctly setting the label (GH10119)
  • Bug in plot not defaulting to matplotlib axes.grid setting (GH9792)
  • Bug causing strings containing an exponent, but no decimal to be parsed as int instead of float in engine='python' for the read_csv parser (GH9565)
  • Bug in Series.align resets name when fill_value is specified (GH10067)
  • Bug in read_csv causing index name not to be set on an empty DataFrame (GH10184)
  • Bug in SparseSeries.abs resets name (GH10241)
  • Bug in TimedeltaIndex slicing may reset freq (GH10292)
  • Bug in GroupBy.get_group raises ValueError when group key contains NaT (GH6992)
  • Bug in SparseSeries constructor ignores input data name (GH10258)
  • Bug in Categorical.remove_categories causing a ValueError when removing the NaN category if underlying dtype is floating-point (GH10156)
  • Bug where infer_freq infers timerule (WOM-5XXX) unsupported by to_offset (GH9425)
  • Bug in DataFrame.to_hdf() where table format would raise a seemingly unrelated error for invalid (non-string) column names. This is now explicitly forbidden. (GH9057)
  • Bug to handle masking empty DataFrame (GH10126).
  • Bug where MySQL interface could not handle numeric table/column names (GH10255)
  • Bug in read_csv with a date_parser that returned a datetime64 array of other time resolution than [ns] (GH10245)
  • Bug in Panel.apply when the result has ndim=0 (GH10332)
  • Bug in read_hdf where auto_close could not be passed (GH9327).
  • Bug in read_hdf where open stores could not be used (GH10330).
  • Bug in adding empty DataFrames, now results in a DataFrame that .equals an empty DataFrame (GH10181).
  • Bug in to_hdf and HDFStore which did not check that complib choices were valid (GH4582, GH8874).

v0.16.1 (May 11, 2015)

This is a minor bug-fix release from 0.16.0 and includes a a large number of bug fixes along several new features, enhancements, and performance improvements. We recommend that all users upgrade to this version.

Highlights include:

  • Support for a CategoricalIndex, a category based index, see here
  • New section on how-to-contribute to pandas, see here
  • Revised “Merge, join, and concatenate” documentation, including graphical examples to make it easier to understand each operations, see here
  • New method sample for drawing random samples from Series, DataFrames and Panels. See here
  • The default Index printing has changed to a more uniform format, see here
  • BusinessHour datetime-offset is now supported, see here
  • Further enhancement to the .str accessor to make string operations easier, see here

Warning

In pandas 0.17.0, the sub-package pandas.io.data will be removed in favor of a separately installable package (GH8961).

Enhancements

CategoricalIndex

We introduce a CategoricalIndex, a new type of index object that is useful for supporting indexing with duplicates. This is a container around a Categorical (introduced in v0.15.0) and allows efficient indexing and storage of an index with a large number of duplicated elements. Prior to 0.16.1, setting the index of a DataFrame/Series with a category dtype would convert this to regular object-based Index.

In [1]: df = DataFrame({'A' : np.arange(6),
   ...:                 'B' : Series(list('aabbca')).astype('category',
   ...:                                                     categories=list('cab'))
   ...:                })
   ...:

In [2]: df
Out[2]:
   A  B
0  0  a
1  1  a
2  2  b
3  3  b
4  4  c
5  5  a

In [3]: df.dtypes
Out[3]:
A       int64
B    category
dtype: object

In [4]: df.B.cat.categories
Out[4]: Index(['c', 'a', 'b'], dtype='object')

setting the index, will create create a CategoricalIndex

In [5]: df2 = df.set_index('B')

In [6]: df2.index
Out[6]: CategoricalIndex(['a', 'a', 'b', 'b', 'c', 'a'], categories=['c', 'a', 'b'], ordered=False, name='B', dtype='category')

indexing with __getitem__/.iloc/.loc/.ix works similarly to an Index with duplicates. The indexers MUST be in the category or the operation will raise.

In [7]: df2.loc['a']
Out[7]:
   A
B
a  0
a  1
a  5

and preserves the CategoricalIndex

In [8]: df2.loc['a'].index
Out[8]: CategoricalIndex(['a', 'a', 'a'], categories=['c', 'a', 'b'], ordered=False, name='B', dtype='category')

sorting will order by the order of the categories

In [9]: df2.sort_index()
Out[9]:
   A
B
c  4
a  0
a  1
a  5
b  2
b  3

groupby operations on the index will preserve the index nature as well

In [10]: df2.groupby(level=0).sum()
Out[10]:
   A
B
c  4
a  6
b  5

In [11]: df2.groupby(level=0).sum().index
Out[11]: CategoricalIndex(['c', 'a', 'b'], categories=['c', 'a', 'b'], ordered=False, name='B', dtype='category')

reindexing operations, will return a resulting index based on the type of the passed indexer, meaning that passing a list will return a plain-old-Index; indexing with a Categorical will return a CategoricalIndex, indexed according to the categories of the PASSED Categorical dtype. This allows one to arbitrarly index these even with values NOT in the categories, similarly to how you can reindex ANY pandas index.

In [12]: df2.reindex(['a','e'])
Out[12]:
     A
B
a  0.0
a  1.0
a  5.0
e  NaN

In [13]: df2.reindex(['a','e']).index
Out[13]: Index(['a', 'a', 'a', 'e'], dtype='object', name='B')

In [14]: df2.reindex(pd.Categorical(['a','e'],categories=list('abcde')))
Out[14]:
     A
B
a  0.0
a  1.0
a  5.0
e  NaN

In [15]: df2.reindex(pd.Categorical(['a','e'],categories=list('abcde'))).index
Out[15]: CategoricalIndex(['a', 'a', 'a', 'e'], categories=['a', 'b', 'c', 'd', 'e'], ordered=False, name='B', dtype='category')

See the documentation for more. (GH7629, GH10038, GH10039)

Sample

Series, DataFrames, and Panels now have a new method: sample(). The method accepts a specific number of rows or columns to return, or a fraction of the total number or rows or columns. It also has options for sampling with or without replacement, for passing in a column for weights for non-uniform sampling, and for setting seed values to facilitate replication. (GH2419)

In [1]: example_series = Series([0,1,2,3,4,5])

# When no arguments are passed, returns 1
In [2]: example_series.sample()
Out[2]: 
3    3
dtype: int64

# One may specify either a number of rows:
In [3]: example_series.sample(n=3)
Out[3]: 
5    5
1    1
4    4
dtype: int64

# Or a fraction of the rows:
In [4]: example_series.sample(frac=0.5)
Out[4]: 
4    4
1    1
0    0
dtype: int64

# weights are accepted.
In [5]: example_weights = [0, 0, 0.2, 0.2, 0.2, 0.4]

In [6]: example_series.sample(n=3, weights=example_weights)
Out[6]: 
2    2
3    3
5    5
dtype: int64

# weights will also be normalized if they do not sum to one,
# and missing values will be treated as zeros.
In [7]: example_weights2 = [0.5, 0, 0, 0, None, np.nan]

In [8]: example_series.sample(n=1, weights=example_weights2)
Out[8]: 
0    0
dtype: int64

When applied to a DataFrame, one may pass the name of a column to specify sampling weights when sampling from rows.

In [9]: df = DataFrame({'col1':[9,8,7,6], 'weight_column':[0.5, 0.4, 0.1, 0]})

In [10]: df.sample(n=3, weights='weight_column')
Out[10]: 
   col1  weight_column
0     9            0.5
1     8            0.4
2     7            0.1

String Methods Enhancements

Continuing from v0.16.0, the following enhancements make string operations easier and more consistent with standard python string operations.

  • Added StringMethods (.str accessor) to Index (GH9068)

    The .str accessor is now available for both Series and Index.

    In [11]: idx = Index([' jack', 'jill ', ' jesse ', 'frank'])
    
    In [12]: idx.str.strip()
    Out[12]: Index(['jack', 'jill', 'jesse', 'frank'], dtype='object')
    

    One special case for the .str accessor on Index is that if a string method returns bool, the .str accessor will return a np.array instead of a boolean Index (GH8875). This enables the following expression to work naturally:

    In [13]: idx = Index(['a1', 'a2', 'b1', 'b2'])
    
    In [14]: s = Series(range(4), index=idx)
    
    In [15]: s
    Out[15]: 
    a1    0
    a2    1
    b1    2
    b2    3
    dtype: int64
    
    In [16]: idx.str.startswith('a')
    Out[16]: array([ True,  True, False, False], dtype=bool)
    
    In [17]: s[s.index.str.startswith('a')]
    Out[17]: 
    a1    0
    a2    1
    dtype: int64
    
  • The following new methods are accesible via .str accessor to apply the function to each values. (GH9766, GH9773, GH10031, GH10045, GH10052)

    Methods
    capitalize() swapcase() normalize() partition() rpartition()
    index() rindex() translate()    
  • split now takes expand keyword to specify whether to expand dimensionality. return_type is deprecated. (GH9847)

    In [18]: s = Series(['a,b', 'a,c', 'b,c'])
    
    # return Series
    In [19]: s.str.split(',')
    Out[19]: 
    0    [a, b]
    1    [a, c]
    2    [b, c]
    dtype: object
    
    # return DataFrame
    In [20]: s.str.split(',', expand=True)
    Out[20]: 
       0  1
    0  a  b
    1  a  c
    2  b  c
    
    In [21]: idx = Index(['a,b', 'a,c', 'b,c'])
    
    # return Index
    In [22]: idx.str.split(',')
    Out[22]: Index([['a', 'b'], ['a', 'c'], ['b', 'c']], dtype='object')
    
    # return MultiIndex
    In [23]: idx.str.split(',', expand=True)
    Out[23]: 
    MultiIndex(levels=[['a', 'b'], ['b', 'c']],
               labels=[[0, 0, 1], [0, 1, 1]])
    
  • Improved extract and get_dummies methods for Index.str (GH9980)

Other Enhancements

  • BusinessHour offset is now supported, which represents business hours starting from 09:00 - 17:00 on BusinessDay by default. See Here for details. (GH7905)

    In [24]: from pandas.tseries.offsets import BusinessHour
    
    In [25]: Timestamp('2014-08-01 09:00') + BusinessHour()
    Out[25]: Timestamp('2014-08-01 10:00:00')
    
    In [26]: Timestamp('2014-08-01 07:00') + BusinessHour()
    Out[26]: Timestamp('2014-08-01 10:00:00')
    
    In [27]: Timestamp('2014-08-01 16:30') + BusinessHour()
    Out[27]: Timestamp('2014-08-04 09:30:00')
    
  • DataFrame.diff now takes an axis parameter that determines the direction of differencing (GH9727)

  • Allow clip, clip_lower, and clip_upper to accept array-like arguments as thresholds (This is a regression from 0.11.0). These methods now have an axis parameter which determines how the Series or DataFrame will be aligned with the threshold(s). (GH6966)

  • DataFrame.mask() and Series.mask() now support same keywords as where (GH8801)

  • drop function can now accept errors keyword to suppress ValueError raised when any of label does not exist in the target data. (GH6736)

    In [28]: df = DataFrame(np.random.randn(3, 3), columns=['A', 'B', 'C'])
    
    In [29]: df.drop(['A', 'X'], axis=1, errors='ignore')
    Out[29]: 
              B         C
    0  1.058969 -0.397840
    1  1.047579  1.045938
    2 -0.122092  0.124713
    
  • Add support for separating years and quarters using dashes, for example 2014-Q1. (GH9688)

  • Allow conversion of values with dtype datetime64 or timedelta64 to strings using astype(str) (GH9757)

  • get_dummies function now accepts sparse keyword. If set to True, the return DataFrame is sparse, e.g. SparseDataFrame. (GH8823)

  • Period now accepts datetime64 as value input. (GH9054)

  • Allow timedelta string conversion when leading zero is missing from time definition, ie 0:00:00 vs 00:00:00. (GH9570)

  • Allow Panel.shift with axis='items' (GH9890)

  • Trying to write an excel file now raises NotImplementedError if the DataFrame has a MultiIndex instead of writing a broken Excel file. (GH9794)

  • Allow Categorical.add_categories to accept Series or np.array. (GH9927)

  • Add/delete str/dt/cat accessors dynamically from __dir__. (GH9910)

  • Add normalize as a dt accessor method. (GH10047)

  • DataFrame and Series now have _constructor_expanddim property as overridable constructor for one higher dimensionality data. This should be used only when it is really needed, see here

  • pd.lib.infer_dtype now returns 'bytes' in Python 3 where appropriate. (GH10032)

API changes

  • When passing in an ax to df.plot( ..., ax=ax), the sharex kwarg will now default to False. The result is that the visibility of xlabels and xticklabels will not anymore be changed. You have to do that by yourself for the right axes in your figure or set sharex=True explicitly (but this changes the visible for all axes in the figure, not only the one which is passed in!). If pandas creates the subplots itself (e.g. no passed in ax kwarg), then the default is still sharex=True and the visibility changes are applied.
  • assign() now inserts new columns in alphabetical order. Previously the order was arbitrary. (GH9777)
  • By default, read_csv and read_table will now try to infer the compression type based on the file extension. Set compression=None to restore the previous behavior (no decompression). (GH9770)

Deprecations

  • Series.str.split’s return_type keyword was removed in favor of expand (GH9847)

Index Representation

The string representation of Index and its sub-classes have now been unified. These will show a single-line display if there are few values; a wrapped multi-line display for a lot of values (but less than display.max_seq_items; if lots of items (> display.max_seq_items) will show a truncated display (the head and tail of the data). The formatting for MultiIndex is unchanges (a multi-line wrapped display). The display width responds to the option display.max_seq_items, which is defaulted to 100. (GH6482)

Previous Behavior

In [2]: pd.Index(range(4),name='foo')
Out[2]: Int64Index([0, 1, 2, 3], dtype='int64')

In [3]: pd.Index(range(104),name='foo')
Out[3]: Int64Index([0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, ...], dtype='int64')

In [4]: pd.date_range('20130101',periods=4,name='foo',tz='US/Eastern')
Out[4]:
<class 'pandas.tseries.index.DatetimeIndex'>
[2013-01-01 00:00:00-05:00, ..., 2013-01-04 00:00:00-05:00]
Length: 4, Freq: D, Timezone: US/Eastern

In [5]: pd.date_range('20130101',periods=104,name='foo',tz='US/Eastern')
Out[5]:
<class 'pandas.tseries.index.DatetimeIndex'>
[2013-01-01 00:00:00-05:00, ..., 2013-04-14 00:00:00-04:00]
Length: 104, Freq: D, Timezone: US/Eastern

New Behavior

In [30]: pd.set_option('display.width', 80)

In [31]: pd.Index(range(4), name='foo')
Out[31]: RangeIndex(start=0, stop=4, step=1, name='foo')

In [32]: pd.Index(range(30), name='foo')
Out[32]: RangeIndex(start=0, stop=30, step=1, name='foo')

In [33]: pd.Index(range(104), name='foo')
Out[33]: RangeIndex(start=0, stop=104, step=1, name='foo')

In [34]: pd.CategoricalIndex(['a','bb','ccc','dddd'], ordered=True, name='foobar')
Out[34]: CategoricalIndex(['a', 'bb', 'ccc', 'dddd'], categories=['a', 'bb', 'ccc', 'dddd'], ordered=True, name='foobar', dtype='category')

In [35]: pd.CategoricalIndex(['a','bb','ccc','dddd']*10, ordered=True, name='foobar')
Out[35]: 
CategoricalIndex(['a', 'bb', 'ccc', 'dddd', 'a', 'bb', 'ccc', 'dddd', 'a',
                  'bb', 'ccc', 'dddd', 'a', 'bb', 'ccc', 'dddd', 'a', 'bb',
                  'ccc', 'dddd', 'a', 'bb', 'ccc', 'dddd', 'a', 'bb', 'ccc',
                  'dddd', 'a', 'bb', 'ccc', 'dddd', 'a', 'bb', 'ccc', 'dddd',
                  'a', 'bb', 'ccc', 'dddd'],
                 categories=['a', 'bb', 'ccc', 'dddd'], ordered=True, name='foobar', dtype='category')

In [36]: pd.CategoricalIndex(['a','bb','ccc','dddd']*100, ordered=True, name='foobar')
Out[36]: 
CategoricalIndex(['a', 'bb', 'ccc', 'dddd', 'a', 'bb', 'ccc', 'dddd', 'a',
                  'bb',
                  ...
                  'ccc', 'dddd', 'a', 'bb', 'ccc', 'dddd', 'a', 'bb', 'ccc',
                  'dddd'],
                 categories=['a', 'bb', 'ccc', 'dddd'], ordered=True, name='foobar', dtype='category', length=400)

In [37]: pd.date_range('20130101',periods=4, name='foo', tz='US/Eastern')
Out[37]: 
DatetimeIndex(['2013-01-01 00:00:00-05:00', '2013-01-02 00:00:00-05:00',
               '2013-01-03 00:00:00-05:00', '2013-01-04 00:00:00-05:00'],
              dtype='datetime64[ns, US/Eastern]', name='foo', freq='D')

In [38]: pd.date_range('20130101',periods=25, freq='D')
Out[38]: 
DatetimeIndex(['2013-01-01', '2013-01-02', '2013-01-03', '2013-01-04',
               '2013-01-05', '2013-01-06', '2013-01-07', '2013-01-08',
               '2013-01-09', '2013-01-10', '2013-01-11', '2013-01-12',
               '2013-01-13', '2013-01-14', '2013-01-15', '2013-01-16',
               '2013-01-17', '2013-01-18', '2013-01-19', '2013-01-20',
               '2013-01-21', '2013-01-22', '2013-01-23', '2013-01-24',
               '2013-01-25'],
              dtype='datetime64[ns]', freq='D')

In [39]: pd.date_range('20130101',periods=104, name='foo', tz='US/Eastern')
Out[39]: 
DatetimeIndex(['2013-01-01 00:00:00-05:00', '2013-01-02 00:00:00-05:00',
               '2013-01-03 00:00:00-05:00', '2013-01-04 00:00:00-05:00',
               '2013-01-05 00:00:00-05:00', '2013-01-06 00:00:00-05:00',
               '2013-01-07 00:00:00-05:00', '2013-01-08 00:00:00-05:00',
               '2013-01-09 00:00:00-05:00', '2013-01-10 00:00:00-05:00',
               ...
               '2013-04-05 00:00:00-04:00', '2013-04-06 00:00:00-04:00',
               '2013-04-07 00:00:00-04:00', '2013-04-08 00:00:00-04:00',
               '2013-04-09 00:00:00-04:00', '2013-04-10 00:00:00-04:00',
               '2013-04-11 00:00:00-04:00', '2013-04-12 00:00:00-04:00',
               '2013-04-13 00:00:00-04:00', '2013-04-14 00:00:00-04:00'],
              dtype='datetime64[ns, US/Eastern]', name='foo', length=104, freq='D')

Performance Improvements

  • Improved csv write performance with mixed dtypes, including datetimes by up to 5x (GH9940)
  • Improved csv write performance generally by 2x (GH9940)
  • Improved the performance of pd.lib.max_len_string_array by 5-7x (GH10024)

Bug Fixes

  • Bug where labels did not appear properly in the legend of DataFrame.plot(), passing label= arguments works, and Series indices are no longer mutated. (GH9542)
  • Bug in json serialization causing a segfault when a frame had zero length. (GH9805)
  • Bug in read_csv where missing trailing delimiters would cause segfault. (GH5664)
  • Bug in retaining index name on appending (GH9862)
  • Bug in scatter_matrix draws unexpected axis ticklabels (GH5662)
  • Fixed bug in StataWriter resulting in changes to input DataFrame upon save (GH9795).
  • Bug in transform causing length mismatch when null entries were present and a fast aggregator was being used (GH9697)
  • Bug in equals causing false negatives when block order differed (GH9330)
  • Bug in grouping with multiple pd.Grouper where one is non-time based (GH10063)
  • Bug in read_sql_table error when reading postgres table with timezone (GH7139)
  • Bug in DataFrame slicing may not retain metadata (GH9776)
  • Bug where TimdeltaIndex were not properly serialized in fixed HDFStore (GH9635)
  • Bug with TimedeltaIndex constructor ignoring name when given another TimedeltaIndex as data (GH10025).
  • Bug in DataFrameFormatter._get_formatted_index with not applying max_colwidth to the DataFrame index (GH7856)
  • Bug in .loc with a read-only ndarray data source (GH10043)
  • Bug in groupby.apply() that would raise if a passed user defined function either returned only None (for all input). (GH9685)
  • Always use temporary files in pytables tests (GH9992)
  • Bug in plotting continuously using secondary_y may not show legend properly. (GH9610, GH9779)
  • Bug in DataFrame.plot(kind="hist") results in TypeError when DataFrame contains non-numeric columns (GH9853)
  • Bug where repeated plotting of DataFrame with a DatetimeIndex may raise TypeError (GH9852)
  • Bug in setup.py that would allow an incompat cython version to build (GH9827)
  • Bug in plotting secondary_y incorrectly attaches right_ax property to secondary axes specifying itself recursively. (GH9861)
  • Bug in Series.quantile on empty Series of type Datetime or Timedelta (GH9675)
  • Bug in where causing incorrect results when upcasting was required (GH9731)
  • Bug in FloatArrayFormatter where decision boundary for displaying “small” floats in decimal format is off by one order of magnitude for a given display.precision (GH9764)
  • Fixed bug where DataFrame.plot() raised an error when both color and style keywords were passed and there was no color symbol in the style strings (GH9671)
  • Not showing a DeprecationWarning on combining list-likes with an Index (GH10083)
  • Bug in read_csv and read_table when using skip_rows parameter if blank lines are present. (GH9832)
  • Bug in read_csv() interprets index_col=True as 1 (GH9798)
  • Bug in index equality comparisons using == failing on Index/MultiIndex type incompatibility (GH9785)
  • Bug in which SparseDataFrame could not take nan as a column name (GH8822)
  • Bug in to_msgpack and read_msgpack zlib and blosc compression support (GH9783)
  • Bug GroupBy.size doesn’t attach index name properly if grouped by TimeGrouper (GH9925)
  • Bug causing an exception in slice assignments because length_of_indexer returns wrong results (GH9995)
  • Bug in csv parser causing lines with initial whitespace plus one non-space character to be skipped. (GH9710)
  • Bug in C csv parser causing spurious NaNs when data started with newline followed by whitespace. (GH10022)
  • Bug causing elements with a null group to spill into the final group when grouping by a Categorical (GH9603)
  • Bug where .iloc and .loc behavior is not consistent on empty dataframes (GH9964)
  • Bug in invalid attribute access on a TimedeltaIndex incorrectly raised ValueError instead of AttributeError (GH9680)
  • Bug in unequal comparisons between categorical data and a scalar, which was not in the categories (e.g. Series(Categorical(list("abc"), ordered=True)) > "d". This returned False for all elements, but now raises a TypeError. Equality comparisons also now return False for == and True for !=. (GH9848)
  • Bug in DataFrame __setitem__ when right hand side is a dictionary (GH9874)
  • Bug in where when dtype is datetime64/timedelta64, but dtype of other is not (GH9804)
  • Bug in MultiIndex.sortlevel() results in unicode level name breaks (GH9856)
  • Bug in which groupby.transform incorrectly enforced output dtypes to match input dtypes. (GH9807)
  • Bug in DataFrame constructor when columns parameter is set, and data is an empty list (GH9939)
  • Bug in bar plot with log=True raises TypeError if all values are less than 1 (GH9905)
  • Bug in horizontal bar plot ignores log=True (GH9905)
  • Bug in PyTables queries that did not return proper results using the index (GH8265, GH9676)
  • Bug where dividing a dataframe containing values of type Decimal by another Decimal would raise. (GH9787)
  • Bug where using DataFrames asfreq would remove the name of the index. (GH9885)
  • Bug causing extra index point when resample BM/BQ (GH9756)
  • Changed caching in AbstractHolidayCalendar to be at the instance level rather than at the class level as the latter can result in unexpected behaviour. (GH9552)
  • Fixed latex output for multi-indexed dataframes (GH9778)
  • Bug causing an exception when setting an empty range using DataFrame.loc (GH9596)
  • Bug in hiding ticklabels with subplots and shared axes when adding a new plot to an existing grid of axes (GH9158)
  • Bug in transform and filter when grouping on a categorical variable (GH9921)
  • Bug in transform when groups are equal in number and dtype to the input index (GH9700)
  • Google BigQuery connector now imports dependencies on a per-method basis.(GH9713)
  • Updated BigQuery connector to no longer use deprecated oauth2client.tools.run() (GH8327)
  • Bug in subclassed DataFrame. It may not return the correct class, when slicing or subsetting it. (GH9632)
  • Bug in .median() where non-float null values are not handled correctly (GH10040)
  • Bug in Series.fillna() where it raises if a numerically convertible string is given (GH10092)

v0.16.0 (March 22, 2015)

This is a major release from 0.15.2 and includes a small number of API changes, several new features, enhancements, and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this version.

Highlights include:

  • DataFrame.assign method, see here
  • Series.to_coo/from_coo methods to interact with scipy.sparse, see here
  • Backwards incompatible change to Timedelta to conform the .seconds attribute with datetime.timedelta, see here
  • Changes to the .loc slicing API to conform with the behavior of .ix see here
  • Changes to the default for ordering in the Categorical constructor, see here
  • Enhancement to the .str accessor to make string operations easier, see here
  • The pandas.tools.rplot, pandas.sandbox.qtpandas and pandas.rpy modules are deprecated. We refer users to external packages like seaborn, pandas-qt and rpy2 for similar or equivalent functionality, see here

Check the API Changes and deprecations before updating.

New features

DataFrame Assign

Inspired by dplyr’s mutate verb, DataFrame has a new assign() method. The function signature for assign is simply **kwargs. The keys are the column names for the new fields, and the values are either a value to be inserted (for example, a Series or NumPy array), or a function of one argument to be called on the DataFrame. The new values are inserted, and the entire DataFrame (with all original and new columns) is returned.

In [1]: iris = read_csv('data/iris.data')

In [2]: iris.head()
Out[2]: 
   SepalLength  SepalWidth  PetalLength  PetalWidth         Name
0          5.1         3.5          1.4         0.2  Iris-setosa
1          4.9         3.0          1.4         0.2  Iris-setosa
2          4.7         3.2          1.3         0.2  Iris-setosa
3          4.6         3.1          1.5         0.2  Iris-setosa
4          5.0         3.6          1.4         0.2  Iris-setosa

In [3]: iris.assign(sepal_ratio=iris['SepalWidth'] / iris['SepalLength']).head()
Out[3]: 
   SepalLength  SepalWidth  PetalLength  PetalWidth         Name  sepal_ratio
0          5.1         3.5          1.4         0.2  Iris-setosa     0.686275
1          4.9         3.0          1.4         0.2  Iris-setosa     0.612245
2          4.7         3.2          1.3         0.2  Iris-setosa     0.680851
3          4.6         3.1          1.5         0.2  Iris-setosa     0.673913
4          5.0         3.6          1.4         0.2  Iris-setosa     0.720000

Above was an example of inserting a precomputed value. We can also pass in a function to be evaluated.

In [4]: iris.assign(sepal_ratio = lambda x: (x['SepalWidth'] /
   ...:                                      x['SepalLength'])).head()
   ...: 
Out[4]: 
   SepalLength  SepalWidth  PetalLength  PetalWidth         Name  sepal_ratio
0          5.1         3.5          1.4         0.2  Iris-setosa     0.686275
1          4.9         3.0          1.4         0.2  Iris-setosa     0.612245
2          4.7         3.2          1.3         0.2  Iris-setosa     0.680851
3          4.6         3.1          1.5         0.2  Iris-setosa     0.673913
4          5.0         3.6          1.4         0.2  Iris-setosa     0.720000

The power of assign comes when used in chains of operations. For example, we can limit the DataFrame to just those with a Sepal Length greater than 5, calculate the ratio, and plot

In [5]: (iris.query('SepalLength > 5')
   ...:      .assign(SepalRatio = lambda x: x.SepalWidth / x.SepalLength,
   ...:              PetalRatio = lambda x: x.PetalWidth / x.PetalLength)
   ...:      .plot(kind='scatter', x='SepalRatio', y='PetalRatio'))
   ...: 
Out[5]: <matplotlib.axes._subplots.AxesSubplot at 0x1c41df41d0>
_images/whatsnew_assign.png

See the documentation for more. (GH9229)

Interaction with scipy.sparse

Added SparseSeries.to_coo() and SparseSeries.from_coo() methods (GH8048) for converting to and from scipy.sparse.coo_matrix instances (see here). For example, given a SparseSeries with MultiIndex we can convert to a scipy.sparse.coo_matrix by specifying the row and column labels as index levels:

In [6]: from numpy import nan

In [7]: s = Series([3.0, nan, 1.0, 3.0, nan, nan])

In [8]: s.index = MultiIndex.from_tuples([(1, 2, 'a', 0),
   ...:                                   (1, 2, 'a', 1),
   ...:                                   (1, 1, 'b', 0),
   ...:                                   (1, 1, 'b', 1),
   ...:                                   (2, 1, 'b', 0),
   ...:                                   (2, 1, 'b', 1)],
   ...:                                   names=['A', 'B', 'C', 'D'])
   ...: 

In [9]: s
Out[9]: 
A  B  C  D
1  2  a  0    3.0
         1    NaN
   1  b  0    1.0
         1    3.0
2  1  b  0    NaN
         1    NaN
dtype: float64

# SparseSeries
In [10]: ss = s.to_sparse()

In [11]: ss
Out[11]: 
A  B  C  D
1  2  a  0    3.0
         1    NaN
   1  b  0    1.0
         1    3.0
2  1  b  0    NaN
         1    NaN
dtype: float64
BlockIndex
Block locations: array([0, 2], dtype=int32)
Block lengths: array([1, 2], dtype=int32)

In [12]: A, rows, columns = ss.to_coo(row_levels=['A', 'B'],
   ....:                              column_levels=['C', 'D'],
   ....:                              sort_labels=False)
   ....: 

In [13]: A
Out[13]: 
<3x4 sparse matrix of type '<class 'numpy.float64'>'
	with 3 stored elements in COOrdinate format>

In [14]: A.todense()
Out[14]: 
matrix([[ 3.,  0.,  0.,  0.],
        [ 0.,  0.,  1.,  3.],
        [ 0.,  0.,  0.,  0.]])

In [15]: rows
Out[15]: [(1, 2), (1, 1), (2, 1)]

In [16]: columns
Out[16]: [('a', 0), ('a', 1), ('b', 0), ('b', 1)]

The from_coo method is a convenience method for creating a SparseSeries from a scipy.sparse.coo_matrix:

In [17]: from scipy import sparse

In [18]: A = sparse.coo_matrix(([3.0, 1.0, 2.0], ([1, 0, 0], [0, 2, 3])),
   ....:                             shape=(3, 4))
   ....: 

In [19]: A
Out[19]: 
<3x4 sparse matrix of type '<class 'numpy.float64'>'
	with 3 stored elements in COOrdinate format>

In [20]: A.todense()
Out[20]: 
matrix([[ 0.,  0.,  1.,  2.],
        [ 3.,  0.,  0.,  0.],
        [ 0.,  0.,  0.,  0.]])

In [21]: ss = SparseSeries.from_coo(A)

In [22]: ss
Out[22]: 
0  2    1.0
   3    2.0
1  0    3.0
dtype: float64
BlockIndex
Block locations: array([0], dtype=int32)
Block lengths: array([3], dtype=int32)

String Methods Enhancements

  • Following new methods are accesible via .str accessor to apply the function to each values. This is intended to make it more consistent with standard methods on strings. (GH9282, GH9352, GH9386, GH9387, GH9439)

    Methods
    isalnum() isalpha() isdigit() isdigit() isspace()
    islower() isupper() istitle() isnumeric() isdecimal()
    find() rfind() ljust() rjust() zfill()
    In [23]: s = Series(['abcd', '3456', 'EFGH'])
    
    In [24]: s.str.isalpha()
    Out[24]: 
    0     True
    1    False
    2     True
    dtype: bool
    
    In [25]: s.str.find('ab')
    Out[25]: 
    0    0
    1   -1
    2   -1
    dtype: int64
    
  • Series.str.pad() and Series.str.center() now accept fillchar option to specify filling character (GH9352)

    In [26]: s = Series(['12', '300', '25'])
    
    In [27]: s.str.pad(5, fillchar='_')
    Out[27]: 
    0    ___12
    1    __300
    2    ___25
    dtype: object
    
  • Added Series.str.slice_replace(), which previously raised NotImplementedError (GH8888)

    In [28]: s = Series(['ABCD', 'EFGH', 'IJK'])
    
    In [29]: s.str.slice_replace(1, 3, 'X')
    Out[29]: 
    0    AXD
    1    EXH
    2     IX
    dtype: object
    
    # replaced with empty char
    In [30]: s.str.slice_replace(0, 1)
    Out[30]: 
    0    BCD
    1    FGH
    2     JK
    dtype: object
    

Other enhancements

  • Reindex now supports method='nearest' for frames or series with a monotonic increasing or decreasing index (GH9258):

    In [31]: df = pd.DataFrame({'x': range(5)})
    
    In [32]: df.reindex([0.2, 1.8, 3.5], method='nearest')
    Out[32]: 
         x
    0.2  0
    1.8  2
    3.5  4
    

    This method is also exposed by the lower level Index.get_indexer and Index.get_loc methods.

  • The read_excel() function’s sheetname argument now accepts a list and None, to get multiple or all sheets respectively. If more than one sheet is specified, a dictionary is returned. (GH9450)

    # Returns the 1st and 4th sheet, as a dictionary of DataFrames.
    pd.read_excel('path_to_file.xls',sheetname=['Sheet1',3])
    
  • Allow Stata files to be read incrementally with an iterator; support for long strings in Stata files. See the docs here (GH9493:).

  • Paths beginning with ~ will now be expanded to begin with the user’s home directory (GH9066)

  • Added time interval selection in get_data_yahoo (GH9071)

  • Added Timestamp.to_datetime64() to complement Timedelta.to_timedelta64() (GH9255)

  • tseries.frequencies.to_offset() now accepts Timedelta as input (GH9064)

  • Lag parameter was added to the autocorrelation method of Series, defaults to lag-1 autocorrelation (GH9192)

  • Timedelta will now accept nanoseconds keyword in constructor (GH9273)

  • SQL code now safely escapes table and column names (GH8986)

  • Added auto-complete for Series.str.<tab>, Series.dt.<tab> and Series.cat.<tab> (GH9322)

  • Index.get_indexer now supports method='pad' and method='backfill' even for any target array, not just monotonic targets. These methods also work for monotonic decreasing as well as monotonic increasing indexes (GH9258).

  • Index.asof now works on all index types (GH9258).

  • A verbose argument has been augmented in io.read_excel(), defaults to False. Set to True to print sheet names as they are parsed. (GH9450)

  • Added days_in_month (compatibility alias daysinmonth) property to Timestamp, DatetimeIndex, Period, PeriodIndex, and Series.dt (GH9572)

  • Added decimal option in to_csv to provide formatting for non-‘.’ decimal separators (GH781)

  • Added normalize option for Timestamp to normalized to midnight (GH8794)

  • Added example for DataFrame import to R using HDF5 file and rhdf5 library. See the documentation for more (GH9636).

Backwards incompatible API changes

Changes in Timedelta

In v0.15.0 a new scalar type Timedelta was introduced, that is a sub-class of datetime.timedelta. Mentioned here was a notice of an API change w.r.t. the .seconds accessor. The intent was to provide a user-friendly set of accessors that give the ‘natural’ value for that unit, e.g. if you had a Timedelta('1 day, 10:11:12'), then .seconds would return 12. However, this is at odds with the definition of datetime.timedelta, which defines .seconds as 10 * 3600 + 11 * 60 + 12 == 36672.

So in v0.16.0, we are restoring the API to match that of datetime.timedelta. Further, the component values are still available through the .components accessor. This affects the .seconds and .microseconds accessors, and removes the .hours, .minutes, .milliseconds accessors. These changes affect TimedeltaIndex and the Series .dt accessor as well. (GH9185, GH9139)

Previous Behavior

In [2]: t = pd.Timedelta('1 day, 10:11:12.100123')

In [3]: t.days
Out[3]: 1

In [4]: t.seconds
Out[4]: 12

In [5]: t.microseconds
Out[5]: 123

New Behavior

In [33]: t = pd.Timedelta('1 day, 10:11:12.100123')

In [34]: t.days
Out[34]: 1

In [35]: t.seconds
Out[35]: 36672

In [36]: t.microseconds
Out[36]: 100123

Using .components allows the full component access

In [37]: t.components
Out[37]: Components(days=1, hours=10, minutes=11, seconds=12, milliseconds=100, microseconds=123, nanoseconds=0)

In [38]: t.components.seconds
Out[38]: 12

Indexing Changes

The behavior of a small sub-set of edge cases for using .loc have changed (GH8613). Furthermore we have improved the content of the error messages that are raised:

  • Slicing with .loc where the start and/or stop bound is not found in the index is now allowed; this previously would raise a KeyError. This makes the behavior the same as .ix in this case. This change is only for slicing, not when indexing with a single label.

    In [39]: df = DataFrame(np.random.randn(5,4),
       ....:                columns=list('ABCD'),
       ....:                index=date_range('20130101',periods=5))
       ....: 
    
    In [40]: df
    Out[40]: 
                       A         B         C         D
    2013-01-01 -0.322795  0.841675  2.390961  0.076200
    2013-01-02 -0.566446  0.036142 -2.074978  0.247792
    2013-01-03 -0.897157 -0.136795  0.018289  0.755414
    2013-01-04  0.215269  0.841009 -1.445810 -1.401973
    2013-01-05 -0.100918 -0.548242 -0.144620  0.354020
    
    In [41]: s = Series(range(5),[-2,-1,1,2,3])
    
    In [42]: s
    Out[42]: 
    -2    0
    -1    1
     1    2
     2    3
     3    4
    dtype: int64
    

    Previous Behavior

    In [4]: df.loc['2013-01-02':'2013-01-10']
    KeyError: 'stop bound [2013-01-10] is not in the [index]'
    
    In [6]: s.loc[-10:3]
    KeyError: 'start bound [-10] is not the [index]'
    

    New Behavior

    In [43]: df.loc['2013-01-02':'2013-01-10']
    Out[43]: 
                       A         B         C         D
    2013-01-02 -0.566446  0.036142 -2.074978  0.247792
    2013-01-03 -0.897157 -0.136795  0.018289  0.755414
    2013-01-04  0.215269  0.841009 -1.445810 -1.401973
    2013-01-05 -0.100918 -0.548242 -0.144620  0.354020
    
    In [44]: s.loc[-10:3]
    Out[44]: 
    -2    0
    -1    1
     1    2
     2    3
     3    4
    dtype: int64
    
  • Allow slicing with float-like values on an integer index for .ix. Previously this was only enabled for .loc:

    Previous Behavior

    In [8]: s.ix[-1.0:2]
    TypeError: the slice start value [-1.0] is not a proper indexer for this index type (Int64Index)
    

    New Behavior

    In [2]: s.ix[-1.0:2]
    Out[2]:
    -1    1
     1    2
     2    3
    dtype: int64
    
  • Provide a useful exception for indexing with an invalid type for that index when using .loc. For example trying to use .loc on an index of type DatetimeIndex or PeriodIndex or TimedeltaIndex, with an integer (or a float).

    Previous Behavior

    In [4]: df.loc[2:3]
    KeyError: 'start bound [2] is not the [index]'
    

    New Behavior

    In [4]: df.loc[2:3]
    TypeError: Cannot do slice indexing on <class 'pandas.tseries.index.DatetimeIndex'> with <type 'int'> keys
    

Categorical Changes

In prior versions, Categoricals that had an unspecified ordering (meaning no ordered keyword was passed) were defaulted as ordered Categoricals. Going forward, the ordered keyword in the Categorical constructor will default to False. Ordering must now be explicit.

Furthermore, previously you could change the ordered attribute of a Categorical by just setting the attribute, e.g. cat.ordered=True; This is now deprecated and you should use cat.as_ordered() or cat.as_unordered(). These will by default return a new object and not modify the existing object. (GH9347, GH9190)

Previous Behavior

In [3]: s = Series([0,1,2], dtype='category')

In [4]: s
Out[4]:
0    0
1    1
2    2
dtype: category
Categories (3, int64): [0 < 1 < 2]

In [5]: s.cat.ordered
Out[5]: True

In [6]: s.cat.ordered = False

In [7]: s
Out[7]:
0    0
1    1
2    2
dtype: category
Categories (3, int64): [0, 1, 2]

New Behavior

In [45]: s = Series([0,1,2], dtype='category')

In [46]: s
Out[46]: 
0    0
1    1
2    2
dtype: category
Categories (3, int64): [0, 1, 2]

In [47]: s.cat.ordered
Out[47]: False

In [48]: s = s.cat.as_ordered()

In [49]: s
Out[49]: 
0    0
1    1
2    2
dtype: category
Categories (3, int64): [0 < 1 < 2]

In [50]: s.cat.ordered
Out[50]: True

# you can set in the constructor of the Categorical
In [51]: s = Series(Categorical([0,1,2],ordered=True))

In [52]: s
Out[52]: 
0    0
1    1
2    2
dtype: category
Categories (3, int64): [0 < 1 < 2]

In [53]: s.cat.ordered
Out[53]: True

For ease of creation of series of categorical data, we have added the ability to pass keywords when calling .astype(). These are passed directly to the constructor.

In [54]: s = Series(["a","b","c","a"]).astype('category',ordered=True)

In [55]: s
Out[55]:
0    a
1    b
2    c
3    a
dtype: category
Categories (3, object): [a < b < c]

In [56]: s = Series(["a","b","c","a"]).astype('category',categories=list('abcdef'),ordered=False)

In [57]: s
Out[57]:
0    a
1    b
2    c
3    a
dtype: category
Categories (6, object): [a, b, c, d, e, f]

Other API Changes

  • Index.duplicated now returns np.array(dtype=bool) rather than Index(dtype=object) containing bool values. (GH8875)

  • DataFrame.to_json now returns accurate type serialisation for each column for frames of mixed dtype (GH9037)

    Previously data was coerced to a common dtype before serialisation, which for example resulted in integers being serialised to floats:

    In [2]: pd.DataFrame({'i': [1,2], 'f': [3.0, 4.2]}).to_json()
    Out[2]: '{"f":{"0":3.0,"1":4.2},"i":{"0":1.0,"1":2.0}}'
    

    Now each column is serialised using its correct dtype:

    In [2]:  pd.DataFrame({'i': [1,2], 'f': [3.0, 4.2]}).to_json()
    Out[2]: '{"f":{"0":3.0,"1":4.2},"i":{"0":1,"1":2}}'
    
  • DatetimeIndex, PeriodIndex and TimedeltaIndex.summary now output the same format. (GH9116)

  • TimedeltaIndex.freqstr now output the same string format as DatetimeIndex. (GH9116)

  • Bar and horizontal bar plots no longer add a dashed line along the info axis. The prior style can be achieved with matplotlib’s axhline or axvline methods (GH9088).

  • Series accessors .dt, .cat and .str now raise AttributeError instead of TypeError if the series does not contain the appropriate type of data (GH9617). This follows Python’s built-in exception hierarchy more closely and ensures that tests like hasattr(s, 'cat') are consistent on both Python 2 and 3.

  • Series now supports bitwise operation for integral types (GH9016). Previously even if the input dtypes were integral, the output dtype was coerced to bool.

    Previous Behavior

    In [2]: pd.Series([0,1,2,3], list('abcd')) | pd.Series([4,4,4,4], list('abcd'))
    Out[2]:
    a    True
    b    True
    c    True
    d    True
    dtype: bool
    

    New Behavior. If the input dtypes are integral, the output dtype is also integral and the output values are the result of the bitwise operation.

    In [2]: pd.Series([0,1,2,3], list('abcd')) | pd.Series([4,4,4,4], list('abcd'))
    Out[2]:
    a    4
    b    5
    c    6
    d    7
    dtype: int64
    
  • During division involving a Series or DataFrame, 0/0 and 0//0 now give np.nan instead of np.inf. (GH9144, GH8445)

    Previous Behavior

    In [2]: p = pd.Series([0, 1])
    
    In [3]: p / 0
    Out[3]:
    0    inf
    1    inf
    dtype: float64
    
    In [4]: p // 0
    Out[4]:
    0    inf
    1    inf
    dtype: float64
    

    New Behavior

    In [54]: p = pd.Series([0, 1])
    
    In [55]: p / 0
    Out[55]: 
    0    NaN
    1    inf
    dtype: float64
    
    In [56]: p // 0
    Out[56]: 
    0    NaN
    1    inf
    dtype: float64
    
  • Series.values_counts and Series.describe for categorical data will now put NaN entries at the end. (GH9443)

  • Series.describe for categorical data will now give counts and frequencies of 0, not NaN, for unused categories (GH9443)

  • Due to a bug fix, looking up a partial string label with DatetimeIndex.asof now includes values that match the string, even if they are after the start of the partial string label (GH9258).

    Old behavior:

    In [4]: pd.to_datetime(['2000-01-31', '2000-02-28']).asof('2000-02')
    Out[4]: Timestamp('2000-01-31 00:00:00')
    

    Fixed behavior:

    In [57]: pd.to_datetime(['2000-01-31', '2000-02-28']).asof('2000-02')
    Out[57]: Timestamp('2000-02-28 00:00:00')
    

    To reproduce the old behavior, simply add more precision to the label (e.g., use 2000-02-01 instead of 2000-02).

Deprecations

  • The rplot trellis plotting interface is deprecated and will be removed in a future version. We refer to external packages like seaborn for similar but more refined functionality (GH3445). The documentation includes some examples how to convert your existing code using rplot to seaborn: rplot docs.
  • The pandas.sandbox.qtpandas interface is deprecated and will be removed in a future version. We refer users to the external package pandas-qt. (GH9615)
  • The pandas.rpy interface is deprecated and will be removed in a future version. Similar functionaility can be accessed thru the rpy2 project (GH9602)
  • Adding DatetimeIndex/PeriodIndex to another DatetimeIndex/PeriodIndex is being deprecated as a set-operation. This will be changed to a TypeError in a future version. .union() should be used for the union set operation. (GH9094)
  • Subtracting DatetimeIndex/PeriodIndex from another DatetimeIndex/PeriodIndex is being deprecated as a set-operation. This will be changed to an actual numeric subtraction yielding a TimeDeltaIndex in a future version. .difference() should be used for the differencing set operation. (GH9094)

Removal of prior version deprecations/changes

  • DataFrame.pivot_table and crosstab’s rows and cols keyword arguments were removed in favor of index and columns (GH6581)
  • DataFrame.to_excel and DataFrame.to_csv cols keyword argument was removed in favor of columns (GH6581)
  • Removed convert_dummies in favor of get_dummies (GH6581)
  • Removed value_range in favor of describe (GH6581)

Performance Improvements

  • Fixed a performance regression for .loc indexing with an array or list-like (GH9126:).
  • DataFrame.to_json 30x performance improvement for mixed dtype frames. (GH9037)
  • Performance improvements in MultiIndex.duplicated by working with labels instead of values (GH9125)
  • Improved the speed of nunique by calling unique instead of value_counts (GH9129, GH7771)
  • Performance improvement of up to 10x in DataFrame.count and DataFrame.dropna by taking advantage of homogeneous/heterogeneous dtypes appropriately (GH9136)
  • Performance improvement of up to 20x in DataFrame.count when using a MultiIndex and the level keyword argument (GH9163)
  • Performance and memory usage improvements in merge when key space exceeds int64 bounds (GH9151)
  • Performance improvements in multi-key groupby (GH9429)
  • Performance improvements in MultiIndex.sortlevel (GH9445)
  • Performance and memory usage improvements in DataFrame.duplicated (GH9398)
  • Cythonized Period (GH9440)
  • Decreased memory usage on to_hdf (GH9648)

Bug Fixes

  • Changed .to_html to remove leading/trailing spaces in table body (GH4987)
  • Fixed issue using read_csv on s3 with Python 3 (GH9452)
  • Fixed compatibility issue in DatetimeIndex affecting architectures where numpy.int_ defaults to numpy.int32 (GH8943)
  • Bug in Panel indexing with an object-like (GH9140)
  • Bug in the returned Series.dt.components index was reset to the default index (GH9247)
  • Bug in Categorical.__getitem__/__setitem__ with listlike input getting incorrect results from indexer coercion (GH9469)
  • Bug in partial setting with a DatetimeIndex (GH9478)
  • Bug in groupby for integer and datetime64 columns when applying an aggregator that caused the value to be changed when the number was sufficiently large (GH9311, GH6620)
  • Fixed bug in to_sql when mapping a Timestamp object column (datetime column with timezone info) to the appropriate sqlalchemy type (GH9085).
  • Fixed bug in to_sql dtype argument not accepting an instantiated SQLAlchemy type (GH9083).
  • Bug in .loc partial setting with a np.datetime64 (GH9516)
  • Incorrect dtypes inferred on datetimelike looking Series & on .xs slices (GH9477)
  • Items in Categorical.unique() (and s.unique() if s is of dtype category) now appear in the order in which they are originally found, not in sorted order (GH9331). This is now consistent with the behavior for other dtypes in pandas.
  • Fixed bug on big endian platforms which produced incorrect results in StataReader (GH8688).
  • Bug in MultiIndex.has_duplicates when having many levels causes an indexer overflow (GH9075, GH5873)
  • Bug in pivot and unstack where nan values would break index alignment (GH4862, GH7401, GH7403, GH7405, GH7466, GH9497)
  • Bug in left join on multi-index with sort=True or null values (GH9210).
  • Bug in MultiIndex where inserting new keys would fail (GH9250).
  • Bug in groupby when key space exceeds int64 bounds (GH9096).
  • Bug in unstack with TimedeltaIndex or DatetimeIndex and nulls (GH9491).
  • Bug in rank where comparing floats with tolerance will cause inconsistent behaviour (GH8365).
  • Fixed character encoding bug in read_stata and StataReader when loading data from a URL (GH9231).
  • Bug in adding offsets.Nano to other offsets raises TypeError (GH9284)
  • Bug in DatetimeIndex iteration, related to (GH8890), fixed in (GH9100)
  • Bugs in resample around DST transitions. This required fixing offset classes so they behave correctly on DST transitions. (GH5172, GH8744, GH8653, GH9173, GH9468).
  • Bug in binary operator method (eg .mul()) alignment with integer levels (GH9463).
  • Bug in boxplot, scatter and hexbin plot may show an unnecessary warning (GH8877)
  • Bug in subplot with layout kw may show unnecessary warning (GH9464)
  • Bug in using grouper functions that need passed thru arguments (e.g. axis), when using wrapped function (e.g. fillna), (GH9221)
  • DataFrame now properly supports simultaneous copy and dtype arguments in constructor (GH9099)
  • Bug in read_csv when using skiprows on a file with CR line endings with the c engine. (GH9079)
  • isnull now detects NaT in PeriodIndex (GH9129)
  • Bug in groupby .nth() with a multiple column groupby (GH8979)
  • Bug in DataFrame.where and Series.where coerce numerics to string incorrectly (GH9280)
  • Bug in DataFrame.where and Series.where raise ValueError when string list-like is passed. (GH9280)
  • Accessing Series.str methods on with non-string values now raises TypeError instead of producing incorrect results (GH9184)
  • Bug in DatetimeIndex.__contains__ when index has duplicates and is not monotonic increasing (GH9512)
  • Fixed division by zero error for Series.kurt() when all values are equal (GH9197)
  • Fixed issue in the xlsxwriter engine where it added a default ‘General’ format to cells if no other format was applied. This prevented other row or column formatting being applied. (GH9167)
  • Fixes issue with index_col=False when usecols is also specified in read_csv. (GH9082)
  • Bug where wide_to_long would modify the input stubnames list (GH9204)
  • Bug in to_sql not storing float64 values using double precision. (GH9009)
  • SparseSeries and SparsePanel now accept zero argument constructors (same as their non-sparse counterparts) (GH9272).
  • Regression in merging Categorical and object dtypes (GH9426)
  • Bug in read_csv with buffer overflows with certain malformed input files (GH9205)
  • Bug in groupby MultiIndex with missing pair (GH9049, GH9344)
  • Fixed bug in Series.groupby where grouping on MultiIndex levels would ignore the sort argument (GH9444)
  • Fix bug in DataFrame.Groupby where sort=False is ignored in the case of Categorical columns. (GH8868)
  • Fixed bug with reading CSV files from Amazon S3 on python 3 raising a TypeError (GH9452)
  • Bug in the Google BigQuery reader where the ‘jobComplete’ key may be present but False in the query results (GH8728)
  • Bug in Series.values_counts with excluding NaN for categorical type Series with dropna=True (GH9443)
  • Fixed mising numeric_only option for DataFrame.std/var/sem (GH9201)
  • Support constructing Panel or Panel4D with scalar data (GH8285)
  • Series text representation disconnected from max_rows/max_columns (GH7508).

  • Series number formatting inconsistent when truncated (GH8532).

    Previous Behavior

    In [2]: pd.options.display.max_rows = 10
    In [3]: s = pd.Series([1,1,1,1,1,1,1,1,1,1,0.9999,1,1]*10)
    In [4]: s
    Out[4]:
    0    1
    1    1
    2    1
    ...
    127    0.9999
    128    1.0000
    129    1.0000
    Length: 130, dtype: float64
    

    New Behavior

    0      1.0000
    1      1.0000
    2      1.0000
    3      1.0000
    4      1.0000
    ...
    125    1.0000
    126    1.0000
    127    0.9999
    128    1.0000
    129    1.0000
    dtype: float64
    
  • A Spurious SettingWithCopy Warning was generated when setting a new item in a frame in some cases (GH8730)

    The following would previously report a SettingWithCopy Warning.

    In [1]: df1 = DataFrame({'x': Series(['a','b','c']), 'y': Series(['d','e','f'])})
    
    In [2]: df2 = df1[['x']]
    
    In [3]: df2['y'] = ['g', 'h', 'i']
    

v0.15.2 (December 12, 2014)

This is a minor release from 0.15.1 and includes a large number of bug fixes along with several new features, enhancements, and performance improvements. A small number of API changes were necessary to fix existing bugs. We recommend that all users upgrade to this version.

API changes

  • Indexing in MultiIndex beyond lex-sort depth is now supported, though a lexically sorted index will have a better performance. (GH2646)

    In [1]: df = pd.DataFrame({'jim':[0, 0, 1, 1],
       ...:                    'joe':['x', 'x', 'z', 'y'],
       ...:                    'jolie':np.random.rand(4)}).set_index(['jim', 'joe'])
       ...: 
    
    In [2]: df
    Out[2]: 
                jolie
    jim joe          
    0   x    0.123943
        x    0.119381
    1   z    0.738523
        y    0.587304
    
    In [3]: df.index.lexsort_depth
    Out[3]: 1
    
    # in prior versions this would raise a KeyError
    # will now show a PerformanceWarning
    In [4]: df.loc[(1, 'z')]
    Out[4]: 
                jolie
    jim joe          
    1   z    0.738523
    
    # lexically sorting
    In [5]: df2 = df.sort_index()
    
    In [6]: df2
    Out[6]: 
                jolie
    jim joe          
    0   x    0.123943
        x    0.119381
    1   y    0.587304
        z    0.738523
    
    In [7]: df2.index.lexsort_depth
    Out[7]: 2
    
    In [8]: df2.loc[(1,'z')]
    Out[8]: 
                jolie
    jim joe          
    1   z    0.738523
    
  • Bug in unique of Series with category dtype, which returned all categories regardless whether they were “used” or not (see GH8559 for the discussion). Previous behaviour was to return all categories:

    In [3]: cat = pd.Categorical(['a', 'b', 'a'], categories=['a', 'b', 'c'])
    
    In [4]: cat
    Out[4]:
    [a, b, a]
    Categories (3, object): [a < b < c]
    
    In [5]: cat.unique()
    Out[5]: array(['a', 'b', 'c'], dtype=object)
    

    Now, only the categories that do effectively occur in the array are returned:

    In [9]: cat = pd.Categorical(['a', 'b', 'a'], categories=['a', 'b', 'c'])
    
    In [10]: cat.unique()
    Out[10]: 
    [a, b]
    Categories (2, object): [a, b]
    
  • Series.all and Series.any now support the level and skipna parameters. Series.all, Series.any, Index.all, and Index.any no longer support the out and keepdims parameters, which existed for compatibility with ndarray. Various index types no longer support the all and any aggregation functions and will now raise TypeError. (GH8302).

  • Allow equality comparisons of Series with a categorical dtype and object dtype; previously these would raise TypeError (GH8938)

  • Bug in NDFrame: conflicting attribute/column names now behave consistently between getting and setting. Previously, when both a column and attribute named y existed, data.y would return the attribute, while data.y = z would update the column (GH8994)

    In [11]: data = pd.DataFrame({'x':[1, 2, 3]})
    
    In [12]: data.y = 2
    
    In [13]: data['y'] = [2, 4, 6]
    
    In [14]: data
    Out[14]: 
       x  y
    0  1  2
    1  2  4
    2  3  6
    
    # this assignment was inconsistent
    In [15]: data.y = 5
    

    Old behavior:

    In [6]: data.y
    Out[6]: 2
    
    In [7]: data['y'].values
    Out[7]: array([5, 5, 5])
    

    New behavior:

    In [16]: data.y
    Out[16]: 5
    
    In [17]: data['y'].values
    Out[17]: array([2, 4, 6])
    
  • Timestamp('now') is now equivalent to Timestamp.now() in that it returns the local time rather than UTC. Also, Timestamp('today') is now equivalent to Timestamp.today() and both have tz as a possible argument. (GH9000)

  • Fix negative step support for label-based slices (GH8753)

    Old behavior:

    In [1]: s = pd.Series(np.arange(3), ['a', 'b', 'c'])
    Out[1]:
    a    0
    b    1
    c    2
    dtype: int64
    
    In [2]: s.loc['c':'a':-1]
    Out[2]:
    c    2
    dtype: int64
    

    New behavior:

    In [18]: s = pd.Series(np.arange(3), ['a', 'b', 'c'])
    
    In [19]: s.loc['c':'a':-1]
    Out[19]: 
    c    2
    b    1
    a    0
    dtype: int64
    

Enhancements

Categorical enhancements:

  • Added ability to export Categorical data to Stata (GH8633). See here for limitations of categorical variables exported to Stata data files.
  • Added flag order_categoricals to StataReader and read_stata to select whether to order imported categorical data (GH8836). See here for more information on importing categorical variables from Stata data files.
  • Added ability to export Categorical data to to/from HDF5 (GH7621). Queries work the same as if it was an object array. However, the category dtyped data is stored in a more efficient manner. See here for an example and caveats w.r.t. prior versions of pandas.
  • Added support for searchsorted() on Categorical class (GH8420).

Other enhancements:

  • Added the ability to specify the SQL type of columns when writing a DataFrame to a database (GH8778). For example, specifying to use the sqlalchemy String type instead of the default Text type for string columns:

    from sqlalchemy.types import String
    data.to_sql('data_dtype', engine, dtype={'Col_1': String})
    
  • Series.all and Series.any now support the level and skipna parameters (GH8302):

    In [20]: s = pd.Series([False, True, False], index=[0, 0, 1])
    
    In [21]: s.any(level=0)
    Out[21]: 
    0     True
    1    False
    dtype: bool
    
  • Panel now supports the all and any aggregation functions. (GH8302):

    In [22]: p = pd.Panel(np.random.rand(2, 5, 4) > 0.1)
    
    In [23]: p.all()
    Out[23]: 
          0     1      2     3
    0  True  True   True  True
    1  True  True  False  True
    2  True  True   True  True
    3  True  True  False  True
    4  True  True   True  True
    
  • Added support for utcfromtimestamp(), fromtimestamp(), and combine() on Timestamp class (GH5351).

  • Added Google Analytics (pandas.io.ga) basic documentation (GH8835). See here.

  • Timedelta arithmetic returns NotImplemented in unknown cases, allowing extensions by custom classes (GH8813).

  • Timedelta now supports arithemtic with numpy.ndarray objects of the appropriate dtype (numpy 1.8 or newer only) (GH8884).

  • Added Timedelta.to_timedelta64() method to the public API (GH8884).

  • Added gbq.generate_bq_schema() function to the gbq module (GH8325).

  • Series now works with map objects the same way as generators (GH8909).

  • Added context manager to HDFStore for automatic closing (GH8791).

  • to_datetime gains an exact keyword to allow for a format to not require an exact match for a provided format string (if its False). exact defaults to True (meaning that exact matching is still the default) (GH8904)

  • Added axvlines boolean option to parallel_coordinates plot function, determines whether vertical lines will be printed, default is True

  • Added ability to read table footers to read_html (GH8552)

  • to_sql now infers datatypes of non-NA values for columns that contain NA values and have dtype object (GH8778).

Performance

  • Reduce memory usage when skiprows is an integer in read_csv (GH8681)
  • Performance boost for to_datetime conversions with a passed format=, and the exact=False (GH8904)

Bug Fixes

  • Bug in concat of Series with category dtype which were coercing to object. (GH8641)
  • Bug in Timestamp-Timestamp not returning a Timedelta type and datelike-datelike ops with timezones (GH8865)
  • Made consistent a timezone mismatch exception (either tz operated with None or incompatible timezone), will now return TypeError rather than ValueError (a couple of edge cases only), (GH8865)
  • Bug in using a pd.Grouper(key=...) with no level/axis or level only (GH8795, GH8866)
  • Report a TypeError when invalid/no parameters are passed in a groupby (GH8015)
  • Bug in packaging pandas with py2app/cx_Freeze (GH8602, GH8831)
  • Bug in groupby signatures that didn’t include *args or **kwargs (GH8733).
  • io.data.Options now raises RemoteDataError when no expiry dates are available from Yahoo and when it receives no data from Yahoo (GH8761), (GH8783).
  • Unclear error message in csv parsing when passing dtype and names and the parsed data is a different data type (GH8833)
  • Bug in slicing a multi-index with an empty list and at least one boolean indexer (GH8781)
  • io.data.Options now raises RemoteDataError when no expiry dates are available from Yahoo (GH8761).
  • Timedelta kwargs may now be numpy ints and floats (GH8757).
  • Fixed several outstanding bugs for Timedelta arithmetic and comparisons (GH8813, GH5963, GH5436).
  • sql_schema now generates dialect appropriate CREATE TABLE statements (GH8697)
  • slice string method now takes step into account (GH8754)
  • Bug in BlockManager where setting values with different type would break block integrity (GH8850)
  • Bug in DatetimeIndex when using time object as key (GH8667)
  • Bug in merge where how='left' and sort=False would not preserve left frame order (GH7331)
  • Bug in MultiIndex.reindex where reindexing at level would not reorder labels (GH4088)
  • Bug in certain operations with dateutil timezones, manifesting with dateutil 2.3 (GH8639)
  • Regression in DatetimeIndex iteration with a Fixed/Local offset timezone (GH8890)
  • Bug in to_datetime when parsing a nanoseconds using the %f format (GH8989)
  • io.data.Options now raises RemoteDataError when no expiry dates are available from Yahoo and when it receives no data from Yahoo (GH8761), (GH8783).
  • Fix: The font size was only set on x axis if vertical or the y axis if horizontal. (GH8765)
  • Fixed division by 0 when reading big csv files in python 3 (GH8621)
  • Bug in outputting a Multindex with to_html,index=False which would add an extra column (GH8452)
  • Imported categorical variables from Stata files retain the ordinal information in the underlying data (GH8836).
  • Defined .size attribute across NDFrame objects to provide compat with numpy >= 1.9.1; buggy with np.array_split (GH8846)
  • Skip testing of histogram plots for matplotlib <= 1.2 (GH8648).
  • Bug where get_data_google returned object dtypes (GH3995)
  • Bug in DataFrame.stack(..., dropna=False) when the DataFrame’s columns is a MultiIndex whose labels do not reference all its levels. (GH8844)
  • Bug in that Option context applied on __enter__ (GH8514)
  • Bug in resample that causes a ValueError when resampling across multiple days and the last offset is not calculated from the start of the range (GH8683)
  • Bug where DataFrame.plot(kind='scatter') fails when checking if an np.array is in the DataFrame (GH8852)
  • Bug in pd.infer_freq/DataFrame.inferred_freq that prevented proper sub-daily frequency inference when the index contained DST days (GH8772).
  • Bug where index name was still used when plotting a series with use_index=False (GH8558).
  • Bugs when trying to stack multiple columns, when some (or all) of the level names are numbers (GH8584).
  • Bug in MultiIndex where __contains__ returns wrong result if index is not lexically sorted or unique (GH7724)
  • BUG CSV: fix problem with trailing whitespace in skipped rows, (GH8679), (GH8661), (GH8983)
  • Regression in Timestamp does not parse ‘Z’ zone designator for UTC (GH8771)
  • Bug in StataWriter the produces writes strings with 244 characters irrespective of actual size (GH8969)
  • Fixed ValueError raised by cummin/cummax when datetime64 Series contains NaT. (GH8965)
  • Bug in Datareader returns object dtype if there are missing values (GH8980)
  • Bug in plotting if sharex was enabled and index was a timeseries, would show labels on multiple axes (GH3964).
  • Bug where passing a unit to the TimedeltaIndex constructor applied the to nano-second conversion twice. (GH9011).
  • Bug in plotting of a period-like array (GH9012)

v0.15.1 (November 9, 2014)

This is a minor bug-fix release from 0.15.0 and includes a small number of API changes, several new features, enhancements, and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this version.

API changes

  • s.dt.hour and other .dt accessors will now return np.nan for missing values (rather than previously -1), (GH8689)

    In [1]: s = Series(date_range('20130101',periods=5,freq='D'))
    
    In [2]: s.iloc[2] = np.nan
    
    In [3]: s
    Out[3]: 
    0   2013-01-01
    1   2013-01-02
    2          NaT
    3   2013-01-04
    4   2013-01-05
    dtype: datetime64[ns]
    

    previous behavior:

    In [6]: s.dt.hour
    Out[6]:
    0    0
    1    0
    2   -1
    3    0
    4    0
    dtype: int64
    

    current behavior:

    In [4]: s.dt.hour
    Out[4]: 
    0    0.0
    1    0.0
    2    NaN
    3    0.0
    4    0.0
    dtype: float64
    
  • groupby with as_index=False will not add erroneous extra columns to result (GH8582):

    In [5]: np.random.seed(2718281)
    
    In [6]: df = pd.DataFrame(np.random.randint(0, 100, (10, 2)),
       ...:                   columns=['jim', 'joe'])
       ...: 
    
    In [7]: df.head()
    Out[7]: 
       jim  joe
    0   61   81
    1   96   49
    2   55   65
    3   72   51
    4   77   12
    
    In [8]: ts = pd.Series(5 * np.random.randint(0, 3, 10))
    

    previous behavior:

    In [4]: df.groupby(ts, as_index=False).max()
    Out[4]:
       NaN  jim  joe
    0    0   72   83
    1    5   77   84
    2   10   96   65
    

    current behavior:

    In [9]: df.groupby(ts, as_index=False).max()
    Out[9]: 
       jim  joe
    0   72   83
    1   77   84
    2   96   65
    
  • groupby will not erroneously exclude columns if the column name conflics with the grouper name (GH8112):

    In [10]: df = pd.DataFrame({'jim': range(5), 'joe': range(5, 10)})
    
    In [11]: df
    Out[11]: 
       jim  joe
    0    0    5
    1    1    6
    2    2    7
    3    3    8
    4    4    9
    
    In [12]: gr = df.groupby(df['jim'] < 2)
    

    previous behavior (excludes 1st column from output):

    In [4]: gr.apply(sum)
    Out[4]:
           joe
    jim
    False   24
    True    11
    

    current behavior:

    In [13]: gr.apply(sum)
    Out[13]: 
           jim  joe
    jim            
    False    9   24
    True     1   11
    
  • Support for slicing with monotonic decreasing indexes, even if start or stop is not found in the index (GH7860):

    In [14]: s = pd.Series(['a', 'b', 'c', 'd'], [4, 3, 2, 1])
    
    In [15]: s
    Out[15]: 
    4    a
    3    b
    2    c
    1    d
    dtype: object
    

    previous behavior:

    In [8]: s.loc[3.5:1.5]
    KeyError: 3.5
    

    current behavior:

    In [16]: s.loc[3.5:1.5]
    Out[16]: 
    3    b
    2    c
    dtype: object
    
  • io.data.Options has been fixed for a change in the format of the Yahoo Options page (GH8612), (GH8741)

    Note

    As a result of a change in Yahoo’s option page layout, when an expiry date is given, Options methods now return data for a single expiry date. Previously, methods returned all data for the selected month.

    The month and year parameters have been undeprecated and can be used to get all options data for a given month.

    If an expiry date that is not valid is given, data for the next expiry after the given date is returned.

    Option data frames are now saved on the instance as callsYYMMDD or putsYYMMDD. Previously they were saved as callsMMYY and putsMMYY. The next expiry is saved as calls and puts.

    New features:

    • The expiry parameter can now be a single date or a list-like object containing dates.
    • A new property expiry_dates was added, which returns all available expiry dates.

    Current behavior:

    In [17]: from pandas.io.data import Options
    
    In [18]: aapl = Options('aapl','yahoo')
    
    In [19]: aapl.get_call_data().iloc[0:5,0:1]
    Out[19]:
                                                 Last
    Strike Expiry     Type Symbol
    80     2014-11-14 call AAPL141114C00080000  29.05
    84     2014-11-14 call AAPL141114C00084000  24.80
    85     2014-11-14 call AAPL141114C00085000  24.05
    86     2014-11-14 call AAPL141114C00086000  22.76
    87     2014-11-14 call AAPL141114C00087000  21.74
    
    In [20]: aapl.expiry_dates
    Out[20]:
    [datetime.date(2014, 11, 14),
     datetime.date(2014, 11, 22),
     datetime.date(2014, 11, 28),
     datetime.date(2014, 12, 5),
     datetime.date(2014, 12, 12),
     datetime.date(2014, 12, 20),
     datetime.date(2015, 1, 17),
     datetime.date(2015, 2, 20),
     datetime.date(2015, 4, 17),
     datetime.date(2015, 7, 17),
     datetime.date(2016, 1, 15),
     datetime.date(2017, 1, 20)]
    
    In [21]: aapl.get_near_stock_price(expiry=aapl.expiry_dates[0:3]).iloc[0:5,0:1]
    Out[21]:
                                                Last
    Strike Expiry     Type Symbol
    109    2014-11-22 call AAPL141122C00109000  1.48
           2014-11-28 call AAPL141128C00109000  1.79
    110    2014-11-14 call AAPL141114C00110000  0.55
           2014-11-22 call AAPL141122C00110000  1.02
           2014-11-28 call AAPL141128C00110000  1.32
    
  • pandas now also registers the datetime64 dtype in matplotlib’s units registry to plot such values as datetimes. This is activated once pandas is imported. In previous versions, plotting an array of datetime64 values will have resulted in plotted integer values. To keep the previous behaviour, you can do del matplotlib.units.registry[np.datetime64] (GH8614).

Enhancements

  • concat permits a wider variety of iterables of pandas objects to be passed as the first parameter (GH8645):

    In [17]: from collections import deque
    
    In [18]: df1 = pd.DataFrame([1, 2, 3])
    
    In [19]: df2 = pd.DataFrame([4, 5, 6])
    

    previous behavior:

    In [7]: pd.concat(deque((df1, df2)))
    TypeError: first argument must be a list-like of pandas objects, you passed an object of type "deque"
    

    current behavior:

    In [20]: pd.concat(deque((df1, df2)))
    Out[20]: 
       0
    0  1
    1  2
    2  3
    0  4
    1  5
    2  6
    
  • Represent MultiIndex labels with a dtype that utilizes memory based on the level size. In prior versions, the memory usage was a constant 8 bytes per element in each level. In addition, in prior versions, the reported memory usage was incorrect as it didn’t show the usage for the memory occupied by the underling data array. (GH8456)

    In [21]: dfi = DataFrame(1,index=pd.MultiIndex.from_product([['a'],range(1000)]),columns=['A'])
    

    previous behavior:

    # this was underreported in prior versions
    In [1]: dfi.memory_usage(index=True)
    Out[1]:
    Index    8000 # took about 24008 bytes in < 0.15.1
    A        8000
    dtype: int64
    

    current behavior:

    In [22]: dfi.memory_usage(index=True)
    Out[22]: 
    Index    52080
    A         8000
    dtype: int64
    
  • Added Index properties is_monotonic_increasing and is_monotonic_decreasing (GH8680).

  • Added option to select columns when importing Stata files (GH7935)

  • Qualify memory usage in DataFrame.info() by adding + if it is a lower bound (GH8578)

  • Raise errors in certain aggregation cases where an argument such as numeric_only is not handled (GH8592).

  • Added support for 3-character ISO and non-standard country codes in io.wb.download() (GH8482)

  • World Bank data requests now will warn/raise based on an errors argument, as well as a list of hard-coded country codes and the World Bank’s JSON response. In prior versions, the error messages didn’t look at the World Bank’s JSON response. Problem-inducing input were simply dropped prior to the request. The issue was that many good countries were cropped in the hard-coded approach. All countries will work now, but some bad countries will raise exceptions because some edge cases break the entire response. (GH8482)

  • Added option to Series.str.split() to return a DataFrame rather than a Series (GH8428)

  • Added option to df.info(null_counts=None|True|False) to override the default display options and force showing of the null-counts (GH8701)

Bug Fixes

  • Bug in unpickling of a CustomBusinessDay object (GH8591)
  • Bug in coercing Categorical to a records array, e.g. df.to_records() (GH8626)
  • Bug in Categorical not created properly with Series.to_frame() (GH8626)
  • Bug in coercing in astype of a Categorical of a passed pd.Categorical (this now raises TypeError correctly), (GH8626)
  • Bug in cut/qcut when using Series and retbins=True (GH8589)
  • Bug in writing Categorical columns to an SQL database with to_sql (GH8624).
  • Bug in comparing Categorical of datetime raising when being compared to a scalar datetime (GH8687)
  • Bug in selecting from a Categorical with .iloc (GH8623)
  • Bug in groupby-transform with a Categorical (GH8623)
  • Bug in duplicated/drop_duplicates with a Categorical (GH8623)
  • Bug in Categorical reflected comparison operator raising if the first argument was a numpy array scalar (e.g. np.int64) (GH8658)
  • Bug in Panel indexing with a list-like (GH8710)
  • Compat issue is DataFrame.dtypes when options.mode.use_inf_as_null is True (GH8722)
  • Bug in read_csv, dialect parameter would not take a string (GH8703)
  • Bug in slicing a multi-index level with an empty-list (GH8737)
  • Bug in numeric index operations of add/sub with Float/Index Index with numpy arrays (GH8608)
  • Bug in setitem with empty indexer and unwanted coercion of dtypes (GH8669)
  • Bug in ix/loc block splitting on setitem (manifests with integer-like dtypes, e.g. datetime64) (GH8607)
  • Bug when doing label based indexing with integers not found in the index for non-unique but monotonic indexes (GH8680).
  • Bug when indexing a Float64Index with np.nan on numpy 1.7 (GH8980).
  • Fix shape attribute for MultiIndex (GH8609)
  • Bug in GroupBy where a name conflict between the grouper and columns would break groupby operations (GH7115, GH8112)
  • Fixed a bug where plotting a column y and specifying a label would mutate the index name of the original DataFrame (GH8494)
  • Fix regression in plotting of a DatetimeIndex directly with matplotlib (GH8614).
  • Bug in date_range where partially-specified dates would incorporate current date (GH6961)
  • Bug in Setting by indexer to a scalar value with a mixed-dtype Panel4d was failing (GH8702)
  • Bug where DataReader’s would fail if one of the symbols passed was invalid. Now returns data for valid symbols and np.nan for invalid (GH8494)
  • Bug in get_quote_yahoo that wouldn’t allow non-float return values (GH5229).

v0.15.0 (October 18, 2014)

This is a major release from 0.14.1 and includes a small number of API changes, several new features, enhancements, and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this version.

Warning

pandas >= 0.15.0 will no longer support compatibility with NumPy versions < 1.7.0. If you want to use the latest versions of pandas, please upgrade to NumPy >= 1.7.0 (GH7711)

Warning

In 0.15.0 Index has internally been refactored to no longer sub-class ndarray but instead subclass PandasObject, similarly to the rest of the pandas objects. This change allows very easy sub-classing and creation of new index types. This should be a transparent change with only very limited API implications (See the Internal Refactoring)

Warning

The refactorings in Categorical changed the two argument constructor from “codes/labels and levels” to “values and levels (now called ‘categories’)”. This can lead to subtle bugs. If you use Categorical directly, please audit your code before updating to this pandas version and change it to use the from_codes() constructor. See more on Categorical here

New features

Categoricals in Series/DataFrame

Categorical can now be included in Series and DataFrames and gained new methods to manipulate. Thanks to Jan Schulz for much of this API/implementation. (GH3943, GH5313, GH5314, GH7444, GH7839, GH7848, GH7864, GH7914, GH7768, GH8006, GH3678, GH8075, GH8076, GH8143, GH8453, GH8518).

For full docs, see the categorical introduction and the API documentation.

In [1]: df = DataFrame({"id":[1,2,3,4,5,6], "raw_grade":['a', 'b', 'b', 'a', 'a', 'e']})

In [2]: df["grade"] = df["raw_grade"].astype("category")

In [3]: df["grade"]
Out[3]: 
0    a
1    b
2    b
3    a
4    a
5    e
Name: grade, dtype: category
Categories (3, object): [a, b, e]

# Rename the categories
In [4]: df["grade"].cat.categories = ["very good", "good", "very bad"]

# Reorder the categories and simultaneously add the missing categories
In [5]: df["grade"] = df["grade"].cat.set_categories(["very bad", "bad", "medium", "good", "very good"])

In [6]: df["grade"]
Out[6]: 
0    very good
1         good
2         good
3    very good
4    very good
5     very bad
Name: grade, dtype: category
Categories (5, object): [very bad, bad, medium, good, very good]

In [7]: df.sort_values("grade")
Out[7]: 
   id raw_grade      grade
5   6         e   very bad
1   2         b       good
2   3         b       good
0   1         a  very good
3   4         a  very good
4   5         a  very good

In [8]: df.groupby("grade").size()
Out[8]: 
grade
very bad     1
bad          0
medium       0
good         2
very good    3
dtype: int64
  • pandas.core.group_agg and pandas.core.factor_agg were removed. As an alternative, construct a dataframe and use df.groupby(<group>).agg(<func>).
  • Supplying “codes/labels and levels” to the Categorical constructor is not supported anymore. Supplying two arguments to the constructor is now interpreted as “values and levels (now called ‘categories’)”. Please change your code to use the from_codes() constructor.
  • The Categorical.labels attribute was renamed to Categorical.codes and is read only. If you want to manipulate codes, please use one of the API methods on Categoricals.
  • The Categorical.levels attribute is renamed to Categorical.categories.

TimedeltaIndex/Scalar

We introduce a new scalar type Timedelta, which is a subclass of datetime.timedelta, and behaves in a similar manner, but allows compatibility with np.timedelta64 types as well as a host of custom representation, parsing, and attributes. This type is very similar to how Timestamp works for datetimes. It is a nice-API box for the type. See the docs. (GH3009, GH4533, GH8209, GH8187, GH8190, GH7869, GH7661, GH8345, GH8471)

Warning

Timedelta scalars (and TimedeltaIndex) component fields are not the same as the component fields on a datetime.timedelta object. For example, .seconds on a datetime.timedelta object returns the total number of seconds combined between hours, minutes and seconds. In contrast, the pandas Timedelta breaks out hours, minutes, microseconds and nanoseconds separately.

# Timedelta accessor
In [9]: tds = Timedelta('31 days 5 min 3 sec')

In [10]: tds.minutes
Out[10]: 5L

In [11]: tds.seconds
Out[11]: 3L

# datetime.timedelta accessor
# this is 5 minutes * 60 + 3 seconds
In [12]: tds.to_pytimedelta().seconds
Out[12]: 303

Note: this is no longer true starting from v0.16.0, where full compatibility with datetime.timedelta is introduced. See the 0.16.0 whatsnew entry

Warning

Prior to 0.15.0 pd.to_timedelta would return a Series for list-like/Series input, and a np.timedelta64 for scalar input. It will now return a TimedeltaIndex for list-like input, Series for Series input, and Timedelta for scalar input.

The arguments to pd.to_timedelta are now (arg,unit='ns',box=True,coerce=False), previously were (arg,box=True,unit='ns') as these are more logical.

Consruct a scalar

In [9]: Timedelta('1 days 06:05:01.00003')
Out[9]: Timedelta('1 days 06:05:01.000030')

In [10]: Timedelta('15.5us')
Out[10]: Timedelta('0 days 00:00:00.000015')

In [11]: Timedelta('1 hour 15.5us')
Out[11]: Timedelta('0 days 01:00:00.000015')

# negative Timedeltas have this string repr
# to be more consistent with datetime.timedelta conventions
In [12]: Timedelta('-1us')
Out[12]: Timedelta('-1 days +23:59:59.999999')

# a NaT
In [13]: Timedelta('nan')
Out[13]: NaT

Access fields for a Timedelta

In [14]: td = Timedelta('1 hour 3m 15.5us')

In [15]: td.seconds
Out[15]: 3780

In [16]: td.microseconds
Out[16]: 16

In [17]: td.nanoseconds
Out[17]: 500

Construct a TimedeltaIndex

In [18]: TimedeltaIndex(['1 days','1 days, 00:00:05',
   ....:                 np.timedelta64(2,'D'),timedelta(days=2,seconds=2)])
   ....: 
Out[18]: 
TimedeltaIndex(['1 days 00:00:00', '1 days 00:00:05', '2 days 00:00:00',
                '2 days 00:00:02'],
               dtype='timedelta64[ns]', freq=None)

Constructing a TimedeltaIndex with a regular range

In [19]: timedelta_range('1 days',periods=5,freq='D')
Out[19]: TimedeltaIndex(['1 days', '2 days', '3 days', '4 days', '5 days'], dtype='timedelta64[ns]', freq='D')

In [20]: timedelta_range(start='1 days',end='2 days',freq='30T')
Out[20]: 
TimedeltaIndex(['1 days 00:00:00', '1 days 00:30:00', '1 days 01:00:00',
                '1 days 01:30:00', '1 days 02:00:00', '1 days 02:30:00',
                '1 days 03:00:00', '1 days 03:30:00', '1 days 04:00:00',
                '1 days 04:30:00', '1 days 05:00:00', '1 days 05:30:00',
                '1 days 06:00:00', '1 days 06:30:00', '1 days 07:00:00',
                '1 days 07:30:00', '1 days 08:00:00', '1 days 08:30:00',
                '1 days 09:00:00', '1 days 09:30:00', '1 days 10:00:00',
                '1 days 10:30:00', '1 days 11:00:00', '1 days 11:30:00',
                '1 days 12:00:00', '1 days 12:30:00', '1 days 13:00:00',
                '1 days 13:30:00', '1 days 14:00:00', '1 days 14:30:00',
                '1 days 15:00:00', '1 days 15:30:00', '1 days 16:00:00',
                '1 days 16:30:00', '1 days 17:00:00', '1 days 17:30:00',
                '1 days 18:00:00', '1 days 18:30:00', '1 days 19:00:00',
                '1 days 19:30:00', '1 days 20:00:00', '1 days 20:30:00',
                '1 days 21:00:00', '1 days 21:30:00', '1 days 22:00:00',
                '1 days 22:30:00', '1 days 23:00:00', '1 days 23:30:00',
                '2 days 00:00:00'],
               dtype='timedelta64[ns]', freq='30T')

You can now use a TimedeltaIndex as the index of a pandas object

In [21]: s = Series(np.arange(5),
   ....:            index=timedelta_range('1 days',periods=5,freq='s'))
   ....: 

In [22]: s
Out[22]: 
1 days 00:00:00    0
1 days 00:00:01    1
1 days 00:00:02    2
1 days 00:00:03    3
1 days 00:00:04    4
Freq: S, dtype: int64

You can select with partial string selections

In [23]: s['1 day 00:00:02']
Out[23]: 2

In [24]: s['1 day':'1 day 00:00:02']
Out[24]: 
1 days 00:00:00    0
1 days 00:00:01    1
1 days 00:00:02    2
Freq: S, dtype: int64

Finally, the combination of TimedeltaIndex with DatetimeIndex allow certain combination operations that are NaT preserving:

In [25]: tdi = TimedeltaIndex(['1 days',pd.NaT,'2 days'])

In [26]: tdi.tolist()
Out[26]: [Timedelta('1 days 00:00:00'), NaT, Timedelta('2 days 00:00:00')]

In [27]: dti = date_range('20130101',periods=3)

In [28]: dti.tolist()
Out[28]: 
[Timestamp('2013-01-01 00:00:00', freq='D'),
 Timestamp('2013-01-02 00:00:00', freq='D'),
 Timestamp('2013-01-03 00:00:00', freq='D')]

In [29]: (dti + tdi).tolist()
Out[29]: [Timestamp('2013-01-02 00:00:00'), NaT, Timestamp('2013-01-05 00:00:00')]

In [30]: (dti - tdi).tolist()
Out[30]: [Timestamp('2012-12-31 00:00:00'), NaT, Timestamp('2013-01-01 00:00:00')]
  • iteration of a Series e.g. list(Series(...)) of timedelta64[ns] would prior to v0.15.0 return np.timedelta64 for each element. These will now be wrapped in Timedelta.

Memory Usage

Implemented methods to find memory usage of a DataFrame. See the FAQ for more. (GH6852).

A new display option display.memory_usage (see Options and Settings) sets the default behavior of the memory_usage argument in the df.info() method. By default display.memory_usage is True.

In [31]: dtypes = ['int64', 'float64', 'datetime64[ns]', 'timedelta64[ns]',
   ....:           'complex128', 'object', 'bool']
   ....: 

In [32]: n = 5000

In [33]: data = dict([ (t, np.random.randint(100, size=n).astype(t))
   ....:                 for t in dtypes])
   ....: 

In [34]: df = DataFrame(data)

In [35]: df['categorical'] = df['object'].astype('category')

In [36]: df.info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 5000 entries, 0 to 4999
Data columns (total 8 columns):
int64              5000 non-null int64
float64            5000 non-null float64
datetime64[ns]     5000 non-null datetime64[ns]
timedelta64[ns]    5000 non-null timedelta64[ns]
complex128         5000 non-null complex128
object             5000 non-null object
bool               5000 non-null bool
categorical        5000 non-null category
dtypes: bool(1), category(1), complex128(1), datetime64[ns](1), float64(1), int64(1), object(1), timedelta64[ns](1)
memory usage: 289.1+ KB

Additionally memory_usage() is an available method for a dataframe object which returns the memory usage of each column.

In [37]: df.memory_usage(index=True)
Out[37]: 
Index                 80
int64              40000
float64            40000
datetime64[ns]     40000
timedelta64[ns]    40000
complex128         80000
object             40000
bool                5000
categorical        10920
dtype: int64

.dt accessor

Series has gained an accessor to succinctly return datetime like properties for the values of the Series, if its a datetime/period like Series. (GH7207) This will return a Series, indexed like the existing Series. See the docs

# datetime
In [38]: s = Series(date_range('20130101 09:10:12',periods=4))

In [39]: s
Out[39]: 
0   2013-01-01 09:10:12
1   2013-01-02 09:10:12
2   2013-01-03 09:10:12
3   2013-01-04 09:10:12
dtype: datetime64[ns]

In [40]: s.dt.hour
Out[40]: 
0    9
1    9
2    9
3    9
dtype: int64

In [41]: s.dt.second
Out[41]: 
0    12
1    12
2    12
3    12
dtype: int64

In [42]: s.dt.day
Out[42]: 
0    1
1    2
2    3
3    4
dtype: int64

In [43]: s.dt.freq
Out[43]: 'D'

This enables nice expressions like this:

In [44]: s[s.dt.day==2]
Out[44]: 
1   2013-01-02 09:10:12
dtype: datetime64[ns]

You can easily produce tz aware transformations:

In [45]: stz = s.dt.tz_localize('US/Eastern')

In [46]: stz
Out[46]: 
0   2013-01-01 09:10:12-05:00
1   2013-01-02 09:10:12-05:00
2   2013-01-03 09:10:12-05:00
3   2013-01-04 09:10:12-05:00
dtype: datetime64[ns, US/Eastern]

In [47]: stz.dt.tz
Out[47]: <DstTzInfo 'US/Eastern' LMT-1 day, 19:04:00 STD>

You can also chain these types of operations:

In [48]: s.dt.tz_localize('UTC').dt.tz_convert('US/Eastern')
Out[48]: 
0   2013-01-01 04:10:12-05:00
1   2013-01-02 04:10:12-05:00
2   2013-01-03 04:10:12-05:00
3   2013-01-04 04:10:12-05:00
dtype: datetime64[ns, US/Eastern]

The .dt accessor works for period and timedelta dtypes.

# period
In [49]: s = Series(period_range('20130101',periods=4,freq='D'))

In [50]: s
Out[50]: 
0   2013-01-01
1   2013-01-02
2   2013-01-03
3   2013-01-04
dtype: object

In [51]: s.dt.year
Out[51]: 
0    2013
1    2013
2    2013
3    2013
dtype: int64

In [52]: s.dt.day
Out[52]: 
0    1
1    2
2    3
3    4
dtype: int64
# timedelta
In [53]: s = Series(timedelta_range('1 day 00:00:05',periods=4,freq='s'))

In [54]: s
Out[54]: 
0   1 days 00:00:05
1   1 days 00:00:06
2   1 days 00:00:07
3   1 days 00:00:08
dtype: timedelta64[ns]

In [55]: s.dt.days
Out[55]: 
0    1
1    1
2    1
3    1
dtype: int64

In [56]: s.dt.seconds
Out[56]: 
0    5
1    6
2    7
3    8
dtype: int64

In [57]: s.dt.components
Out[57]: 
   days  hours  minutes  seconds  milliseconds  microseconds  nanoseconds
0     1      0        0        5             0             0            0
1     1      0        0        6             0             0            0
2     1      0        0        7             0             0            0
3     1      0        0        8             0             0            0

Timezone handling improvements

  • tz_localize(None) for tz-aware Timestamp and DatetimeIndex now removes timezone holding local time, previously this resulted in Exception or TypeError (GH7812)

    In [58]: ts = Timestamp('2014-08-01 09:00', tz='US/Eastern')
    
    In [59]: ts
    Out[59]: Timestamp('2014-08-01 09:00:00-0400', tz='US/Eastern')
    
    In [60]: ts.tz_localize(None)
    Out[60]: Timestamp('2014-08-01 09:00:00')
    
    In [61]: didx = DatetimeIndex(start='2014-08-01 09:00', freq='H', periods=10, tz='US/Eastern')
    
    In [62]: didx
    Out[62]: 
    DatetimeIndex(['2014-08-01 09:00:00-04:00', '2014-08-01 10:00:00-04:00',
                   '2014-08-01 11:00:00-04:00', '2014-08-01 12:00:00-04:00',
                   '2014-08-01 13:00:00-04:00', '2014-08-01 14:00:00-04:00',
                   '2014-08-01 15:00:00-04:00', '2014-08-01 16:00:00-04:00',
                   '2014-08-01 17:00:00-04:00', '2014-08-01 18:00:00-04:00'],
                  dtype='datetime64[ns, US/Eastern]', freq='H')
    
    In [63]: didx.tz_localize(None)
    Out[63]: 
    DatetimeIndex(['2014-08-01 09:00:00', '2014-08-01 10:00:00',
                   '2014-08-01 11:00:00', '2014-08-01 12:00:00',
                   '2014-08-01 13:00:00', '2014-08-01 14:00:00',
                   '2014-08-01 15:00:00', '2014-08-01 16:00:00',
                   '2014-08-01 17:00:00', '2014-08-01 18:00:00'],
                  dtype='datetime64[ns]', freq='H')
    
  • tz_localize now accepts the ambiguous keyword which allows for passing an array of bools indicating whether the date belongs in DST or not, ‘NaT’ for setting transition times to NaT, ‘infer’ for inferring DST/non-DST, and ‘raise’ (default) for an AmbiguousTimeError to be raised. See the docs for more details (GH7943)

  • DataFrame.tz_localize and DataFrame.tz_convert now accepts an optional level argument for localizing a specific level of a MultiIndex (GH7846)

  • Timestamp.tz_localize and Timestamp.tz_convert now raise TypeError in error cases, rather than Exception (GH8025)

  • a timeseries/index localized to UTC when inserted into a Series/DataFrame will preserve the UTC timezone (rather than being a naive datetime64[ns]) as object dtype (GH8411)

  • Timestamp.__repr__ displays dateutil.tz.tzoffset info (GH7907)

Rolling/Expanding Moments improvements

  • rolling_min(), rolling_max(), rolling_cov(), and rolling_corr() now return objects with all NaN when len(arg) < min_periods <= window rather than raising. (This makes all rolling functions consistent in this behavior). (GH7766)

    Prior to 0.15.0

    In [64]: s = Series([10, 11, 12, 13])
    
    In [15]: rolling_min(s, window=10, min_periods=5)
    ValueError: min_periods (5) must be <= window (4)
    

    New behavior

    In [4]: pd.rolling_min(s, window=10, min_periods=5)
    Out[4]:
    0   NaN
    1   NaN
    2   NaN
    3   NaN
    dtype: float64
    
  • rolling_max(), rolling_min(), rolling_sum(), rolling_mean(), rolling_median(), rolling_std(), rolling_var(), rolling_skew(), rolling_kurt(), rolling_quantile(), rolling_cov(), rolling_corr(), rolling_corr_pairwise(), rolling_window(), and rolling_apply() with center=True previously would return a result of the same structure as the input arg with NaN in the final (window-1)/2 entries.

    Now the final (window-1)/2 entries of the result are calculated as if the input arg were followed by (window-1)/2 NaN values (or with shrinking windows, in the case of rolling_apply()). (GH7925, GH8269)

    Prior behavior (note final value is NaN):

    In [7]: rolling_sum(Series(range(4)), window=3, min_periods=0, center=True)
    Out[7]:
    0     1
    1     3
    2     6
    3   NaN
    dtype: float64
    

    New behavior (note final value is 5 = sum([2, 3, NaN])):

    In [7]: rolling_sum(Series(range(4)), window=3, min_periods=0, center=True)
    Out[7]:
    0    1
    1    3
    2    6
    3    5
    dtype: float64
    
  • rolling_window() now normalizes the weights properly in rolling mean mode (mean=True) so that the calculated weighted means (e.g. ‘triang’, ‘gaussian’) are distributed about the same means as those calculated without weighting (i.e. ‘boxcar’). See the note on normalization for further details. (GH7618)

    In [65]: s = Series([10.5, 8.8, 11.4, 9.7, 9.3])
    

    Behavior prior to 0.15.0:

    In [39]: rolling_window(s, window=3, win_type='triang', center=True)
    Out[39]:
    0         NaN
    1    6.583333
    2    6.883333
    3    6.683333
    4         NaN
    dtype: float64
    

    New behavior

    In [10]: pd.rolling_window(s, window=3, win_type='triang', center=True)
    Out[10]:
    0       NaN
    1     9.875
    2    10.325
    3    10.025
    4       NaN
    dtype: float64
    
  • Removed center argument from all expanding_ functions (see list), as the results produced when center=True did not make much sense. (GH7925)

  • Added optional ddof argument to expanding_cov() and rolling_cov(). The default value of 1 is backwards-compatible. (GH8279)

  • Documented the ddof argument to expanding_var(), expanding_std(), rolling_var(), and rolling_std(). These functions’ support of a ddof argument (with a default value of 1) was previously undocumented. (GH8064)

  • ewma(), ewmstd(), ewmvol(), ewmvar(), ewmcov(), and ewmcorr() now interpret min_periods in the same manner that the rolling_*() and expanding_*() functions do: a given result entry will be NaN if the (expanding, in this case) window does not contain at least min_periods values. The previous behavior was to set to NaN the min_periods entries starting with the first non- NaN value. (GH7977)

    Prior behavior (note values start at index 2, which is min_periods after index 0 (the index of the first non-empty value)):

    In [66]: s  = Series([1, None, None, None, 2, 3])
    
    In [51]: ewma(s, com=3., min_periods=2)
    Out[51]:
    0         NaN
    1         NaN
    2    1.000000
    3    1.000000
    4    1.571429
    5    2.189189
    dtype: float64
    

    New behavior (note values start at index 4, the location of the 2nd (since min_periods=2) non-empty value):

    In [2]: pd.ewma(s, com=3., min_periods=2)
    Out[2]:
    0         NaN
    1         NaN
    2         NaN
    3         NaN
    4    1.759644
    5    2.383784
    dtype: float64
    
  • ewmstd(), ewmvol(), ewmvar(), ewmcov(), and ewmcorr() now have an optional adjust argument, just like ewma() does, affecting how the weights are calculated. The default value of adjust is True, which is backwards-compatible. See Exponentially weighted moment functions for details. (GH7911)

  • ewma(), ewmstd(), ewmvol(), ewmvar(), ewmcov(), and ewmcorr() now have an optional ignore_na argument. When ignore_na=False (the default), missing values are taken into account in the weights calculation. When ignore_na=True (which reproduces the pre-0.15.0 behavior), missing values are ignored in the weights calculation. (GH7543)

    In [7]: pd.ewma(Series([None, 1., 8.]), com=2.)
    Out[7]:
    0    NaN
    1    1.0
    2    5.2
    dtype: float64
    
    In [8]: pd.ewma(Series([1., None, 8.]), com=2., ignore_na=True)  # pre-0.15.0 behavior
    Out[8]:
    0    1.0
    1    1.0
    2    5.2
    dtype: float64
    
    In [9]: pd.ewma(Series([1., None, 8.]), com=2., ignore_na=False)  # new default
    Out[9]:
    0    1.000000
    1    1.000000
    2    5.846154
    dtype: float64
    

    Warning

    By default (ignore_na=False) the ewm*() functions’ weights calculation in the presence of missing values is different than in pre-0.15.0 versions. To reproduce the pre-0.15.0 calculation of weights in the presence of missing values one must specify explicitly ignore_na=True.

  • Bug in expanding_cov(), expanding_corr(), rolling_cov(), rolling_cor(), ewmcov(), and ewmcorr() returning results with columns sorted by name and producing an error for non-unique columns; now handles non-unique columns and returns columns in original order (except for the case of two DataFrames with pairwise=False, where behavior is unchanged) (GH7542)

  • Bug in rolling_count() and expanding_*() functions unnecessarily producing error message for zero-length data (GH8056)

  • Bug in rolling_apply() and expanding_apply() interpreting min_periods=0 as min_periods=1 (GH8080)

  • Bug in expanding_std() and expanding_var() for a single value producing a confusing error message (GH7900)

  • Bug in rolling_std() and rolling_var() for a single value producing 0 rather than NaN (GH7900)

  • Bug in ewmstd(), ewmvol(), ewmvar(), and ewmcov() calculation of de-biasing factors when bias=False (the default). Previously an incorrect constant factor was used, based on adjust=True, ignore_na=True, and an infinite number of observations. Now a different factor is used for each entry, based on the actual weights (analogous to the usual N/(N-1) factor). In particular, for a single point a value of NaN is returned when bias=False, whereas previously a value of (approximately) 0 was returned.

    For example, consider the following pre-0.15.0 results for ewmvar(..., bias=False), and the corresponding debiasing factors:

    In [67]: s = Series([1., 2., 0., 4.])
    
    In [89]: ewmvar(s, com=2., bias=False)
    Out[89]:
    0   -2.775558e-16
    1    3.000000e-01
    2    9.556787e-01
    3    3.585799e+00
    dtype: float64
    
    In [90]: ewmvar(s, com=2., bias=False) / ewmvar(s, com=2., bias=True)
    Out[90]:
    0    1.25
    1    1.25
    2    1.25
    3    1.25
    dtype: float64
    

    Note that entry 0 is approximately 0, and the debiasing factors are a constant 1.25. By comparison, the following 0.15.0 results have a NaN for entry 0, and the debiasing factors are decreasing (towards 1.25):

    In [14]: pd.ewmvar(s, com=2., bias=False)
    Out[14]:
    0         NaN
    1    0.500000
    2    1.210526
    3    4.089069
    dtype: float64
    
    In [15]: pd.ewmvar(s, com=2., bias=False) / pd.ewmvar(s, com=2., bias=True)
    Out[15]:
    0         NaN
    1    2.083333
    2    1.583333
    3    1.425439
    dtype: float64
    

    See Exponentially weighted moment functions for details. (GH7912)

Improvements in the sql io module

  • Added support for a chunksize parameter to to_sql function. This allows DataFrame to be written in chunks and avoid packet-size overflow errors (GH8062).

  • Added support for a chunksize parameter to read_sql function. Specifying this argument will return an iterator through chunks of the query result (GH2908).

  • Added support for writing datetime.date and datetime.time object columns with to_sql (GH6932).

  • Added support for specifying a schema to read from/write to with read_sql_table and to_sql (GH7441, GH7952). For example:

    df.to_sql('table', engine, schema='other_schema')
    pd.read_sql_table('table', engine, schema='other_schema')
    
  • Added support for writing NaN values with to_sql (GH2754).

  • Added support for writing datetime64 columns with to_sql for all database flavors (GH7103).

Backwards incompatible API changes

Breaking changes

API changes related to Categorical (see here for more details):

  • The Categorical constructor with two arguments changed from “codes/labels and levels” to “values and levels (now called ‘categories’)”. This can lead to subtle bugs. If you use Categorical directly, please audit your code by changing it to use the from_codes() constructor.

    An old function call like (prior to 0.15.0):

    pd.Categorical([0,1,0,2,1], levels=['a', 'b', 'c'])
    

    will have to adapted to the following to keep the same behaviour:

    In [2]: pd.Categorical.from_codes([0,1,0,2,1], categories=['a', 'b', 'c'])
    Out[2]:
    [a, b, a, c, b]
    Categories (3, object): [a, b, c]
    

API changes related to the introduction of the Timedelta scalar (see above for more details):

  • Prior to 0.15.0 to_timedelta() would return a Series for list-like/Series input, and a np.timedelta64 for scalar input. It will now return a TimedeltaIndex for list-like input, Series for Series input, and Timedelta for scalar input.

For API changes related to the rolling and expanding functions, see detailed overview above.

Other notable API changes:

  • Consistency when indexing with .loc and a list-like indexer when no values are found.

    In [68]: df = DataFrame([['a'],['b']],index=[1,2])
    
    In [69]: df
    Out[69]: 
       0
    1  a
    2  b
    

    In prior versions there was a difference in these two constructs:

    • df.loc[[3]] would return a frame reindexed by 3 (with all np.nan values)
    • df.loc[[3],:] would raise KeyError.

    Both will now raise a KeyError. The rule is that at least 1 indexer must be found when using a list-like and .loc (GH7999)

    Furthermore in prior versions these were also different:

    • df.loc[[1,3]] would return a frame reindexed by [1,3]
    • df.loc[[1,3],:] would raise KeyError.

    Both will now return a frame reindex by [1,3]. E.g.

    In [3]: df.loc[[1,3]]
    Out[3]:
         0
    1    a
    3  NaN
    
    In [4]: df.loc[[1,3],:]
    Out[4]:
         0
    1    a
    3  NaN
    

    This can also be seen in multi-axis indexing with a Panel.

    In [70]: p = Panel(np.arange(2*3*4).reshape(2,3,4),
       ....:           items=['ItemA','ItemB'],
       ....:           major_axis=[1,2,3],
       ....:           minor_axis=['A','B','C','D'])
       ....: 
    
    In [71]: p
    Out[71]: 
    <class 'pandas.core.panel.Panel'>
    Dimensions: 2 (items) x 3 (major_axis) x 4 (minor_axis)
    Items axis: ItemA to ItemB
    Major_axis axis: 1 to 3
    Minor_axis axis: A to D
    

    The following would raise KeyError prior to 0.15.0:

    In [5]:
    Out[5]:
       ItemA  ItemD
    1      3    NaN
    2      7    NaN
    3     11    NaN
    

    Furthermore, .loc will raise If no values are found in a multi-index with a list-like indexer:

    In [72]: s = Series(np.arange(3,dtype='int64'),
       ....:            index=MultiIndex.from_product([['A'],['foo','bar','baz']],
       ....:                                          names=['one','two'])
       ....:           ).sort_index()
       ....: 
    
    In [73]: s
    Out[73]: 
    one  two
    A    bar    1
         baz    2
         foo    0
    dtype: int64
    
    In [74]: try:
       ....:    s.loc[['D']]
       ....: except KeyError as e:
       ....:    print("KeyError: " + str(e))
       ....: 
    KeyError: "['D'] not in index"
    
  • Assigning values to None now considers the dtype when choosing an ‘empty’ value (GH7941).

    Previously, assigning to None in numeric containers changed the dtype to object (or errored, depending on the call). It now uses NaN:

    In [75]: s = Series([1, 2, 3])
    
    In [76]: s.loc[0] = None
    
    In [77]: s
    Out[77]: 
    0    NaN
    1    2.0
    2    3.0
    dtype: float64
    

    NaT is now used similarly for datetime containers.

    For object containers, we now preserve None values (previously these were converted to NaN values).

    In [78]: s = Series(["a", "b", "c"])
    
    In [79]: s.loc[0] = None
    
    In [80]: s
    Out[80]: 
    0    None
    1       b
    2       c
    dtype: object
    

    To insert a NaN, you must explicitly use np.nan. See the docs.

  • In prior versions, updating a pandas object inplace would not reflect in other python references to this object. (GH8511, GH5104)

    In [81]: s = Series([1, 2, 3])
    
    In [82]: s2 = s
    
    In [83]: s += 1.5
    

    Behavior prior to v0.15.0

    # the original object
    In [5]: s
    Out[5]:
    0    2.5
    1    3.5
    2    4.5
    dtype: float64
    
    
    # a reference to the original object
    In [7]: s2
    Out[7]:
    0    1
    1    2
    2    3
    dtype: int64
    

    This is now the correct behavior

    # the original object
    In [84]: s
    Out[84]: 
    0    2.5
    1    3.5
    2    4.5
    dtype: float64
    
    # a reference to the original object
    In [85]: s2
    Out[85]: 
    0    2.5
    1    3.5
    2    4.5
    dtype: float64
    
  • Made both the C-based and Python engines for read_csv and read_table ignore empty lines in input as well as whitespace-filled lines, as long as sep is not whitespace. This is an API change that can be controlled by the keyword parameter skip_blank_lines. See the docs (GH4466)

  • A timeseries/index localized to UTC when inserted into a Series/DataFrame will preserve the UTC timezone and inserted as object dtype rather than being converted to a naive datetime64[ns] (GH8411).

  • Bug in passing a DatetimeIndex with a timezone that was not being retained in DataFrame construction from a dict (GH7822)

    In prior versions this would drop the timezone, now it retains the timezone, but gives a column of object dtype:

    In [86]: i = date_range('1/1/2011', periods=3, freq='10s', tz = 'US/Eastern')
    
    In [87]: i
    Out[87]: 
    DatetimeIndex(['2011-01-01 00:00:00-05:00', '2011-01-01 00:00:10-05:00',
                   '2011-01-01 00:00:20-05:00'],
                  dtype='datetime64[ns, US/Eastern]', freq='10S')
    
    In [88]: df = DataFrame( {'a' : i } )
    
    In [89]: df
    Out[89]: 
                              a
    0 2011-01-01 00:00:00-05:00
    1 2011-01-01 00:00:10-05:00
    2 2011-01-01 00:00:20-05:00
    
    In [90]: df.dtypes
    Out[90]: 
    a    datetime64[ns, US/Eastern]
    dtype: object
    

    Previously this would have yielded a column of datetime64 dtype, but without timezone info.

    The behaviour of assigning a column to an existing dataframe as df[‘a’] = i remains unchanged (this already returned an object column with a timezone).

  • When passing multiple levels to stack(), it will now raise a ValueError when the levels aren’t all level names or all level numbers (GH7660). See Reshaping by stacking and unstacking.

  • Raise a ValueError in df.to_hdf with ‘fixed’ format, if df has non-unique columns as the resulting file will be broken (GH7761)

  • SettingWithCopy raise/warnings (according to the option mode.chained_assignment) will now be issued when setting a value on a sliced mixed-dtype DataFrame using chained-assignment. (GH7845, GH7950)

    In [1]: df = DataFrame(np.arange(0,9), columns=['count'])
    
    In [2]: df['group'] = 'b'
    
    In [3]: df.iloc[0:5]['group'] = 'a'
    /usr/local/bin/ipython:1: SettingWithCopyWarning:
    A value is trying to be set on a copy of a slice from a DataFrame.
    Try using .loc[row_indexer,col_indexer] = value instead
    
    See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy
    
  • merge, DataFrame.merge, and ordered_merge now return the same type as the left argument (GH7737).

  • Previously an enlargement with a mixed-dtype frame would act unlike .append which will preserve dtypes (related GH2578, GH8176):

    In [91]: df = DataFrame([[True, 1],[False, 2]],
       ....:                columns=["female","fitness"])
       ....: 
    
    In [92]: df
    Out[92]: 
       female  fitness
    0    True        1
    1   False        2
    
    In [93]: df.dtypes
    Out[93]: 
    female      bool
    fitness    int64
    dtype: object
    
    # dtypes are now preserved
    In [94]: df.loc[2] = df.loc[1]
    
    In [95]: df
    Out[95]: 
       female  fitness
    0    True        1
    1   False        2
    2   False        2
    
    In [96]: df.dtypes
    Out[96]: 
    female      bool
    fitness    int64
    dtype: object
    
  • Series.to_csv() now returns a string when path=None, matching the behaviour of DataFrame.to_csv() (GH8215).

  • read_hdf now raises IOError when a file that doesn’t exist is passed in. Previously, a new, empty file was created, and a KeyError raised (GH7715).

  • DataFrame.info() now ends its output with a newline character (GH8114)

  • Concatenating no objects will now raise a ValueError rather than a bare Exception.

  • Merge errors will now be sub-classes of ValueError rather than raw Exception (GH8501)

  • DataFrame.plot and Series.plot keywords are now have consistent orders (GH8037)

Internal Refactoring

In 0.15.0 Index has internally been refactored to no longer sub-class ndarray but instead subclass PandasObject, similarly to the rest of the pandas objects. This change allows very easy sub-classing and creation of new index types. This should be a transparent change with only very limited API implications (GH5080, GH7439, GH7796, GH8024, GH8367, GH7997, GH8522):

  • you may need to unpickle pandas version < 0.15.0 pickles using pd.read_pickle rather than pickle.load. See pickle docs
  • when plotting with a PeriodIndex, the matplotlib internal axes will now be arrays of Period rather than a PeriodIndex (this is similar to how a DatetimeIndex passes arrays of datetimes now)
  • MultiIndexes will now raise similarly to other pandas objects w.r.t. truth testing, see here (GH7897).
  • When plotting a DatetimeIndex directly with matplotlib’s plot function, the axis labels will no longer be formatted as dates but as integers (the internal representation of a datetime64). UPDATE This is fixed in 0.15.1, see here.

Deprecations

  • The attributes Categorical labels and levels attributes are deprecated and renamed to codes and categories.
  • The outtype argument to pd.DataFrame.to_dict has been deprecated in favor of orient. (GH7840)
  • The convert_dummies method has been deprecated in favor of get_dummies (GH8140)
  • The infer_dst argument in tz_localize will be deprecated in favor of ambiguous to allow for more flexibility in dealing with DST transitions. Replace infer_dst=True with ambiguous='infer' for the same behavior (GH7943). See the docs for more details.
  • The top-level pd.value_range has been deprecated and can be replaced by .describe() (GH8481)
  • The Index set operations + and - were deprecated in order to provide these for numeric type operations on certain index types. + can be replaced by .union() or |, and - by .difference(). Further the method name Index.diff() is deprecated and can be replaced by Index.difference() (GH8226)

    # +
    Index(['a','b','c']) + Index(['b','c','d'])
    
    # should be replaced by
    Index(['a','b','c']).union(Index(['b','c','d']))
    
    # -
    Index(['a','b','c']) - Index(['b','c','d'])
    
    # should be replaced by
    Index(['a','b','c']).difference(Index(['b','c','d']))
    
  • The infer_types argument to read_html() now has no effect and is deprecated (GH7762, GH7032).

Removal of prior version deprecations/changes

  • Remove DataFrame.delevel method in favor of DataFrame.reset_index

Enhancements

Enhancements in the importing/exporting of Stata files:

  • Added support for bool, uint8, uint16 and uint32 datatypes in to_stata (GH7097, GH7365)
  • Added conversion option when importing Stata files (GH8527)
  • DataFrame.to_stata and StataWriter check string length for compatibility with limitations imposed in dta files where fixed-width strings must contain 244 or fewer characters. Attempting to write Stata dta files with strings longer than 244 characters raises a ValueError. (GH7858)
  • read_stata and StataReader can import missing data information into a DataFrame by setting the argument convert_missing to True. When using this options, missing values are returned as StataMissingValue objects and columns containing missing values have object data type. (GH8045)

Enhancements in the plotting functions:

  • Added layout keyword to DataFrame.plot. You can pass a tuple of (rows, columns), one of which can be -1 to automatically infer (GH6667, GH8071).
  • Allow to pass multiple axes to DataFrame.plot, hist and boxplot (GH5353, GH6970, GH7069)
  • Added support for c, colormap and colorbar arguments for DataFrame.plot with kind='scatter' (GH7780)
  • Histogram from DataFrame.plot with kind='hist' (GH7809), See the docs.
  • Boxplot from DataFrame.plot with kind='box' (GH7998), See the docs.

Other:

  • read_csv now has a keyword parameter float_precision which specifies which floating-point converter the C engine should use during parsing, see here (GH8002, GH8044)

  • Added searchsorted method to Series objects (GH7447)

  • describe() on mixed-types DataFrames is more flexible. Type-based column filtering is now possible via the include/exclude arguments. See the docs (GH8164).

    In [97]: df = DataFrame({'catA': ['foo', 'foo', 'bar'] * 8,
       ....:                 'catB': ['a', 'b', 'c', 'd'] * 6,
       ....:                 'numC': np.arange(24),
       ....:                 'numD': np.arange(24.) + .5})
       ....: 
    
    In [98]: df.describe(include=["object"])
    Out[98]: 
           catA catB
    count    24   24
    unique    2    4
    top     foo    a
    freq     16    6
    
    In [99]: df.describe(include=["number", "object"], exclude=["float"])
    Out[99]: 
           catA catB       numC
    count    24   24  24.000000
    unique    2    4        NaN
    top     foo    a        NaN
    freq     16    6        NaN
    mean    NaN  NaN  11.500000
    std     NaN  NaN   7.071068
    min     NaN  NaN   0.000000
    25%     NaN  NaN   5.750000
    50%     NaN  NaN  11.500000
    75%     NaN  NaN  17.250000
    max     NaN  NaN  23.000000
    

    Requesting all columns is possible with the shorthand ‘all’

    In [100]: df.describe(include='all')
    Out[100]: 
           catA catB       numC       numD
    count    24   24  24.000000  24.000000
    unique    2    4        NaN        NaN
    top     foo    a        NaN        NaN
    freq     16    6        NaN        NaN
    mean    NaN  NaN  11.500000  12.000000
    std     NaN  NaN   7.071068   7.071068
    min     NaN  NaN   0.000000   0.500000
    25%     NaN  NaN   5.750000   6.250000
    50%     NaN  NaN  11.500000  12.000000
    75%     NaN  NaN  17.250000  17.750000
    max     NaN  NaN  23.000000  23.500000
    

    Without those arguments, describe will behave as before, including only numerical columns or, if none are, only categorical columns. See also the docs

  • Added split as an option to the orient argument in pd.DataFrame.to_dict. (GH7840)

  • The get_dummies method can now be used on DataFrames. By default only catagorical columns are encoded as 0’s and 1’s, while other columns are left untouched.

    In [101]: df = DataFrame({'A': ['a', 'b', 'a'], 'B': ['c', 'c', 'b'],
       .....:                 'C': [1, 2, 3]})
       .....: 
    
    In [102]: pd.get_dummies(df)
    Out[102]: 
       C  A_a  A_b  B_b  B_c
    0  1    1    0    0    1
    1  2    0    1    0    1
    2  3    1    0    1    0
    
  • PeriodIndex supports resolution as the same as DatetimeIndex (GH7708)

  • pandas.tseries.holiday has added support for additional holidays and ways to observe holidays (GH7070)

  • pandas.tseries.holiday.Holiday now supports a list of offsets in Python3 (GH7070)

  • pandas.tseries.holiday.Holiday now supports a days_of_week parameter (GH7070)

  • GroupBy.nth() now supports selecting multiple nth values (GH7910)

    In [103]: business_dates = date_range(start='4/1/2014', end='6/30/2014', freq='B')
    
    In [104]: df = DataFrame(1, index=business_dates, columns=['a', 'b'])
    
    # get the first, 4th, and last date index for each month
    In [105]: df.groupby((df.index.year, df.index.month)).nth([0, 3, -1])
    Out[105]: 
            a  b
    2014 4  1  1
         4  1  1
         4  1  1
         5  1  1
         5  1  1
         5  1  1
         6  1  1
         6  1  1
         6  1  1
    
  • Period and PeriodIndex supports addition/subtraction with timedelta-likes (GH7966)

    If Period freq is D, H, T, S, L, U, N, Timedelta-like can be added if the result can have same freq. Otherwise, only the same offsets can be added.

    In [106]: idx = pd.period_range('2014-07-01 09:00', periods=5, freq='H')
    
    In [107]: idx
    Out[107]: 
    PeriodIndex(['2014-07-01 09:00', '2014-07-01 10:00', '2014-07-01 11:00',
                 '2014-07-01 12:00', '2014-07-01 13:00'],
                dtype='period[H]', freq='H')
    
    In [108]: idx + pd.offsets.Hour(2)
    Out[108]: 
    PeriodIndex(['2014-07-01 11:00', '2014-07-01 12:00', '2014-07-01 13:00',
                 '2014-07-01 14:00', '2014-07-01 15:00'],
                dtype='period[H]', freq='H')
    
    In [109]: idx + Timedelta('120m')
    Out[109]: 
    PeriodIndex(['2014-07-01 11:00', '2014-07-01 12:00', '2014-07-01 13:00',
                 '2014-07-01 14:00', '2014-07-01 15:00'],
                dtype='period[H]', freq='H')
    
    In [110]: idx = pd.period_range('2014-07', periods=5, freq='M')
    
    In [111]: idx
    Out[111]: PeriodIndex(['2014-07', '2014-08', '2014-09', '2014-10', '2014-11'], dtype='period[M]', freq='M')
    
    In [112]: idx + pd.offsets.MonthEnd(3)
    Out[112]: PeriodIndex(['2014-10', '2014-11', '2014-12', '2015-01', '2015-02'], dtype='period[M]', freq='M')
    
  • Added experimental compatibility with openpyxl for versions >= 2.0. The DataFrame.to_excel method engine keyword now recognizes openpyxl1 and openpyxl2 which will explicitly require openpyxl v1 and v2 respectively, failing if the requested version is not available. The openpyxl engine is a now a meta-engine that automatically uses whichever version of openpyxl is installed. (GH7177)

  • DataFrame.fillna can now accept a DataFrame as a fill value (GH8377)

  • Passing multiple levels to stack() will now work when multiple level numbers are passed (GH7660). See Reshaping by stacking and unstacking.

  • set_names(), set_labels(), and set_levels() methods now take an optional level keyword argument to all modification of specific level(s) of a MultiIndex. Additionally set_names() now accepts a scalar string value when operating on an Index or on a specific level of a MultiIndex (GH7792)

    In [113]: idx = MultiIndex.from_product([['a'], range(3), list("pqr")], names=['foo', 'bar', 'baz'])
    
    In [114]: idx.set_names('qux', level=0)
    Out[114]: 
    MultiIndex(levels=[['a'], [0, 1, 2], ['p', 'q', 'r']],
               labels=[[0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 1, 1, 1, 2, 2, 2], [0, 1, 2, 0, 1, 2, 0, 1, 2]],
               names=['qux', 'bar', 'baz'])
    
    In [115]: idx.set_names(['qux','corge'], level=[0,1])
    Out[115]: 
    MultiIndex(levels=[['a'], [0, 1, 2], ['p', 'q', 'r']],
               labels=[[0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 1, 1, 1, 2, 2, 2], [0, 1, 2, 0, 1, 2, 0, 1, 2]],
               names=['qux', 'corge', 'baz'])
    
    In [116]: idx.set_levels(['a','b','c'], level='bar')
    Out[116]: 
    MultiIndex(levels=[['a'], ['a', 'b', 'c'], ['p', 'q', 'r']],
               labels=[[0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 1, 1, 1, 2, 2, 2], [0, 1, 2, 0, 1, 2, 0, 1, 2]],
               names=['foo', 'bar', 'baz'])
    
    In [117]: idx.set_levels([['a','b','c'],[1,2,3]], level=[1,2])
    Out[117]: 
    MultiIndex(levels=[['a'], ['a', 'b', 'c'], [1, 2, 3]],
               labels=[[0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 1, 1, 1, 2, 2, 2], [0, 1, 2, 0, 1, 2, 0, 1, 2]],
               names=['foo', 'bar', 'baz'])
    
  • Index.isin now supports a level argument to specify which index level to use for membership tests (GH7892, GH7890)

    In [1]: idx = MultiIndex.from_product([[0, 1], ['a', 'b', 'c']])
    
    In [2]: idx.values
    Out[2]: array([(0, 'a'), (0, 'b'), (0, 'c'), (1, 'a'), (1, 'b'), (1, 'c')], dtype=object)
    
    In [3]: idx.isin(['a', 'c', 'e'], level=1)
    Out[3]: array([ True, False,  True,  True, False,  True], dtype=bool)
    
  • Index now supports duplicated and drop_duplicates. (GH4060)

    In [118]: idx = Index([1, 2, 3, 4, 1, 2])
    
    In [119]: idx
    Out[119]: Int64Index([1, 2, 3, 4, 1, 2], dtype='int64')
    
    In [120]: idx.duplicated()
    Out[120]: array([False, False, False, False,  True,  True], dtype=bool)
    
    In [121]: idx.drop_duplicates()
    Out[121]: Int64Index([1, 2, 3, 4], dtype='int64')
    
  • add copy=True argument to pd.concat to enable pass thru of complete blocks (GH8252)

  • Added support for numpy 1.8+ data types (bool_, int_, float_, string_) for conversion to R dataframe (GH8400)

Performance

  • Performance improvements in DatetimeIndex.__iter__ to allow faster iteration (GH7683)
  • Performance improvements in Period creation (and PeriodIndex setitem) (GH5155)
  • Improvements in Series.transform for significant performance gains (revised) (GH6496)
  • Performance improvements in StataReader when reading large files (GH8040, GH8073)
  • Performance improvements in StataWriter when writing large files (GH8079)
  • Performance and memory usage improvements in multi-key groupby (GH8128)
  • Performance improvements in groupby .agg and .apply where builtins max/min were not mapped to numpy/cythonized versions (GH7722)
  • Performance improvement in writing to sql (to_sql) of up to 50% (GH8208).
  • Performance benchmarking of groupby for large value of ngroups (GH6787)
  • Performance improvement in CustomBusinessDay, CustomBusinessMonth (GH8236)
  • Performance improvement for MultiIndex.values for multi-level indexes containing datetimes (GH8543)

Bug Fixes

  • Bug in pivot_table, when using margins and a dict aggfunc (GH8349)
  • Bug in read_csv where squeeze=True would return a view (GH8217)
  • Bug in checking of table name in read_sql in certain cases (GH7826).
  • Bug in DataFrame.groupby where Grouper does not recognize level when frequency is specified (GH7885)
  • Bug in multiindexes dtypes getting mixed up when DataFrame is saved to SQL table (GH8021)
  • Bug in Series 0-division with a float and integer operand dtypes (GH7785)
  • Bug in Series.astype("unicode") not calling unicode on the values correctly (GH7758)
  • Bug in DataFrame.as_matrix() with mixed datetime64[ns] and timedelta64[ns] dtypes (GH7778)
  • Bug in HDFStore.select_column() not preserving UTC timezone info when selecting a DatetimeIndex (GH7777)
  • Bug in to_datetime when format='%Y%m%d' and coerce=True are specified, where previously an object array was returned (rather than a coerced time-series with NaT), (GH7930)
  • Bug in DatetimeIndex and PeriodIndex in-place addition and subtraction cause different result from normal one (GH6527)
  • Bug in adding and subtracting PeriodIndex with PeriodIndex raise TypeError (GH7741)
  • Bug in combine_first with PeriodIndex data raises TypeError (GH3367)
  • Bug in multi-index slicing with missing indexers (GH7866)
  • Bug in multi-index slicing with various edge cases (GH8132)
  • Regression in multi-index indexing with a non-scalar type object (GH7914)
  • Bug in Timestamp comparisons with == and int64 dtype (GH8058)
  • Bug in pickles contains DateOffset may raise AttributeError when normalize attribute is referred internally (GH7748)
  • Bug in Panel when using major_xs and copy=False is passed (deprecation warning fails because of missing warnings) (GH8152).
  • Bug in pickle deserialization that failed for pre-0.14.1 containers with dup items trying to avoid ambiguity when matching block and manager items, when there’s only one block there’s no ambiguity (GH7794)
  • Bug in putting a PeriodIndex into a Series would convert to int64 dtype, rather than object of Periods (GH7932)
  • Bug in HDFStore iteration when passing a where (GH8014)
  • Bug in DataFrameGroupby.transform when transforming with a passed non-sorted key (GH8046, GH8430)
  • Bug in repeated timeseries line and area plot may result in ValueError or incorrect kind (GH7733)
  • Bug in inference in a MultiIndex with datetime.date inputs (GH7888)
  • Bug in get where an IndexError would not cause the default value to be returned (GH7725)
  • Bug in offsets.apply, rollforward and rollback may reset nanosecond (GH7697)
  • Bug in offsets.apply, rollforward and rollback may raise AttributeError if Timestamp has dateutil tzinfo (GH7697)
  • Bug in sorting a multi-index frame with a Float64Index (GH8017)
  • Bug in inconsistent panel setitem with a rhs of a DataFrame for alignment (GH7763)
  • Bug in is_superperiod and is_subperiod cannot handle higher frequencies than S (GH7760, GH7772, GH7803)
  • Bug in 32-bit platforms with Series.shift (GH8129)
  • Bug in PeriodIndex.unique returns int64 np.ndarray (GH7540)
  • Bug in groupby.apply with a non-affecting mutation in the function (GH8467)
  • Bug in DataFrame.reset_index which has MultiIndex contains PeriodIndex or DatetimeIndex with tz raises ValueError (GH7746, GH7793)
  • Bug in DataFrame.plot with subplots=True may draw unnecessary minor xticks and yticks (GH7801)
  • Bug in StataReader which did not read variable labels in 117 files due to difference between Stata documentation and implementation (GH7816)
  • Bug in StataReader where strings were always converted to 244 characters-fixed width irrespective of underlying string size (GH7858)
  • Bug in DataFrame.plot and Series.plot may ignore rot and fontsize keywords (GH7844)
  • Bug in DatetimeIndex.value_counts doesn’t preserve tz (GH7735)
  • Bug in PeriodIndex.value_counts results in Int64Index (GH7735)
  • Bug in DataFrame.join when doing left join on index and there are multiple matches (GH5391)
  • Bug in GroupBy.transform() where int groups with a transform that didn’t preserve the index were incorrectly truncated (GH7972).
  • Bug in groupby where callable objects without name attributes would take the wrong path, and produce a DataFrame instead of a Series (GH7929)
  • Bug in groupby error message when a DataFrame grouping column is duplicated (GH7511)
  • Bug in read_html where the infer_types argument forced coercion of date-likes incorrectly (GH7762, GH7032).
  • Bug in Series.str.cat with an index which was filtered as to not include the first item (GH7857)
  • Bug in Timestamp cannot parse nanosecond from string (GH7878)
  • Bug in Timestamp with string offset and tz results incorrect (GH7833)
  • Bug in tslib.tz_convert and tslib.tz_convert_single may return different results (GH7798)
  • Bug in DatetimeIndex.intersection of non-overlapping timestamps with tz raises IndexError (GH7880)
  • Bug in alignment with TimeOps and non-unique indexes (GH8363)
  • Bug in GroupBy.filter() where fast path vs. slow path made the filter return a non scalar value that appeared valid but wasn’t (GH7870).
  • Bug in date_range()/DatetimeIndex() when the timezone was inferred from input dates yet incorrect times were returned when crossing DST boundaries (GH7835, GH7901).
  • Bug in to_excel() where a negative sign was being prepended to positive infinity and was absent for negative infinity (GH7949)
  • Bug in area plot draws legend with incorrect alpha when stacked=True (GH8027)
  • Period and PeriodIndex addition/subtraction with np.timedelta64 results in incorrect internal representations (GH7740)
  • Bug in Holiday with no offset or observance (GH7987)
  • Bug in DataFrame.to_latex formatting when columns or index is a MultiIndex (GH7982).
  • Bug in DateOffset around Daylight Savings Time produces unexpected results (GH5175).
  • Bug in DataFrame.shift where empty columns would throw ZeroDivisionError on numpy 1.7 (GH8019)
  • Bug in installation where html_encoding/*.html wasn’t installed and therefore some tests were not running correctly (GH7927).
  • Bug in read_html where bytes objects were not tested for in _read (GH7927).
  • Bug in DataFrame.stack() when one of the column levels was a datelike (GH8039)
  • Bug in broadcasting numpy scalars with DataFrame (GH8116)
  • Bug in pivot_table performed with nameless index and columns raises KeyError (GH8103)
  • Bug in DataFrame.plot(kind='scatter') draws points and errorbars with different colors when the color is specified by c keyword (GH8081)
  • Bug in Float64Index where iat and at were not testing and were failing (GH8092).
  • Bug in DataFrame.boxplot() where y-limits were not set correctly when producing multiple axes (GH7528, GH5517).
  • Bug in read_csv where line comments were not handled correctly given a custom line terminator or delim_whitespace=True (GH8122).
  • Bug in read_html where empty tables caused a StopIteration (GH7575)
  • Bug in casting when setting a column in a same-dtype block (GH7704)
  • Bug in accessing groups from a GroupBy when the original grouper was a tuple (GH8121).
  • Bug in .at that would accept integer indexers on a non-integer index and do fallback (GH7814)
  • Bug with kde plot and NaNs (GH8182)
  • Bug in GroupBy.count with float32 data type were nan values were not excluded (GH8169).
  • Bug with stacked barplots and NaNs (GH8175).
  • Bug in resample with non evenly divisible offsets (e.g. ‘7s’) (GH8371)
  • Bug in interpolation methods with the limit keyword when no values needed interpolating (GH7173).
  • Bug where col_space was ignored in DataFrame.to_string() when header=False (GH8230).
  • Bug with DatetimeIndex.asof incorrectly matching partial strings and returning the wrong date (GH8245).
  • Bug in plotting methods modifying the global matplotlib rcParams (GH8242).
  • Bug in DataFrame.__setitem__ that caused errors when setting a dataframe column to a sparse array (GH8131)
  • Bug where Dataframe.boxplot() failed when entire column was empty (GH8181).
  • Bug with messed variables in radviz visualization (GH8199).
  • Bug in interpolation methods with the limit keyword when no values needed interpolating (GH7173).
  • Bug where col_space was ignored in DataFrame.to_string() when header=False (GH8230).
  • Bug in to_clipboard that would clip long column data (GH8305)
  • Bug in DataFrame terminal display: Setting max_column/max_rows to zero did not trigger auto-resizing of dfs to fit terminal width/height (GH7180).
  • Bug in OLS where running with “cluster” and “nw_lags” parameters did not work correctly, but also did not throw an error (GH5884).
  • Bug in DataFrame.dropna that interpreted non-existent columns in the subset argument as the ‘last column’ (GH8303)
  • Bug in Index.intersection on non-monotonic non-unique indexes (GH8362).
  • Bug in masked series assignment where mismatching types would break alignment (GH8387)
  • Bug in NDFrame.equals gives false negatives with dtype=object (GH8437)
  • Bug in assignment with indexer where type diversity would break alignment (GH8258)
  • Bug in NDFrame.loc indexing when row/column names were lost when target was a list/ndarray (GH6552)
  • Regression in NDFrame.loc indexing when rows/columns were converted to Float64Index if target was an empty list/ndarray (GH7774)
  • Bug in Series that allows it to be indexed by a DataFrame which has unexpected results. Such indexing is no longer permitted (GH8444)
  • Bug in item assignment of a DataFrame with multi-index columns where right-hand-side columns were not aligned (GH7655)
  • Suppress FutureWarning generated by NumPy when comparing object arrays containing NaN for equality (GH7065)
  • Bug in DataFrame.eval() where the dtype of the not operator (~) was not correctly inferred as bool.

v0.14.1 (July 11, 2014)

This is a minor release from 0.14.0 and includes a small number of API changes, several new features, enhancements, and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this version.

API changes

  • Openpyxl now raises a ValueError on construction of the openpyxl writer instead of warning on pandas import (GH7284).

  • For StringMethods.extract, when no match is found, the result - only containing NaN values - now also has dtype=object instead of float (GH7242)

  • Period objects no longer raise a TypeError when compared using == with another object that isn’t a Period. Instead when comparing a Period with another object using == if the other object isn’t a Period False is returned. (GH7376)

  • Previously, the behaviour on resetting the time or not in offsets.apply, rollforward and rollback operations differed between offsets. With the support of the normalize keyword for all offsets(see below) with a default value of False (preserve time), the behaviour changed for certain offsets (BusinessMonthBegin, MonthEnd, BusinessMonthEnd, CustomBusinessMonthEnd, BusinessYearBegin, LastWeekOfMonth, FY5253Quarter, LastWeekOfMonth, Easter):

    In [6]: from pandas.tseries import offsets
    
    In [7]: d = pd.Timestamp('2014-01-01 09:00')
    
    # old behaviour < 0.14.1
    In [8]: d + offsets.MonthEnd()
    Out[8]: Timestamp('2014-01-31 00:00:00')
    

    Starting from 0.14.1 all offsets preserve time by default. The old behaviour can be obtained with normalize=True

    # new behaviour
    In [1]: d + offsets.MonthEnd()
    Out[1]: Timestamp('2014-01-31 09:00:00')
    
    In [2]: d + offsets.MonthEnd(normalize=True)
    Out[2]: Timestamp('2014-01-31 00:00:00')
    

    Note that for the other offsets the default behaviour did not change.

  • Add back #N/A N/A as a default NA value in text parsing, (regression from 0.12) (GH5521)

  • Raise a TypeError on inplace-setting with a .where and a non np.nan value as this is inconsistent with a set-item expression like df[mask] = None (GH7656)

Enhancements

  • Add dropna argument to value_counts and nunique (GH5569).

  • Add select_dtypes() method to allow selection of columns based on dtype (GH7316). See the docs.

  • All offsets supports the normalize keyword to specify whether offsets.apply, rollforward and rollback resets the time (hour, minute, etc) or not (default False, preserves time) (GH7156):

    In [3]: import pandas.tseries.offsets as offsets
    
    In [4]: day = offsets.Day()
    
    In [5]: day.apply(Timestamp('2014-01-01 09:00'))
    Out[5]: Timestamp('2014-01-02 09:00:00')
    
    In [6]: day = offsets.Day(normalize=True)
    
    In [7]: day.apply(Timestamp('2014-01-01 09:00'))
    Out[7]: Timestamp('2014-01-02 00:00:00')
    
  • PeriodIndex is represented as the same format as DatetimeIndex (GH7601)

  • StringMethods now work on empty Series (GH7242)

  • The file parsers read_csv and read_table now ignore line comments provided by the parameter comment, which accepts only a single character for the C reader. In particular, they allow for comments before file data begins (GH2685)

  • Add NotImplementedError for simultaneous use of chunksize and nrows for read_csv() (GH6774).

  • Tests for basic reading of public S3 buckets now exist (GH7281).

  • read_html now sports an encoding argument that is passed to the underlying parser library. You can use this to read non-ascii encoded web pages (GH7323).

  • read_excel now supports reading from URLs in the same way that read_csv does. (GH6809)

  • Support for dateutil timezones, which can now be used in the same way as pytz timezones across pandas. (GH4688)

    In [8]: rng = date_range('3/6/2012 00:00', periods=10, freq='D',
       ...:                  tz='dateutil/Europe/London')
       ...: 
    
    In [9]: rng.tz
    Out[9]: tzfile('/usr/share/zoneinfo/Europe/London')
    

    See the docs.

  • Implemented sem (standard error of the mean) operation for Series, DataFrame, Panel, and Groupby (GH6897)

  • Add nlargest and nsmallest to the Series groupby whitelist, which means you can now use these methods on a SeriesGroupBy object (GH7053).

  • All offsets apply, rollforward and rollback can now handle np.datetime64, previously results in ApplyTypeError (GH7452)

  • Period and PeriodIndex can contain NaT in its values (GH7485)

  • Support pickling Series, DataFrame and Panel objects with non-unique labels along item axis (index, columns and items respectively) (GH7370).

  • Improved inference of datetime/timedelta with mixed null objects. Regression from 0.13.1 in interpretation of an object Index with all null elements (GH7431)

Performance

  • Improvements in dtype inference for numeric operations involving yielding performance gains for dtypes: int64, timedelta64, datetime64 (GH7223)
  • Improvements in Series.transform for significant performance gains (GH6496)
  • Improvements in DataFrame.transform with ufuncs and built-in grouper functions for significant performance gains (GH7383)
  • Regression in groupby aggregation of datetime64 dtypes (GH7555)
  • Improvements in MultiIndex.from_product for large iterables (GH7627)

Experimental

  • pandas.io.data.Options has a new method, get_all_data method, and now consistently returns a multi-indexed DataFrame (GH5602)
  • io.gbq.read_gbq and io.gbq.to_gbq were refactored to remove the dependency on the Google bq.py command line client. This submodule now uses httplib2 and the Google apiclient and oauth2client API client libraries which should be more stable and, therefore, reliable than bq.py. See the docs. (GH6937).

Bug Fixes

  • Bug in DataFrame.where with a symmetric shaped frame and a passed other of a DataFrame (GH7506)
  • Bug in Panel indexing with a multi-index axis (GH7516)
  • Regression in datetimelike slice indexing with a duplicated index and non-exact end-points (GH7523)
  • Bug in setitem with list-of-lists and single vs mixed types (GH7551:)
  • Bug in timeops with non-aligned Series (GH7500)
  • Bug in timedelta inference when assigning an incomplete Series (GH7592)
  • Bug in groupby .nth with a Series and integer-like column name (GH7559)
  • Bug in Series.get with a boolean accessor (GH7407)
  • Bug in value_counts where NaT did not qualify as missing (NaN) (GH7423)
  • Bug in to_timedelta that accepted invalid units and misinterpreted ‘m/h’ (GH7611, GH6423)
  • Bug in line plot doesn’t set correct xlim if secondary_y=True (GH7459)
  • Bug in grouped hist and scatter plots use old figsize default (GH7394)
  • Bug in plotting subplots with DataFrame.plot, hist clears passed ax even if the number of subplots is one (GH7391).
  • Bug in plotting subplots with DataFrame.boxplot with by kw raises ValueError if the number of subplots exceeds 1 (GH7391).
  • Bug in subplots displays ticklabels and labels in different rule (GH5897)
  • Bug in Panel.apply with a multi-index as an axis (GH7469)
  • Bug in DatetimeIndex.insert doesn’t preserve name and tz (GH7299)
  • Bug in DatetimeIndex.asobject doesn’t preserve name (GH7299)
  • Bug in multi-index slicing with datetimelike ranges (strings and Timestamps), (GH7429)
  • Bug in Index.min and max doesn’t handle nan and NaT properly (GH7261)
  • Bug in PeriodIndex.min/max results in int (GH7609)
  • Bug in resample where fill_method was ignored if you passed how (GH2073)
  • Bug in TimeGrouper doesn’t exclude column specified by key (GH7227)
  • Bug in DataFrame and Series bar and barh plot raises TypeError when bottom and left keyword is specified (GH7226)
  • Bug in DataFrame.hist raises TypeError when it contains non numeric column (GH7277)
  • Bug in Index.delete does not preserve name and freq attributes (GH7302)
  • Bug in DataFrame.query()/eval where local string variables with the @ sign were being treated as temporaries attempting to be deleted (GH7300).
  • Bug in Float64Index which didn’t allow duplicates (GH7149).
  • Bug in DataFrame.replace() where truthy values were being replaced (GH7140).
  • Bug in StringMethods.extract() where a single match group Series would use the matcher’s name instead of the group name (GH7313).
  • Bug in isnull() when mode.use_inf_as_null == True where isnull wouldn’t test True when it encountered an inf/-inf (GH7315).
  • Bug in inferred_freq results in None for eastern hemisphere timezones (GH7310)
  • Bug in Easter returns incorrect date when offset is negative (GH7195)
  • Bug in broadcasting with .div, integer dtypes and divide-by-zero (GH7325)
  • Bug in CustomBusinessDay.apply raiases NameError when np.datetime64 object is passed (GH7196)
  • Bug in MultiIndex.append, concat and pivot_table don’t preserve timezone (GH6606)
  • Bug in .loc with a list of indexers on a single-multi index level (that is not nested) (GH7349)
  • Bug in Series.map when mapping a dict with tuple keys of different lengths (GH7333)
  • Bug all StringMethods now work on empty Series (GH7242)
  • Fix delegation of read_sql to read_sql_query when query does not contain ‘select’ (GH7324).
  • Bug where a string column name assignment to a DataFrame with a Float64Index raised a TypeError during a call to np.isnan (GH7366).
  • Bug where NDFrame.replace() didn’t correctly replace objects with Period values (GH7379).
  • Bug in .ix getitem should always return a Series (GH7150)
  • Bug in multi-index slicing with incomplete indexers (GH7399)
  • Bug in multi-index slicing with a step in a sliced level (GH7400)
  • Bug where negative indexers in DatetimeIndex were not correctly sliced (GH7408)
  • Bug where NaT wasn’t repr’d correctly in a MultiIndex (GH7406, GH7409).
  • Bug where bool objects were converted to nan in convert_objects (GH7416).
  • Bug in quantile ignoring the axis keyword argument (GH7306)
  • Bug where nanops._maybe_null_out doesn’t work with complex numbers (GH7353)
  • Bug in several nanops functions when axis==0 for 1-dimensional nan arrays (GH7354)
  • Bug where nanops.nanmedian doesn’t work when axis==None (GH7352)
  • Bug where nanops._has_infs doesn’t work with many dtypes (GH7357)
  • Bug in StataReader.data where reading a 0-observation dta failed (GH7369)
  • Bug in StataReader when reading Stata 13 (117) files containing fixed width strings (GH7360)
  • Bug in StataWriter where encoding was ignored (GH7286)
  • Bug in DatetimeIndex comparison doesn’t handle NaT properly (GH7529)
  • Bug in passing input with tzinfo to some offsets apply, rollforward or rollback resets tzinfo or raises ValueError (GH7465)
  • Bug in DatetimeIndex.to_period, PeriodIndex.asobject, PeriodIndex.to_timestamp doesn’t preserve name (GH7485)
  • Bug in DatetimeIndex.to_period and PeriodIndex.to_timestanp handle NaT incorrectly (GH7228)
  • Bug in offsets.apply, rollforward and rollback may return normal datetime (GH7502)
  • Bug in resample raises ValueError when target contains NaT (GH7227)
  • Bug in Timestamp.tz_localize resets nanosecond info (GH7534)
  • Bug in DatetimeIndex.asobject raises ValueError when it contains NaT (GH7539)
  • Bug in Timestamp.__new__ doesn’t preserve nanosecond properly (GH7610)
  • Bug in Index.astype(float) where it would return an object dtype Index (GH7464).
  • Bug in DataFrame.reset_index loses tz (GH3950)
  • Bug in DatetimeIndex.freqstr raises AttributeError when freq is None (GH7606)
  • Bug in GroupBy.size created by TimeGrouper raises AttributeError (GH7453)
  • Bug in single column bar plot is misaligned (GH7498).
  • Bug in area plot with tz-aware time series raises ValueError (GH7471)
  • Bug in non-monotonic Index.union may preserve name incorrectly (GH7458)
  • Bug in DatetimeIndex.intersection doesn’t preserve timezone (GH4690)
  • Bug in rolling_var where a window larger than the array would raise an error(GH7297)
  • Bug with last plotted timeseries dictating xlim (GH2960)
  • Bug with secondary_y axis not being considered for timeseries xlim (GH3490)
  • Bug in Float64Index assignment with a non scalar indexer (GH7586)
  • Bug in pandas.core.strings.str_contains does not properly match in a case insensitive fashion when regex=False and case=False (GH7505)
  • Bug in expanding_cov, expanding_corr, rolling_cov, and rolling_corr for two arguments with mismatched index (GH7512)
  • Bug in to_sql taking the boolean column as text column (GH7678)
  • Bug in grouped hist doesn’t handle rot kw and sharex kw properly (GH7234)
  • Bug in .loc performing fallback integer indexing with object dtype indices (GH7496)
  • Bug (regression) in PeriodIndex constructor when passed Series objects (GH7701).

v0.14.0 (May 31 , 2014)

This is a major release from 0.13.1 and includes a small number of API changes, several new features, enhancements, and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this version.

Warning

In 0.14.0 all NDFrame based containers have undergone significant internal refactoring. Before that each block of homogeneous data had its own labels and extra care was necessary to keep those in sync with the parent container’s labels. This should not have any visible user/API behavior changes (GH6745)

API changes

  • read_excel uses 0 as the default sheet (GH6573)

  • iloc will now accept out-of-bounds indexers for slices, e.g. a value that exceeds the length of the object being indexed. These will be excluded. This will make pandas conform more with python/numpy indexing of out-of-bounds values. A single indexer that is out-of-bounds and drops the dimensions of the object will still raise IndexError (GH6296, GH6299). This could result in an empty axis (e.g. an empty DataFrame being returned)

    In [1]: dfl = DataFrame(np.random.randn(5,2),columns=list('AB'))
    
    In [2]: dfl
    Out[2]: 
              A         B
    0  1.583584 -0.438313
    1 -0.402537 -0.780572
    2 -0.141685  0.542241
    3  0.370966 -0.251642
    4  0.787484  1.666563
    
    In [3]: dfl.iloc[:,2:3]
    Out[3]: 
    Empty DataFrame
    Columns: []
    Index: [0, 1, 2, 3, 4]
    
    In [4]: dfl.iloc[:,1:3]
    Out[4]: 
              B
    0 -0.438313
    1 -0.780572
    2  0.542241
    3 -0.251642
    4  1.666563
    
    In [5]: dfl.iloc[4:6]
    Out[5]: 
              A         B
    4  0.787484  1.666563
    

    These are out-of-bounds selections

    dfl.iloc[[4,5,6]]
    IndexError: positional indexers are out-of-bounds
    
    dfl.iloc[:,4]
    IndexError: single positional indexer is out-of-bounds
    
  • Slicing with negative start, stop & step values handles corner cases better (GH6531):

    • df.iloc[:-len(df)] is now empty
    • df.iloc[len(df)::-1] now enumerates all elements in reverse
  • The DataFrame.interpolate() keyword downcast default has been changed from infer to None. This is to preseve the original dtype unless explicitly requested otherwise (GH6290).

  • When converting a dataframe to HTML it used to return Empty DataFrame. This special case has been removed, instead a header with the column names is returned (GH6062).

  • Series and Index now internall share more common operations, e.g. factorize(),nunique(),value_counts() are now supported on Index types as well. The Series.weekday property from is removed from Series for API consistency. Using a DatetimeIndex/PeriodIndex method on a Series will now raise a TypeError. (GH4551, GH4056, GH5519, GH6380, GH7206).

  • Add is_month_start, is_month_end, is_quarter_start, is_quarter_end, is_year_start, is_year_end accessors for DateTimeIndex / Timestamp which return a boolean array of whether the timestamp(s) are at the start/end of the month/quarter/year defined by the frequency of the DateTimeIndex / Timestamp (GH4565, GH6998)

  • Local variable usage has changed in pandas.eval()/DataFrame.eval()/DataFrame.query() (GH5987). For the DataFrame methods, two things have changed

    • Column names are now given precedence over locals
    • Local variables must be referred to explicitly. This means that even if you have a local variable that is not a column you must still refer to it with the '@' prefix.
    • You can have an expression like df.query('@a < a') with no complaints from pandas about ambiguity of the name a.
    • The top-level pandas.eval() function does not allow you use the '@' prefix and provides you with an error message telling you so.
    • NameResolutionError was removed because it isn’t necessary anymore.
  • Define and document the order of column vs index names in query/eval (GH6676)

  • concat will now concatenate mixed Series and DataFrames using the Series name or numbering columns as needed (GH2385). See the docs

  • Slicing and advanced/boolean indexing operations on Index classes as well as Index.delete() and Index.drop() methods will no longer change the type of the resulting index (GH6440, GH7040)

    In [6]: i = pd.Index([1, 2, 3, 'a' , 'b', 'c'])
    
    In [7]: i[[0,1,2]]
    Out[7]: Index([1, 2, 3], dtype='object')
    
    In [8]: i.drop(['a', 'b', 'c'])
    Out[8]: Index([1, 2, 3], dtype='object')
    

    Previously, the above operation would return Int64Index. If you’d like to do this manually, use Index.astype()

    In [9]: i[[0,1,2]].astype(np.int_)
    Out[9]: Int64Index([1, 2, 3], dtype='int64')
    
  • set_index no longer converts MultiIndexes to an Index of tuples. For example, the old behavior returned an Index in this case (GH6459):

    # Old behavior, casted MultiIndex to an Index
    In [10]: tuple_ind
    Out[10]: Index([('a', 'c'), ('a', 'd'), ('b', 'c'), ('b', 'd')], dtype='object')
    
    In [11]: df_multi.set_index(tuple_ind)
    Out[11]: 
                   0         1
    (a, c)  0.471435 -1.190976
    (a, d)  1.432707 -0.312652
    (b, c) -0.720589  0.887163
    (b, d)  0.859588 -0.636524
    
    # New behavior
    In [12]: mi
    Out[12]: 
    MultiIndex(levels=[['a', 'b'], ['c', 'd']],
               labels=[[0, 0, 1, 1], [0, 1, 0, 1]])
    
    In [13]: df_multi.set_index(mi)
    Out[13]: 
                0         1
    a c  0.471435 -1.190976
      d  1.432707 -0.312652
    b c -0.720589  0.887163
      d  0.859588 -0.636524
    

    This also applies when passing multiple indices to set_index:

    # Old output, 2-level MultiIndex of tuples
    In [14]: df_multi.set_index([df_multi.index, df_multi.index])
    Out[14]: 
                          0         1
    (a, c) (a, c)  0.471435 -1.190976
    (a, d) (a, d)  1.432707 -0.312652
    (b, c) (b, c) -0.720589  0.887163
    (b, d) (b, d)  0.859588 -0.636524
    
    # New output, 4-level MultiIndex
    In [15]: df_multi.set_index([df_multi.index, df_multi.index])
    Out[15]: 
                    0         1
    a c a c  0.471435 -1.190976
      d a d  1.432707 -0.312652
    b c b c -0.720589  0.887163
      d b d  0.859588 -0.636524
    
  • pairwise keyword was added to the statistical moment functions rolling_cov, rolling_corr, ewmcov, ewmcorr, expanding_cov, expanding_corr to allow the calculation of moving window covariance and correlation matrices (GH4950). See Computing rolling pairwise covariances and correlations in the docs.

    In [1]: df = DataFrame(np.random.randn(10,4),columns=list('ABCD'))
    
    In [4]: covs = pd.rolling_cov(df[['A','B','C']], df[['B','C','D']], 5, pairwise=True)
    
    In [5]: covs[df.index[-1]]
    Out[5]:
              B         C         D
    A  0.035310  0.326593 -0.505430
    B  0.137748 -0.006888 -0.005383
    C -0.006888  0.861040  0.020762
    
  • Series.iteritems() is now lazy (returns an iterator rather than a list). This was the documented behavior prior to 0.14. (GH6760)

  • Added nunique and value_counts functions to Index for counting unique elements. (GH6734)

  • stack and unstack now raise a ValueError when the level keyword refers to a non-unique item in the Index (previously raised a KeyError). (GH6738)

  • drop unused order argument from Series.sort; args now are in the same order as Series.order; add na_position arg to conform to Series.order (GH6847)

  • default sorting algorithm for Series.order is now quicksort, to conform with Series.sort (and numpy defaults)

  • add inplace keyword to Series.order/sort to make them inverses (GH6859)

  • DataFrame.sort now places NaNs at the beginning or end of the sort according to the na_position parameter. (GH3917)

  • accept TextFileReader in concat, which was affecting a common user idiom (GH6583), this was a regression from 0.13.1

  • Added factorize functions to Index and Series to get indexer and unique values (GH7090)

  • describe on a DataFrame with a mix of Timestamp and string like objects returns a different Index (GH7088). Previously the index was unintentionally sorted.

  • Arithmetic operations with only bool dtypes now give a warning indicating that they are evaluated in Python space for +, -, and * operations and raise for all others (GH7011, GH6762, GH7015, GH7210)

    x = pd.Series(np.random.rand(10) > 0.5)
    y = True
    x + y  # warning generated: should do x | y instead
    x / y  # this raises because it doesn't make sense
    
    NotImplementedError: operator '/' not implemented for bool dtypes
    
  • In HDFStore, select_as_multiple will always raise a KeyError, when a key or the selector is not found (GH6177)

  • df['col'] = value and df.loc[:,'col'] = value are now completely equivalent; previously the .loc would not necessarily coerce the dtype of the resultant series (GH6149)

  • dtypes and ftypes now return a series with dtype=object on empty containers (GH5740)

  • df.to_csv will now return a string of the CSV data if neither a target path nor a buffer is provided (GH6061)

  • pd.infer_freq() will now raise a TypeError if given an invalid Series/Index type (GH6407, GH6463)

  • A tuple passed to DataFame.sort_index will be interpreted as the levels of the index, rather than requiring a list of tuple (GH4370)

  • all offset operations now return Timestamp types (rather than datetime), Business/Week frequencies were incorrect (GH4069)

  • to_excel now converts np.inf into a string representation, customizable by the inf_rep keyword argument (Excel has no native inf representation) (GH6782)

  • Replace pandas.compat.scipy.scoreatpercentile with numpy.percentile (GH6810)

  • .quantile on a datetime[ns] series now returns Timestamp instead of np.datetime64 objects (GH6810)

  • change AssertionError to TypeError for invalid types passed to concat (GH6583)

  • Raise a TypeError when DataFrame is passed an iterator as the data argument (GH5357)

Display Changes

  • The default way of printing large DataFrames has changed. DataFrames exceeding max_rows and/or max_columns are now displayed in a centrally truncated view, consistent with the printing of a pandas.Series (GH5603).

    In previous versions, a DataFrame was truncated once the dimension constraints were reached and an ellipse (…) signaled that part of the data was cut off.

    The previous look of truncate.

    In the current version, large DataFrames are centrally truncated, showing a preview of head and tail in both dimensions.

    The new look.
  • allow option 'truncate' for display.show_dimensions to only show the dimensions if the frame is truncated (GH6547).

    The default for display.show_dimensions will now be truncate. This is consistent with how Series display length.

    In [16]: dfd = pd.DataFrame(np.arange(25).reshape(-1,5), index=[0,1,2,3,4], columns=[0,1,2,3,4])
    
    # show dimensions since this is truncated
    In [17]: with pd.option_context('display.max_rows', 2, 'display.max_columns', 2,
       ....:                        'display.show_dimensions', 'truncate'):
       ....:    print(dfd)
       ....: 
         0 ...   4
    0    0 ...   4
    ..  .. ...  ..
    4   20 ...  24
    
    [5 rows x 5 columns]
    
    # will not show dimensions since it is not truncated
    In [18]: with pd.option_context('display.max_rows', 10, 'display.max_columns', 40,
       ....:                        'display.show_dimensions', 'truncate'):
       ....:    print(dfd)
       ....: 
        0   1   2   3   4
    0   0   1   2   3   4
    1   5   6   7   8   9
    2  10  11  12  13  14
    3  15  16  17  18  19
    4  20  21  22  23  24
    
  • Regression in the display of a MultiIndexed Series with display.max_rows is less than the length of the series (GH7101)

  • Fixed a bug in the HTML repr of a truncated Series or DataFrame not showing the class name with the large_repr set to ‘info’ (GH7105)

  • The verbose keyword in DataFrame.info(), which controls whether to shorten the info representation, is now None by default. This will follow the global setting in display.max_info_columns. The global setting can be overridden with verbose=True or verbose=False.

  • Fixed a bug with the info repr not honoring the display.max_info_columns setting (GH6939)

  • Offset/freq info now in Timestamp __repr__ (GH4553)

Text Parsing API Changes

read_csv()/read_table() will now be noiser w.r.t invalid options rather than falling back to the PythonParser.

  • Raise ValueError when sep specified with delim_whitespace=True in read_csv()/read_table() (GH6607)
  • Raise ValueError when engine='c' specified with unsupported options in read_csv()/read_table() (GH6607)
  • Raise ValueError when fallback to python parser causes options to be ignored (GH6607)
  • Produce ParserWarning on fallback to python parser when no options are ignored (GH6607)
  • Translate sep='\s+' to delim_whitespace=True in read_csv()/read_table() if no other C-unsupported options specified (GH6607)

Groupby API Changes

More consistent behaviour for some groupby methods:

  • groupby head and tail now act more like filter rather than an aggregation:

    In [19]: df = pd.DataFrame([[1, 2], [1, 4], [5, 6]], columns=['A', 'B'])
    
    In [20]: g = df.groupby('A')
    
    In [21]: g.head(1)  # filters DataFrame
    Out[21]: 
       A  B
    0  1  2
    2  5  6
    
    In [22]: g.apply(lambda x: x.head(1))  # used to simply fall-through
    Out[22]: 
         A  B
    A        
    1 0  1  2
    5 2  5  6
    
  • groupby head and tail respect column selection:

    In [23]: g[['B']].head(1)
    Out[23]: 
       B
    0  2
    2  6
    
  • groupby nth now reduces by default; filtering can be achieved by passing as_index=False. With an optional dropna argument to ignore NaN. See the docs.

    Reducing

    In [24]: df = DataFrame([[1, np.nan], [1, 4], [5, 6]], columns=['A', 'B'])
    
    In [25]: g = df.groupby('A')
    
    In [26]: g.nth(0)
    Out[26]: 
         B
    A     
    1  NaN
    5  6.0
    
    # this is equivalent to g.first()
    In [27]: g.nth(0, dropna='any')
    Out[27]: 
         B
    A     
    1  4.0
    5  6.0
    
    # this is equivalent to g.last()
    In [28]: g.nth(-1, dropna='any')
    Out[28]: 
         B
    A     
    1  4.0
    5  6.0
    

    Filtering

    In [29]: gf = df.groupby('A',as_index=False)
    
    In [30]: gf.nth(0)
    Out[30]: 
       A    B
    0  1  NaN
    2  5  6.0
    
    In [31]: gf.nth(0, dropna='any')
    Out[31]: 
       A    B
    A        
    1  1  4.0
    5  5  6.0
    
  • groupby will now not return the grouped column for non-cython functions (GH5610, GH5614, GH6732), as its already the index

    In [32]: df = DataFrame([[1, np.nan], [1, 4], [5, 6], [5, 8]], columns=['A', 'B'])
    
    In [33]: g = df.groupby('A')
    
    In [34]: g.count()
    Out[34]: 
       B
    A   
    1  1
    5  2
    
    In [35]: g.describe()
    Out[35]: 
          B                                        
      count mean       std  min  25%  50%  75%  max
    A                                              
    1   1.0  4.0       NaN  4.0  4.0  4.0  4.0  4.0
    5   2.0  7.0  1.414214  6.0  6.5  7.0  7.5  8.0
    
  • passing as_index will leave the grouped column in-place (this is not change in 0.14.0)

    In [36]: df = DataFrame([[1, np.nan], [1, 4], [5, 6], [5, 8]], columns=['A', 'B'])
    
    In [37]: g = df.groupby('A',as_index=False)
    
    In [38]: g.count()
    Out[38]: 
       A  B
    0  1  1
    1  5  2
    
    In [39]: g.describe()
    Out[39]: 
          A                                        B                                        
      count mean  std  min  25%  50%  75%  max count mean       std  min  25%  50%  75%  max
    0   2.0  1.0  0.0  1.0  1.0  1.0  1.0  1.0   1.0  4.0       NaN  4.0  4.0  4.0  4.0  4.0
    1   2.0  5.0  0.0  5.0  5.0  5.0  5.0  5.0   2.0  7.0  1.414214  6.0  6.5  7.0  7.5  8.0
    
  • Allow specification of a more complex groupby via pd.Grouper, such as grouping by a Time and a string field simultaneously. See the docs. (GH3794)

  • Better propagation/preservation of Series names when performing groupby operations:

    • SeriesGroupBy.agg will ensure that the name attribute of the original series is propagated to the result (GH6265).
    • If the function provided to GroupBy.apply returns a named series, the name of the series will be kept as the name of the column index of the DataFrame returned by GroupBy.apply (GH6124). This facilitates DataFrame.stack operations where the name of the column index is used as the name of the inserted column containing the pivoted data.

SQL

The SQL reading and writing functions now support more database flavors through SQLAlchemy (GH2717, GH4163, GH5950, GH6292). All databases supported by SQLAlchemy can be used, such as PostgreSQL, MySQL, Oracle, Microsoft SQL server (see documentation of SQLAlchemy on included dialects).

The functionality of providing DBAPI connection objects will only be supported for sqlite3 in the future. The 'mysql' flavor is deprecated.

The new functions read_sql_query() and read_sql_table() are introduced. The function read_sql() is kept as a convenience wrapper around the other two and will delegate to specific function depending on the provided input (database table name or sql query).

In practice, you have to provide a SQLAlchemy engine to the sql functions. To connect with SQLAlchemy you use the create_engine() function to create an engine object from database URI. You only need to create the engine once per database you are connecting to. For an in-memory sqlite database:

In [40]: from sqlalchemy import create_engine

# Create your connection.
In [41]: engine = create_engine('sqlite:///:memory:')

This engine can then be used to write or read data to/from this database:

In [42]: df = pd.DataFrame({'A': [1,2,3], 'B': ['a', 'b', 'c']})

In [43]: df.to_sql('db_table', engine, index=False)

You can read data from a database by specifying the table name:

In [44]: pd.read_sql_table('db_table', engine)
Out[44]: 
   A  B
0  1  a
1  2  b
2  3  c

or by specifying a sql query:

In [45]: pd.read_sql_query('SELECT * FROM db_table', engine)
Out[45]: 
   A  B
0  1  a
1  2  b
2  3  c

Some other enhancements to the sql functions include:

  • support for writing the index. This can be controlled with the index keyword (default is True).
  • specify the column label to use when writing the index with index_label.
  • specify string columns to parse as datetimes with the parse_dates keyword in read_sql_query() and read_sql_table().

Warning

Some of the existing functions or function aliases have been deprecated and will be removed in future versions. This includes: tquery, uquery, read_frame, frame_query, write_frame.

Warning

The support for the ‘mysql’ flavor when using DBAPI connection objects has been deprecated. MySQL will be further supported with SQLAlchemy engines (GH6900).

MultiIndexing Using Slicers

In 0.14.0 we added a new way to slice multi-indexed objects. You can slice a multi-index by providing multiple indexers.

You can provide any of the selectors as if you are indexing by label, see Selection by Label, including slices, lists of labels, labels, and boolean indexers.

You can use slice(None) to select all the contents of that level. You do not need to specify all the deeper levels, they will be implied as slice(None).

As usual, both sides of the slicers are included as this is label indexing.

See the docs See also issues (GH6134, GH4036, GH3057, GH2598, GH5641, GH7106)

Warning

You should specify all axes in the .loc specifier, meaning the indexer for the index and for the columns. Their are some ambiguous cases where the passed indexer could be mis-interpreted as indexing both axes, rather than into say the MuliIndex for the rows.

You should do this:

df.loc[(slice('A1','A3'),.....),:]

rather than this:

df.loc[(slice('A1','A3'),.....)]

Warning

You will need to make sure that the selection axes are fully lexsorted!

In [46]: def mklbl(prefix,n):
   ....:     return ["%s%s" % (prefix,i)  for i in range(n)]
   ....: 

In [47]: index = MultiIndex.from_product([mklbl('A',4),
   ....:                                  mklbl('B',2),
   ....:                                  mklbl('C',4),
   ....:                                  mklbl('D',2)])
   ....: 

In [48]: columns = MultiIndex.from_tuples([('a','foo'),('a','bar'),
   ....:                                   ('b','foo'),('b','bah')],
   ....:                                    names=['lvl0', 'lvl1'])
   ....: 

In [49]: df = DataFrame(np.arange(len(index)*len(columns)).reshape((len(index),len(columns))),
   ....:                index=index,
   ....:                columns=columns).sort_index().sort_index(axis=1)
   ....: 

In [50]: df
Out[50]: 
lvl0           a         b     
lvl1         bar  foo  bah  foo
A0 B0 C0 D0    1    0    3    2
         D1    5    4    7    6
      C1 D0    9    8   11   10
         D1   13   12   15   14
      C2 D0   17   16   19   18
         D1   21   20   23   22
      C3 D0   25   24   27   26
...          ...  ...  ...  ...
A3 B1 C0 D1  229  228  231  230
      C1 D0  233  232  235  234
         D1  237  236  239  238
      C2 D0  241  240  243  242
         D1  245  244  247  246
      C3 D0  249  248  251  250
         D1  253  252  255  254

[64 rows x 4 columns]

Basic multi-index slicing using slices, lists, and labels.

In [51]: df.loc[(slice('A1','A3'),slice(None), ['C1','C3']),:]
Out[51]: 
lvl0           a         b     
lvl1         bar  foo  bah  foo
A1 B0 C1 D0   73   72   75   74
         D1   77   76   79   78
      C3 D0   89   88   91   90
         D1   93   92   95   94
   B1 C1 D0  105  104  107  106
         D1  109  108  111  110
      C3 D0  121  120  123  122
...          ...  ...  ...  ...
A3 B0 C1 D1  205  204  207  206
      C3 D0  217  216  219  218
         D1  221  220  223  222
   B1 C1 D0  233  232  235  234
         D1  237  236  239  238
      C3 D0  249  248  251  250
         D1  253  252  255  254

[24 rows x 4 columns]

You can use a pd.IndexSlice to shortcut the creation of these slices

In [52]: idx = pd.IndexSlice

In [53]: df.loc[idx[:,:,['C1','C3']],idx[:,'foo']]
Out[53]: 
lvl0           a    b
lvl1         foo  foo
A0 B0 C1 D0    8   10
         D1   12   14
      C3 D0   24   26
         D1   28   30
   B1 C1 D0   40   42
         D1   44   46
      C3 D0   56   58
...          ...  ...
A3 B0 C1 D1  204  206
      C3 D0  216  218
         D1  220  222
   B1 C1 D0  232  234
         D1  236  238
      C3 D0  248  250
         D1  252  254

[32 rows x 2 columns]

It is possible to perform quite complicated selections using this method on multiple axes at the same time.

In [54]: df.loc['A1',(slice(None),'foo')]
Out[54]: 
lvl0        a    b
lvl1      foo  foo
B0 C0 D0   64   66
      D1   68   70
   C1 D0   72   74
      D1   76   78
   C2 D0   80   82
      D1   84   86
   C3 D0   88   90
...       ...  ...
B1 C0 D1  100  102
   C1 D0  104  106
      D1  108  110
   C2 D0  112  114
      D1  116  118
   C3 D0  120  122
      D1  124  126

[16 rows x 2 columns]

In [55]: df.loc[idx[:,:,['C1','C3']],idx[:,'foo']]
Out[55]: 
lvl0           a    b
lvl1         foo  foo
A0 B0 C1 D0    8   10
         D1   12   14
      C3 D0   24   26
         D1   28   30
   B1 C1 D0   40   42
         D1   44   46
      C3 D0   56   58
...          ...  ...
A3 B0 C1 D1  204  206
      C3 D0  216  218
         D1  220  222
   B1 C1 D0  232  234
         D1  236  238
      C3 D0  248  250
         D1  252  254

[32 rows x 2 columns]

Using a boolean indexer you can provide selection related to the values.

In [56]: mask = df[('a','foo')]>200

In [57]: df.loc[idx[mask,:,['C1','C3']],idx[:,'foo']]
Out[57]: 
lvl0           a    b
lvl1         foo  foo
A3 B0 C1 D1  204  206
      C3 D0  216  218
         D1  220  222
   B1 C1 D0  232  234
         D1  236  238
      C3 D0  248  250
         D1  252  254

You can also specify the axis argument to .loc to interpret the passed slicers on a single axis.

In [58]: df.loc(axis=0)[:,:,['C1','C3']]
Out[58]: 
lvl0           a         b     
lvl1         bar  foo  bah  foo
A0 B0 C1 D0    9    8   11   10
         D1   13   12   15   14
      C3 D0   25   24   27   26
         D1   29   28   31   30
   B1 C1 D0   41   40   43   42
         D1   45   44   47   46
      C3 D0   57   56   59   58
...          ...  ...  ...  ...
A3 B0 C1 D1  205  204  207  206
      C3 D0  217  216  219  218
         D1  221  220  223  222
   B1 C1 D0  233  232  235  234
         D1  237  236  239  238
      C3 D0  249  248  251  250
         D1  253  252  255  254

[32 rows x 4 columns]

Furthermore you can set the values using these methods

In [59]: df2 = df.copy()

In [60]: df2.loc(axis=0)[:,:,['C1','C3']] = -10

In [61]: df2
Out[61]: 
lvl0           a         b     
lvl1         bar  foo  bah  foo
A0 B0 C0 D0    1    0    3    2
         D1    5    4    7    6
      C1 D0  -10  -10  -10  -10
         D1  -10  -10  -10  -10
      C2 D0   17   16   19   18
         D1   21   20   23   22
      C3 D0  -10  -10  -10  -10
...          ...  ...  ...  ...
A3 B1 C0 D1  229  228  231  230
      C1 D0  -10  -10  -10  -10
         D1  -10  -10  -10  -10
      C2 D0  241  240  243  242
         D1  245  244  247  246
      C3 D0  -10  -10  -10  -10
         D1  -10  -10  -10  -10

[64 rows x 4 columns]

You can use a right-hand-side of an alignable object as well.

In [62]: df2 = df.copy()

In [63]: df2.loc[idx[:,:,['C1','C3']],:] = df2*1000

In [64]: df2
Out[64]: 
lvl0              a               b        
lvl1            bar     foo     bah     foo
A0 B0 C0 D0       1       0       3       2
         D1       5       4       7       6
      C1 D0    9000    8000   11000   10000
         D1   13000   12000   15000   14000
      C2 D0      17      16      19      18
         D1      21      20      23      22
      C3 D0   25000   24000   27000   26000
...             ...     ...     ...     ...
A3 B1 C0 D1     229     228     231     230
      C1 D0  233000  232000  235000  234000
         D1  237000  236000  239000  238000
      C2 D0     241     240     243     242
         D1     245     244     247     246
      C3 D0  249000  248000  251000  250000
         D1  253000  252000  255000  254000

[64 rows x 4 columns]

Plotting

  • Hexagonal bin plots from DataFrame.plot with kind='hexbin' (GH5478), See the docs.

  • DataFrame.plot and Series.plot now supports area plot with specifying kind='area' (GH6656), See the docs

  • Pie plots from Series.plot and DataFrame.plot with kind='pie' (GH6976), See the docs.

  • Plotting with Error Bars is now supported in the .plot method of DataFrame and Series objects (GH3796, GH6834), See the docs.

  • DataFrame.plot and Series.plot now support a table keyword for plotting matplotlib.Table, See the docs. The table keyword can receive the following values.

    • False: Do nothing (default).
    • True: Draw a table using the DataFrame or Series called plot method. Data will be transposed to meet matplotlib’s default layout.
    • DataFrame or Series: Draw matplotlib.table using the passed data. The data will be drawn as displayed in print method (not transposed automatically). Also, helper function pandas.tools.plotting.table is added to create a table from DataFrame and Series, and add it to an matplotlib.Axes.
  • plot(legend='reverse') will now reverse the order of legend labels for most plot kinds. (GH6014)

  • Line plot and area plot can be stacked by stacked=True (GH6656)

  • Following keywords are now acceptable for DataFrame.plot() with kind='bar' and kind='barh':

    • width: Specify the bar width. In previous versions, static value 0.5 was passed to matplotlib and it cannot be overwritten. (GH6604)
    • align: Specify the bar alignment. Default is center (different from matplotlib). In previous versions, pandas passes align=’edge’ to matplotlib and adjust the location to center by itself, and it results align keyword is not applied as expected. (GH4525)
    • position: Specify relative alignments for bar plot layout. From 0 (left/bottom-end) to 1(right/top-end). Default is 0.5 (center). (GH6604)

    Because of the default align value changes, coordinates of bar plots are now located on integer values (0.0, 1.0, 2.0 …). This is intended to make bar plot be located on the same coordinates as line plot. However, bar plot may differs unexpectedly when you manually adjust the bar location or drawing area, such as using set_xlim, set_ylim, etc. In this cases, please modify your script to meet with new coordinates.

  • The parallel_coordinates() function now takes argument color instead of colors. A FutureWarning is raised to alert that the old colors argument will not be supported in a future release. (GH6956)

  • The parallel_coordinates() and andrews_curves() functions now take positional argument frame instead of data. A FutureWarning is raised if the old data argument is used by name. (GH6956)

  • DataFrame.boxplot() now supports layout keyword (GH6769)

  • DataFrame.boxplot() has a new keyword argument, return_type. It accepts 'dict', 'axes', or 'both', in which case a namedtuple with the matplotlib axes and a dict of matplotlib Lines is returned.

Prior Version Deprecations/Changes

There are prior version deprecations that are taking effect as of 0.14.0.

  • Remove DateRange in favor of DatetimeIndex (GH6816)
  • Remove column keyword from DataFrame.sort (GH4370)
  • Remove precision keyword from set_eng_float_format() (GH395)
  • Remove force_unicode keyword from DataFrame.to_string(), DataFrame.to_latex(), and DataFrame.to_html(); these function encode in unicode by default (GH2224, GH2225)
  • Remove nanRep keyword from DataFrame.to_csv() and DataFrame.to_string() (GH275)
  • Remove unique keyword from HDFStore.select_column() (GH3256)
  • Remove inferTimeRule keyword from Timestamp.offset() (GH391)
  • Remove name keyword from get_data_yahoo() and get_data_google() ( commit b921d1a )
  • Remove offset keyword from DatetimeIndex constructor ( commit 3136390 )
  • Remove time_rule from several rolling-moment statistical functions, such as rolling_sum() (GH1042)
  • Removed neg - boolean operations on numpy arrays in favor of inv ~, as this is going to be deprecated in numpy 1.9 (GH6960)

Deprecations

  • The pivot_table()/DataFrame.pivot_table() and crosstab() functions now take arguments index and columns instead of rows and cols. A FutureWarning is raised to alert that the old rows and cols arguments will not be supported in a future release (GH5505)

  • The DataFrame.drop_duplicates() and DataFrame.duplicated() methods now take argument subset instead of cols to better align with DataFrame.dropna(). A FutureWarning is raised to alert that the old cols arguments will not be supported in a future release (GH6680)

  • The DataFrame.to_csv() and DataFrame.to_excel() functions now takes argument columns instead of cols. A FutureWarning is raised to alert that the old cols arguments will not be supported in a future release (GH6645)

  • Indexers will warn FutureWarning when used with a scalar indexer and a non-floating point Index (GH4892, GH6960)

    # non-floating point indexes can only be indexed by integers / labels
    In [1]: Series(1,np.arange(5))[3.0]
            pandas/core/index.py:469: FutureWarning: scalar indexers for index type Int64Index should be integers and not floating point
    Out[1]: 1
    
    In [2]: Series(1,np.arange(5)).iloc[3.0]
            pandas/core/index.py:469: FutureWarning: scalar indexers for index type Int64Index should be integers and not floating point
    Out[2]: 1
    
    In [3]: Series(1,np.arange(5)).iloc[3.0:4]
            pandas/core/index.py:527: FutureWarning: slice indexers when using iloc should be integers and not floating point
    Out[3]:
            3    1
            dtype: int64
    
    # these are Float64Indexes, so integer or floating point is acceptable
    In [4]: Series(1,np.arange(5.))[3]
    Out[4]: 1
    
    In [5]: Series(1,np.arange(5.))[3.0]
    Out[6]: 1
    
  • Numpy 1.9 compat w.r.t. deprecation warnings (GH6960)

  • Panel.shift() now has a function signature that matches DataFrame.shift(). The old positional argument lags has been changed to a keyword argument periods with a default value of 1. A FutureWarning is raised if the old argument lags is used by name. (GH6910)

  • The order keyword argument of factorize() will be removed. (GH6926).

  • Remove the copy keyword from DataFrame.xs(), Panel.major_xs(), Panel.minor_xs(). A view will be returned if possible, otherwise a copy will be made. Previously the user could think that copy=False would ALWAYS return a view. (GH6894)

  • The parallel_coordinates() function now takes argument color instead of colors. A FutureWarning is raised to alert that the old colors argument will not be supported in a future release. (GH6956)

  • The parallel_coordinates() and andrews_curves() functions now take positional argument frame instead of data. A FutureWarning is raised if the old data argument is used by name. (GH6956)

  • The support for the ‘mysql’ flavor when using DBAPI connection objects has been deprecated. MySQL will be further supported with SQLAlchemy engines (GH6900).

  • The following io.sql functions have been deprecated: tquery, uquery, read_frame, frame_query, write_frame.

  • The percentile_width keyword argument in describe() has been deprecated. Use the percentiles keyword instead, which takes a list of percentiles to display. The default output is unchanged.

  • The default return type of boxplot() will change from a dict to a matpltolib Axes in a future release. You can use the future behavior now by passing return_type='axes' to boxplot.

Known Issues

  • OpenPyXL 2.0.0 breaks backwards compatibility (GH7169)

Enhancements

  • DataFrame and Series will create a MultiIndex object if passed a tuples dict, See the docs (GH3323)

    In [65]: Series({('a', 'b'): 1, ('a', 'a'): 0,
       ....:         ('a', 'c'): 2, ('b', 'a'): 3, ('b', 'b'): 4})
       ....: 
    Out[65]: 
    a  b    1
       a    0
       c    2
    b  a    3
       b    4
    dtype: int64
    
    In [66]: DataFrame({('a', 'b'): {('A', 'B'): 1, ('A', 'C'): 2},
       ....:            ('a', 'a'): {('A', 'C'): 3, ('A', 'B'): 4},
       ....:            ('a', 'c'): {('A', 'B'): 5, ('A', 'C'): 6},
       ....:            ('b', 'a'): {('A', 'C'): 7, ('A', 'B'): 8},
       ....:            ('b', 'b'): {('A', 'D'): 9, ('A', 'B'): 10}})
       ....: 
    Out[66]: 
           a              b      
           b    a    c    a     b
    A B  1.0  4.0  5.0  8.0  10.0
      C  2.0  3.0  6.0  7.0   NaN
      D  NaN  NaN  NaN  NaN   9.0
    
  • Added the sym_diff method to Index (GH5543)

  • DataFrame.to_latex now takes a longtable keyword, which if True will return a table in a longtable environment. (GH6617)

  • Add option to turn off escaping in DataFrame.to_latex (GH6472)

  • pd.read_clipboard will, if the keyword sep is unspecified, try to detect data copied from a spreadsheet and parse accordingly. (GH6223)

  • Joining a singly-indexed DataFrame with a multi-indexed DataFrame (GH3662)

    See the docs. Joining multi-index DataFrames on both the left and right is not yet supported ATM.

    In [67]: household = DataFrame(dict(household_id = [1,2,3],
       ....:                            male = [0,1,0],
       ....:                            wealth = [196087.3,316478.7,294750]),
       ....:                       columns = ['household_id','male','wealth']
       ....:                      ).set_index('household_id')
       ....: 
    
    In [68]: household
    Out[68]: 
                  male    wealth
    household_id                
    1                0  196087.3
    2                1  316478.7
    3                0  294750.0
    
    In [69]: portfolio = DataFrame(dict(household_id = [1,2,2,3,3,3,4],
       ....:                            asset_id = ["nl0000301109","nl0000289783","gb00b03mlx29",
       ....:                                        "gb00b03mlx29","lu0197800237","nl0000289965",np.nan],
       ....:                            name = ["ABN Amro","Robeco","Royal Dutch Shell","Royal Dutch Shell",
       ....:                                    "AAB Eastern Europe Equity Fund","Postbank BioTech Fonds",np.nan],
       ....:                            share = [1.0,0.4,0.6,0.15,0.6,0.25,1.0]),
       ....:                       columns = ['household_id','asset_id','name','share']
       ....:                      ).set_index(['household_id','asset_id'])
       ....: 
    
    In [70]: portfolio
    Out[70]: 
                                                         name  share
    household_id asset_id                                           
    1            nl0000301109                        ABN Amro   1.00
    2            nl0000289783                          Robeco   0.40
                 gb00b03mlx29               Royal Dutch Shell   0.60
    3            gb00b03mlx29               Royal Dutch Shell   0.15
                 lu0197800237  AAB Eastern Europe Equity Fund   0.60
                 nl0000289965          Postbank BioTech Fonds   0.25
    4            NaN                                      NaN   1.00
    
    In [71]: household.join(portfolio, how='inner')
    Out[71]: 
                               male    wealth                            name  share
    household_id asset_id                                                           
    1            nl0000301109     0  196087.3                        ABN Amro   1.00
    2            nl0000289783     1  316478.7                          Robeco   0.40
                 gb00b03mlx29     1  316478.7               Royal Dutch Shell   0.60
    3            gb00b03mlx29     0  294750.0               Royal Dutch Shell   0.15
                 lu0197800237     0  294750.0  AAB Eastern Europe Equity Fund   0.60
                 nl0000289965     0  294750.0          Postbank BioTech Fonds   0.25
    
  • quotechar, doublequote, and escapechar can now be specified when using DataFrame.to_csv (GH5414, GH4528)

  • Partially sort by only the specified levels of a MultiIndex with the sort_remaining boolean kwarg. (GH3984)

  • Added to_julian_date to TimeStamp and DatetimeIndex. The Julian Date is used primarily in astronomy and represents the number of days from noon, January 1, 4713 BC. Because nanoseconds are used to define the time in pandas the actual range of dates that you can use is 1678 AD to 2262 AD. (GH4041)

  • DataFrame.to_stata will now check data for compatibility with Stata data types and will upcast when needed. When it is not possible to losslessly upcast, a warning is issued (GH6327)

  • DataFrame.to_stata and StataWriter will accept keyword arguments time_stamp and data_label which allow the time stamp and dataset label to be set when creating a file. (GH6545)

  • pandas.io.gbq now handles reading unicode strings properly. (GH5940)

  • Holidays Calendars are now available and can be used with the CustomBusinessDay offset (GH6719)

  • Float64Index is now backed by a float64 dtype ndarray instead of an object dtype array (GH6471).

  • Implemented Panel.pct_change (GH6904)

  • Added how option to rolling-moment functions to dictate how to handle resampling; rolling_max() defaults to max, rolling_min() defaults to min, and all others default to mean (GH6297)

  • CustomBuisnessMonthBegin and CustomBusinessMonthEnd are now available (GH6866)

  • Series.quantile() and DataFrame.quantile() now accept an array of quantiles.

  • describe() now accepts an array of percentiles to include in the summary statistics (GH4196)

  • pivot_table can now accept Grouper by index and columns keywords (GH6913)

    In [72]: import datetime
    
    In [73]: df = DataFrame({
       ....:   'Branch' : 'A A A A A B'.split(),
       ....:   'Buyer': 'Carl Mark Carl Carl Joe Joe'.split(),
       ....:   'Quantity': [1, 3, 5, 1, 8, 1],
       ....:   'Date' : [datetime.datetime(2013,11,1,13,0), datetime.datetime(2013,9,1,13,5),
       ....:             datetime.datetime(2013,10,1,20,0), datetime.datetime(2013,10,2,10,0),
       ....:             datetime.datetime(2013,11,1,20,0), datetime.datetime(2013,10,2,10,0)],
       ....:   'PayDay' : [datetime.datetime(2013,10,4,0,0), datetime.datetime(2013,10,15,13,5),
       ....:               datetime.datetime(2013,9,5,20,0), datetime.datetime(2013,11,2,10,0),
       ....:               datetime.datetime(2013,10,7,20,0), datetime.datetime(2013,9,5,10,0)]})
       ....: 
    
    In [74]: df
    Out[74]: 
      Branch Buyer  Quantity                Date              PayDay
    0      A  Carl         1 2013-11-01 13:00:00 2013-10-04 00:00:00
    1      A  Mark         3 2013-09-01 13:05:00 2013-10-15 13:05:00
    2      A  Carl         5 2013-10-01 20:00:00 2013-09-05 20:00:00
    3      A  Carl         1 2013-10-02 10:00:00 2013-11-02 10:00:00
    4      A   Joe         8 2013-11-01 20:00:00 2013-10-07 20:00:00
    5      B   Joe         1 2013-10-02 10:00:00 2013-09-05 10:00:00
    
    In [75]: pivot_table(df, index=Grouper(freq='M', key='Date'),
       ....:             columns=Grouper(freq='M', key='PayDay'),
       ....:             values='Quantity', aggfunc=np.sum)
       ....: 
    Out[75]: 
    PayDay      2013-09-30  2013-10-31  2013-11-30
    Date                                          
    2013-09-30         NaN         3.0         NaN
    2013-10-31         6.0         NaN         1.0
    2013-11-30         NaN         9.0         NaN
    
  • Arrays of strings can be wrapped to a specified width (str.wrap) (GH6999)

  • Add nsmallest() and Series.nlargest() methods to Series, See the docs (GH3960)

  • PeriodIndex fully supports partial string indexing like DatetimeIndex (GH7043)

    In [76]: prng = period_range('2013-01-01 09:00', periods=100, freq='H')
    
    In [77]: ps = Series(np.random.randn(len(prng)), index=prng)
    
    In [78]: ps
    Out[78]: 
    2013-01-01 09:00    0.015696
    2013-01-01 10:00   -2.242685
    2013-01-01 11:00    1.150036
    2013-01-01 12:00    0.991946
    2013-01-01 13:00    0.953324
    2013-01-01 14:00   -2.021255
    2013-01-01 15:00   -0.334077
                          ...   
    2013-01-05 06:00    0.566534
    2013-01-05 07:00    0.503592
    2013-01-05 08:00    0.285296
    2013-01-05 09:00    0.484288
    2013-01-05 10:00    1.363482
    2013-01-05 11:00   -0.781105
    2013-01-05 12:00   -0.468018
    Freq: H, Length: 100, dtype: float64
    
    In [79]: ps['2013-01-02']
    Out[79]: 
    2013-01-02 00:00    0.553439
    2013-01-02 01:00    1.318152
    2013-01-02 02:00   -0.469305
    2013-01-02 03:00    0.675554
    2013-01-02 04:00   -1.817027
    2013-01-02 05:00   -0.183109
    2013-01-02 06:00    1.058969
                          ...   
    2013-01-02 17:00    0.076200
    2013-01-02 18:00   -0.566446
    2013-01-02 19:00    0.036142
    2013-01-02 20:00   -2.074978
    2013-01-02 21:00    0.247792
    2013-01-02 22:00   -0.897157
    2013-01-02 23:00   -0.136795
    Freq: H, Length: 24, dtype: float64
    
  • read_excel can now read milliseconds in Excel dates and times with xlrd >= 0.9.3. (GH5945)

  • pd.stats.moments.rolling_var now uses Welford’s method for increased numerical stability (GH6817)

  • pd.expanding_apply and pd.rolling_apply now take args and kwargs that are passed on to the func (GH6289)

  • DataFrame.rank() now has a percentage rank option (GH5971)

  • Series.rank() now has a percentage rank option (GH5971)

  • Series.rank() and DataFrame.rank() now accept method='dense' for ranks without gaps (GH6514)

  • Support passing encoding with xlwt (GH3710)

  • Refactor Block classes removing Block.items attributes to avoid duplication in item handling (GH6745, GH6988).

  • Testing statements updated to use specialized asserts (GH6175)

Performance

  • Performance improvement when converting DatetimeIndex to floating ordinals using DatetimeConverter (GH6636)
  • Performance improvement for DataFrame.shift (GH5609)
  • Performance improvement in indexing into a multi-indexed Series (GH5567)
  • Performance improvements in single-dtyped indexing (GH6484)
  • Improve performance of DataFrame construction with certain offsets, by removing faulty caching (e.g. MonthEnd,BusinessMonthEnd), (GH6479)
  • Improve performance of CustomBusinessDay (GH6584)
  • improve performance of slice indexing on Series with string keys (GH6341, GH6372)
  • Performance improvement for DataFrame.from_records when reading a specified number of rows from an iterable (GH6700)
  • Performance improvements in timedelta conversions for integer dtypes (GH6754)
  • Improved performance of compatible pickles (GH6899)
  • Improve performance in certain reindexing operations by optimizing take_2d (GH6749)
  • GroupBy.count() is now implemented in Cython and is much faster for large numbers of groups (GH7016).

Experimental

There are no experimental changes in 0.14.0

Bug Fixes

  • Bug in Series ValueError when index doesn’t match data (GH6532)
  • Prevent segfault due to MultiIndex not being supported in HDFStore table format (GH1848)
  • Bug in pd.DataFrame.sort_index where mergesort wasn’t stable when ascending=False (GH6399)
  • Bug in pd.tseries.frequencies.to_offset when argument has leading zeroes (GH6391)
  • Bug in version string gen. for dev versions with shallow clones / install from tarball (GH6127)
  • Inconsistent tz parsing Timestamp / to_datetime for current year (GH5958)
  • Indexing bugs with reordered indexes (GH6252, GH6254)
  • Bug in .xs with a Series multiindex (GH6258, GH5684)
  • Bug in conversion of a string types to a DatetimeIndex with a specified frequency (GH6273, GH6274)
  • Bug in eval where type-promotion failed for large expressions (GH6205)
  • Bug in interpolate with inplace=True (GH6281)
  • HDFStore.remove now handles start and stop (GH6177)
  • HDFStore.select_as_multiple handles start and stop the same way as select (GH6177)
  • HDFStore.select_as_coordinates and select_column works with a where clause that results in filters (GH6177)
  • Regression in join of non_unique_indexes (GH6329)
  • Issue with groupby agg with a single function and a a mixed-type frame (GH6337)
  • Bug in DataFrame.replace() when passing a non- bool to_replace argument (GH6332)
  • Raise when trying to align on different levels of a multi-index assignment (GH3738)
  • Bug in setting complex dtypes via boolean indexing (GH6345)
  • Bug in TimeGrouper/resample when presented with a non-monotonic DatetimeIndex that would return invalid results. (GH4161)
  • Bug in index name propagation in TimeGrouper/resample (GH4161)
  • TimeGrouper has a more compatible API to the rest of the groupers (e.g. groups was missing) (GH3881)
  • Bug in multiple grouping with a TimeGrouper depending on target column order (GH6764)
  • Bug in pd.eval when parsing strings with possible tokens like '&' (GH6351)
  • Bug correctly handle placements of -inf in Panels when dividing by integer 0 (GH6178)
  • DataFrame.shift with axis=1 was raising (GH6371)
  • Disabled clipboard tests until release time (run locally with nosetests -A disabled) (GH6048).
  • Bug in DataFrame.replace() when passing a nested dict that contained keys not in the values to be replaced (GH6342)
  • str.match ignored the na flag (GH6609).
  • Bug in take with duplicate columns that were not consolidated (GH6240)
  • Bug in interpolate changing dtypes (GH6290)
  • Bug in Series.get, was using a buggy access method (GH6383)
  • Bug in hdfstore queries of the form where=[('date', '>=', datetime(2013,1,1)), ('date', '<=', datetime(2014,1,1))] (GH6313)
  • Bug in DataFrame.dropna with duplicate indices (GH6355)
  • Regression in chained getitem indexing with embedded list-like from 0.12 (GH6394)
  • Float64Index with nans not comparing correctly (GH6401)
  • eval/query expressions with strings containing the @ character will now work (GH6366).
  • Bug in Series.reindex when specifying a method with some nan values was inconsistent (noted on a resample) (GH6418)
  • Bug in DataFrame.replace() where nested dicts were erroneously depending on the order of dictionary keys and values (GH5338).
  • Perf issue in concatting with empty objects (GH3259)
  • Clarify sorting of sym_diff on Index objects with NaN values (GH6444)
  • Regression in MultiIndex.from_product with a DatetimeIndex as input (GH6439)
  • Bug in str.extract when passed a non-default index (GH6348)
  • Bug in str.split when passed pat=None and n=1 (GH6466)
  • Bug in io.data.DataReader when passed "F-F_Momentum_Factor" and data_source="famafrench" (GH6460)
  • Bug in sum of a timedelta64[ns] series (GH6462)
  • Bug in resample with a timezone and certain offsets (GH6397)
  • Bug in iat/iloc with duplicate indices on a Series (GH6493)
  • Bug in read_html where nan’s were incorrectly being used to indicate missing values in text. Should use the empty string for consistency with the rest of pandas (GH5129).
  • Bug in read_html tests where redirected invalid URLs would make one test fail (GH6445).
  • Bug in multi-axis indexing using .loc on non-unique indices (GH6504)
  • Bug that caused _ref_locs corruption when slice indexing across columns axis of a DataFrame (GH6525)
  • Regression from 0.13 in the treatment of numpy datetime64 non-ns dtypes in Series creation (GH6529)
  • .names attribute of MultiIndexes passed to set_index are now preserved (GH6459).
  • Bug in setitem with a duplicate index and an alignable rhs (GH6541)
  • Bug in setitem with .loc on mixed integer Indexes (GH6546)
  • Bug in pd.read_stata which would use the wrong data types and missing values (GH6327)
  • Bug in DataFrame.to_stata that lead to data loss in certain cases, and could be exported using the wrong data types and missing values (GH6335)
  • StataWriter replaces missing values in string columns by empty string (GH6802)
  • Inconsistent types in Timestamp addition/subtraction (GH6543)
  • Bug in preserving frequency across Timestamp addition/subtraction (GH4547)
  • Bug in empty list lookup caused IndexError exceptions (GH6536, GH6551)
  • Series.quantile raising on an object dtype (GH6555)
  • Bug in .xs with a nan in level when dropped (GH6574)
  • Bug in fillna with method='bfill/ffill' and datetime64[ns] dtype (GH6587)
  • Bug in sql writing with mixed dtypes possibly leading to data loss (GH6509)
  • Bug in Series.pop (GH6600)
  • Bug in iloc indexing when positional indexer matched Int64Index of the corresponding axis and no reordering happened (GH6612)
  • Bug in fillna with limit and value specified
  • Bug in DataFrame.to_stata when columns have non-string names (GH4558)
  • Bug in compat with np.compress, surfaced in (GH6658)
  • Bug in binary operations with a rhs of a Series not aligning (GH6681)
  • Bug in DataFrame.to_stata which incorrectly handles nan values and ignores with_index keyword argument (GH6685)
  • Bug in resample with extra bins when using an evenly divisible frequency (GH4076)
  • Bug in consistency of groupby aggregation when passing a custom function (GH6715)
  • Bug in resample when how=None resample freq is the same as the axis frequency (GH5955)
  • Bug in downcasting inference with empty arrays (GH6733)
  • Bug in obj.blocks on sparse containers dropping all but the last items of same for dtype (GH6748)
  • Bug in unpickling NaT (NaTType) (GH4606)
  • Bug in DataFrame.replace() where regex metacharacters were being treated as regexs even when regex=False (GH6777).
  • Bug in timedelta ops on 32-bit platforms (GH6808)
  • Bug in setting a tz-aware index directly via .index (GH6785)
  • Bug in expressions.py where numexpr would try to evaluate arithmetic ops (GH6762).
  • Bug in Makefile where it didn’t remove Cython generated C files with make clean (GH6768)
  • Bug with numpy < 1.7.2 when reading long strings from HDFStore (GH6166)
  • Bug in DataFrame._reduce where non bool-like (0/1) integers were being converted into bools. (GH6806)
  • Regression from 0.13 with fillna and a Series on datetime-like (GH6344)
  • Bug in adding np.timedelta64 to DatetimeIndex with timezone outputs incorrect results (GH6818)
  • Bug in DataFrame.replace() where changing a dtype through replacement would only replace the first occurrence of a value (GH6689)
  • Better error message when passing a frequency of ‘MS’ in Period construction (GH5332)
  • Bug in Series.__unicode__ when max_rows=None and the Series has more than 1000 rows. (GH6863)
  • Bug in groupby.get_group where a datetlike wasn’t always accepted (GH5267)
  • Bug in groupBy.get_group created by TimeGrouper raises AttributeError (GH6914)
  • Bug in DatetimeIndex.tz_localize and DatetimeIndex.tz_convert converting NaT incorrectly (GH5546)
  • Bug in arithmetic operations affecting NaT (GH6873)
  • Bug in Series.str.extract where the resulting Series from a single group match wasn’t renamed to the group name
  • Bug in DataFrame.to_csv where setting index=False ignored the header kwarg (GH6186)
  • Bug in DataFrame.plot and Series.plot, where the legend behave inconsistently when plotting to the same axes repeatedly (GH6678)
  • Internal tests for patching __finalize__ / bug in merge not finalizing (GH6923, GH6927)
  • accept TextFileReader in concat, which was affecting a common user idiom (GH6583)
  • Bug in C parser with leading whitespace (GH3374)
  • Bug in C parser with delim_whitespace=True and \r-delimited lines
  • Bug in python parser with explicit multi-index in row following column header (GH6893)
  • Bug in Series.rank and DataFrame.rank that caused small floats (<1e-13) to all receive the same rank (GH6886)
  • Bug in DataFrame.apply with functions that used *args or **kwargs and returned an empty result (GH6952)
  • Bug in sum/mean on 32-bit platforms on overflows (GH6915)
  • Moved Panel.shift to NDFrame.slice_shift and fixed to respect multiple dtypes. (GH6959)
  • Bug in enabling subplots=True in DataFrame.plot only has single column raises TypeError, and Series.plot raises AttributeError (GH6951)
  • Bug in DataFrame.plot draws unnecessary axes when enabling subplots and kind=scatter (GH6951)
  • Bug in read_csv from a filesystem with non-utf-8 encoding (GH6807)
  • Bug in iloc when setting / aligning (GH6766)
  • Bug causing UnicodeEncodeError when get_dummies called with unicode values and a prefix (GH6885)
  • Bug in timeseries-with-frequency plot cursor display (GH5453)
  • Bug surfaced in groupby.plot when using a Float64Index (GH7025)
  • Stopped tests from failing if options data isn’t able to be downloaded from Yahoo (GH7034)
  • Bug in parallel_coordinates and radviz where reordering of class column caused possible color/class mismatch (GH6956)
  • Bug in radviz and andrews_curves where multiple values of ‘color’ were being passed to plotting method (GH6956)
  • Bug in Float64Index.isin() where containing nan s would make indices claim that they contained all the things (GH7066).
  • Bug in DataFrame.boxplot where it failed to use the axis passed as the ax argument (GH3578)
  • Bug in the XlsxWriter and XlwtWriter implementations that resulted in datetime columns being formatted without the time (GH7075) were being passed to plotting method
  • read_fwf() treats None in colspec like regular python slices. It now reads from the beginning or until the end of the line when colspec contains a None (previously raised a TypeError)
  • Bug in cache coherence with chained indexing and slicing; add _is_view property to NDFrame to correctly predict views; mark is_copy on xs only if its an actual copy (and not a view) (GH7084)
  • Bug in DatetimeIndex creation from string ndarray with dayfirst=True (GH5917)
  • Bug in MultiIndex.from_arrays created from DatetimeIndex doesn’t preserve freq and tz (GH7090)
  • Bug in unstack raises ValueError when MultiIndex contains PeriodIndex (GH4342)
  • Bug in boxplot and hist draws unnecessary axes (GH6769)
  • Regression in groupby.nth() for out-of-bounds indexers (GH6621)
  • Bug in quantile with datetime values (GH6965)
  • Bug in Dataframe.set_index, reindex and pivot don’t preserve DatetimeIndex and PeriodIndex attributes (GH3950, GH5878, GH6631)
  • Bug in MultiIndex.get_level_values doesn’t preserve DatetimeIndex and PeriodIndex attributes (GH7092)
  • Bug in Groupby doesn’t preserve tz (GH3950)
  • Bug in PeriodIndex partial string slicing (GH6716)
  • Bug in the HTML repr of a truncated Series or DataFrame not showing the class name with the large_repr set to ‘info’ (GH7105)
  • Bug in DatetimeIndex specifying freq raises ValueError when passed value is too short (GH7098)
  • Fixed a bug with the info repr not honoring the display.max_info_columns setting (GH6939)
  • Bug PeriodIndex string slicing with out of bounds values (GH5407)
  • Fixed a memory error in the hashtable implementation/factorizer on resizing of large tables (GH7157)
  • Bug in isnull when applied to 0-dimensional object arrays (GH7176)
  • Bug in query/eval where global constants were not looked up correctly (GH7178)
  • Bug in recognizing out-of-bounds positional list indexers with iloc and a multi-axis tuple indexer (GH7189)
  • Bug in setitem with a single value, multi-index and integer indices (GH7190, GH7218)
  • Bug in expressions evaluation with reversed ops, showing in series-dataframe ops (GH7198, GH7192)
  • Bug in multi-axis indexing with > 2 ndim and a multi-index (GH7199)
  • Fix a bug where invalid eval/query operations would blow the stack (GH5198)

v0.13.1 (February 3, 2014)

This is a minor release from 0.13.0 and includes a small number of API changes, several new features, enhancements, and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this version.

Highlights include:

  • Added infer_datetime_format keyword to read_csv/to_datetime to allow speedups for homogeneously formatted datetimes.
  • Will intelligently limit display precision for datetime/timedelta formats.
  • Enhanced Panel apply() method.
  • Suggested tutorials in new Tutorials section.
  • Our pandas ecosystem is growing, We now feature related projects in a new Pandas Ecosystem section.
  • Much work has been taking place on improving the docs, and a new Contributing section has been added.
  • Even though it may only be of interest to devs, we <3 our new CI status page: ScatterCI.

Warning

0.13.1 fixes a bug that was caused by a combination of having numpy < 1.8, and doing chained assignment on a string-like array. Please review the docs, chained indexing can have unexpected results and should generally be avoided.

This would previously segfault:

In [1]: df = DataFrame(dict(A = np.array(['foo','bar','bah','foo','bar'])))

In [2]: df['A'].iloc[0] = np.nan

In [3]: df
Out[3]: 
     A
0  NaN
1  bar
2  bah
3  foo
4  bar

The recommended way to do this type of assignment is:

In [4]: df = DataFrame(dict(A = np.array(['foo','bar','bah','foo','bar'])))

In [5]: df.loc[0,'A'] = np.nan

In [6]: df
Out[6]: 
     A
0  NaN
1  bar
2  bah
3  foo
4  bar

Output Formatting Enhancements

  • df.info() view now display dtype info per column (GH5682)

  • df.info() now honors the option max_info_rows, to disable null counts for large frames (GH5974)

    In [7]: max_info_rows = pd.get_option('max_info_rows')
    
    In [8]: df = DataFrame(dict(A = np.random.randn(10),
       ...:                     B = np.random.randn(10),
       ...:                     C = date_range('20130101',periods=10)))
       ...: 
    
    In [9]: df.iloc[3:6,[0,2]] = np.nan
    
    # set to not display the null counts
    In [10]: pd.set_option('max_info_rows',0)
    
    In [11]: df.info()
    <class 'pandas.core.frame.DataFrame'>
    RangeIndex: 10 entries, 0 to 9
    Data columns (total 3 columns):
    A    float64
    B    float64
    C    datetime64[ns]
    dtypes: datetime64[ns](1), float64(2)
    memory usage: 320.0 bytes
    
    # this is the default (same as in 0.13.0)
    In [12]: pd.set_option('max_info_rows',max_info_rows)
    
    In [13]: df.info()
    <class 'pandas.core.frame.DataFrame'>
    RangeIndex: 10 entries, 0 to 9
    Data columns (total 3 columns):
    A    7 non-null float64
    B    10 non-null float64
    C    7 non-null datetime64[ns]
    dtypes: datetime64[ns](1), float64(2)
    memory usage: 320.0 bytes
    
  • Add show_dimensions display option for the new DataFrame repr to control whether the dimensions print.

    In [14]: df = DataFrame([[1, 2], [3, 4]])
    
    In [15]: pd.set_option('show_dimensions', False)
    
    In [16]: df
    Out[16]: 
       0  1
    0  1  2
    1  3  4
    
    In [17]: pd.set_option('show_dimensions', True)
    
    In [18]: df
    Out[18]: 
       0  1
    0  1  2
    1  3  4
    
    [2 rows x 2 columns]
    
  • The ArrayFormatter for datetime and timedelta64 now intelligently limit precision based on the values in the array (GH3401)

    Previously output might look like:

      age                 today               diff
    0 2001-01-01 00:00:00 2013-04-19 00:00:00 4491 days, 00:00:00
    1 2004-06-01 00:00:00 2013-04-19 00:00:00 3244 days, 00:00:00
    

    Now the output looks like:

    In [19]: df = DataFrame([ Timestamp('20010101'),
       ....:                  Timestamp('20040601') ], columns=['age'])
       ....: 
    
    In [20]: df['today'] = Timestamp('20130419')
    
    In [21]: df['diff'] = df['today']-df['age']
    
    In [22]: df
    Out[22]: 
             age      today      diff
    0 2001-01-01 2013-04-19 4491 days
    1 2004-06-01 2013-04-19 3244 days
    
    [2 rows x 3 columns]
    

API changes

  • Add -NaN and -nan to the default set of NA values (GH5952). See NA Values.

  • Added Series.str.get_dummies vectorized string method (GH6021), to extract dummy/indicator variables for separated string columns:

    In [23]: s = Series(['a', 'a|b', np.nan, 'a|c'])
    
    In [24]: s.str.get_dummies(sep='|')
    Out[24]: 
       a  b  c
    0  1  0  0
    1  1  1  0
    2  0  0  0
    3  1  0  1
    
    [4 rows x 3 columns]
    
  • Added the NDFrame.equals() method to compare if two NDFrames are equal have equal axes, dtypes, and values. Added the array_equivalent function to compare if two ndarrays are equal. NaNs in identical locations are treated as equal. (GH5283) See also the docs for a motivating example.

    In [25]: df = DataFrame({'col':['foo', 0, np.nan]})
    
    In [26]: df2 = DataFrame({'col':[np.nan, 0, 'foo']}, index=[2,1,0])
    
    In [27]: df.equals(df2)
    Out[27]: False
    
    In [28]: df.equals(df2.sort_index())
    Out[28]: True
    
    In [29]: import pandas.core.common as com
    
    In [30]: com.array_equivalent(np.array([0, np.nan]), np.array([0, np.nan]))
    ---------------------------------------------------------------------------
    AttributeError                            Traceback (most recent call last)
    <ipython-input-30-18a036193b45> in <module>()
    ----> 1 com.array_equivalent(np.array([0, np.nan]), np.array([0, np.nan]))
    
    AttributeError: module 'pandas.core.common' has no attribute 'array_equivalent'
    
    In [31]: np.array_equal(np.array([0, np.nan]), np.array([0, np.nan]))
    Out[31]: False
    
  • DataFrame.apply will use the reduce argument to determine whether a Series or a DataFrame should be returned when the DataFrame is empty (GH6007).

    Previously, calling DataFrame.apply an empty DataFrame would return either a DataFrame if there were no columns, or the function being applied would be called with an empty Series to guess whether a Series or DataFrame should be returned:

    In [32]: def applied_func(col):
      ....:    print("Apply function being called with: ", col)
      ....:    return col.sum()
      ....:
    
    In [33]: empty = DataFrame(columns=['a', 'b'])
    
    In [34]: empty.apply(applied_func)
    Apply function being called with:  Series([], Length: 0, dtype: float64)
    Out[34]:
    a   NaN
    b   NaN
    Length: 2, dtype: float64
    

    Now, when apply is called on an empty DataFrame: if the reduce argument is True a Series will returned, if it is False a DataFrame will be returned, and if it is None (the default) the function being applied will be called with an empty series to try and guess the return type.

    In [35]: empty.apply(applied_func, reduce=True)
    Out[35]:
    a   NaN
    b   NaN
    Length: 2, dtype: float64
    
    In [36]: empty.apply(applied_func, reduce=False)
    Out[36]:
    Empty DataFrame
    Columns: [a, b]
    Index: []
    
    [0 rows x 2 columns]
    

Prior Version Deprecations/Changes

There are no announced changes in 0.13 or prior that are taking effect as of 0.13.1

Deprecations

There are no deprecations of prior behavior in 0.13.1

Enhancements

  • pd.read_csv and pd.to_datetime learned a new infer_datetime_format keyword which greatly improves parsing perf in many cases. Thanks to @lexual for suggesting and @danbirken for rapidly implementing. (GH5490, GH6021)

    If parse_dates is enabled and this flag is set, pandas will attempt to infer the format of the datetime strings in the columns, and if it can be inferred, switch to a faster method of parsing them. In some cases this can increase the parsing speed by ~5-10x.

    # Try to infer the format for the index column
    df = pd.read_csv('foo.csv', index_col=0, parse_dates=True,
                     infer_datetime_format=True)
    
  • date_format and datetime_format keywords can now be specified when writing to excel files (GH4133)

  • MultiIndex.from_product convenience function for creating a MultiIndex from the cartesian product of a set of iterables (GH6055):

    In [32]: shades = ['light', 'dark']
    
    In [33]: colors = ['red', 'green', 'blue']
    
    In [34]: MultiIndex.from_product([shades, colors], names=['shade', 'color'])
    Out[34]: 
    MultiIndex(levels=[['dark', 'light'], ['blue', 'green', 'red']],
               labels=[[1, 1, 1, 0, 0, 0], [2, 1, 0, 2, 1, 0]],
               names=['shade', 'color'])
    
  • Panel apply() will work on non-ufuncs. See the docs.

    In [35]: import pandas.util.testing as tm
    
    In [36]: panel = tm.makePanel(5)
    
    In [37]: panel
    Out[37]: 
    <class 'pandas.core.panel.Panel'>
    Dimensions: 3 (items) x 5 (major_axis) x 4 (minor_axis)
    Items axis: ItemA to ItemC
    Major_axis axis: 2000-01-03 00:00:00 to 2000-01-07 00:00:00
    Minor_axis axis: A to D
    
    In [38]: panel['ItemA']
    Out[38]: 
                       A         B         C         D
    2000-01-03  0.694103  1.893534 -1.735349 -0.850346
    2000-01-04  0.678630  0.639633  1.210384  1.176812
    2000-01-05  0.239556 -0.962029  0.797435 -0.524336
    2000-01-06  0.151227 -2.085266 -0.379811  0.700908
    2000-01-07  0.816127  1.930247  0.702562  0.984188
    
    [5 rows x 4 columns]
    

    Specifying an apply that operates on a Series (to return a single element)

    In [39]: panel.apply(lambda x: x.dtype, axis='items')
    Out[39]: 
                      A        B        C        D
    2000-01-03  float64  float64  float64  float64
    2000-01-04  float64  float64  float64  float64
    2000-01-05  float64  float64  float64  float64
    2000-01-06  float64  float64  float64  float64
    2000-01-07  float64  float64  float64  float64
    
    [5 rows x 4 columns]
    

    A similar reduction type operation

    In [40]: panel.apply(lambda x: x.sum(), axis='major_axis')
    Out[40]: 
          ItemA     ItemB     ItemC
    A  2.579643  3.062757  0.379252
    B  1.416120 -1.960855  0.923558
    C  0.595222 -1.079772 -3.118269
    D  1.487226 -0.734611 -1.979310
    
    [4 rows x 3 columns]
    

    This is equivalent to

    In [41]: panel.sum('major_axis')
    Out[41]: 
          ItemA     ItemB     ItemC
    A  2.579643  3.062757  0.379252
    B  1.416120 -1.960855  0.923558
    C  0.595222 -1.079772 -3.118269
    D  1.487226 -0.734611 -1.979310
    
    [4 rows x 3 columns]
    

    A transformation operation that returns a Panel, but is computing the z-score across the major_axis

    In [42]: result = panel.apply(
       ....:            lambda x: (x-x.mean())/x.std(),
       ....:            axis='major_axis')
       ....: 
    
    In [43]: result
    Out[43]: 
    <class 'pandas.core.panel.Panel'>
    Dimensions: 3 (items) x 5 (major_axis) x 4 (minor_axis)
    Items axis: ItemA to ItemC
    Major_axis axis: 2000-01-03 00:00:00 to 2000-01-07 00:00:00
    Minor_axis axis: A to D
    
    In [44]: result['ItemA']
    Out[44]: 
                       A         B         C         D
    2000-01-03  0.595800  0.907552 -1.556260 -1.244875
    2000-01-04  0.544058  0.200868  0.915883  0.953747
    2000-01-05 -0.924165 -0.701810  0.569325 -0.891290
    2000-01-06 -1.219530 -1.334852 -0.418654  0.437589
    2000-01-07  1.003837  0.928242  0.489705  0.744830
    
    [5 rows x 4 columns]
    
  • Panel apply() operating on cross-sectional slabs. (GH1148)

    In [45]: f = lambda x: ((x.T-x.mean(1))/x.std(1)).T
    
    In [46]: result = panel.apply(f, axis = ['items','major_axis'])
    
    In [47]: result
    Out[47]: 
    <class 'pandas.core.panel.Panel'>
    Dimensions: 4 (items) x 5 (major_axis) x 3 (minor_axis)
    Items axis: A to D
    Major_axis axis: 2000-01-03 00:00:00 to 2000-01-07 00:00:00
    Minor_axis axis: ItemA to ItemC
    
    In [48]: result.loc[:,:,'ItemA']
    Out[48]: 
                       A         B         C         D
    2000-01-03  0.331409  1.071034 -0.914540 -0.510587
    2000-01-04 -0.741017 -0.118794  0.383277  0.537212
    2000-01-05  0.065042 -0.767353  0.655436  0.069467
    2000-01-06  0.027932 -0.569477  0.908202  0.610585
    2000-01-07  1.116434  1.133591  0.871287  1.004064
    
    [5 rows x 4 columns]
    

    This is equivalent to the following

    In [49]: result = Panel(dict([ (ax,f(panel.loc[:,:,ax]))
       ....:                         for ax in panel.minor_axis ]))
       ....: 
    
    In [50]: result
    Out[50]: 
    <class 'pandas.core.panel.Panel'>
    Dimensions: 4 (items) x 5 (major_axis) x 3 (minor_axis)
    Items axis: A to D
    Major_axis axis: 2000-01-03 00:00:00 to 2000-01-07 00:00:00
    Minor_axis axis: ItemA to ItemC
    
    In [51]: result.loc[:,:,'ItemA']
    Out[51]: 
                       A         B         C         D
    2000-01-03  0.331409  1.071034 -0.914540 -0.510587
    2000-01-04 -0.741017 -0.118794  0.383277  0.537212
    2000-01-05  0.065042 -0.767353  0.655436  0.069467
    2000-01-06  0.027932 -0.569477  0.908202  0.610585
    2000-01-07  1.116434  1.133591  0.871287  1.004064
    
    [5 rows x 4 columns]
    

Performance

Performance improvements for 0.13.1

  • Series datetime/timedelta binary operations (GH5801)
  • DataFrame count/dropna for axis=1
  • Series.str.contains now has a regex=False keyword which can be faster for plain (non-regex) string patterns. (GH5879)
  • Series.str.extract (GH5944)
  • dtypes/ftypes methods (GH5968)
  • indexing with object dtypes (GH5968)
  • DataFrame.apply (GH6013)
  • Regression in JSON IO (GH5765)
  • Index construction from Series (GH6150)

Experimental

There are no experimental changes in 0.13.1

Bug Fixes

See V0.13.1 Bug Fixes for an extensive list of bugs that have been fixed in 0.13.1.

See the full release notes or issue tracker on GitHub for a complete list of all API changes, Enhancements and Bug Fixes.

v0.13.0 (January 3, 2014)

This is a major release from 0.12.0 and includes a number of API changes, several new features and enhancements along with a large number of bug fixes.

Highlights include:

  • support for a new index type Float64Index, and other Indexing enhancements
  • HDFStore has a new string based syntax for query specification
  • support for new methods of interpolation
  • updated timedelta operations
  • a new string manipulation method extract
  • Nanosecond support for Offsets
  • isin for DataFrames

Several experimental features are added, including:

  • new eval/query methods for expression evaluation
  • support for msgpack serialization
  • an i/o interface to Google’s BigQuery

Their are several new or updated docs sections including:

Warning

In 0.13.0 Series has internally been refactored to no longer sub-class ndarray but instead subclass NDFrame, similar to the rest of the pandas containers. This should be a transparent change with only very limited API implications. See Internal Refactoring

API changes

  • read_excel now supports an integer in its sheetname argument giving the index of the sheet to read in (GH4301).

  • Text parser now treats anything that reads like inf (“inf”, “Inf”, “-Inf”, “iNf”, etc.) as infinity. (GH4220, GH4219), affecting read_table, read_csv, etc.

  • pandas now is Python 2/3 compatible without the need for 2to3 thanks to @jtratner. As a result, pandas now uses iterators more extensively. This also led to the introduction of substantive parts of the Benjamin Peterson’s six library into compat. (GH4384, GH4375, GH4372)

  • pandas.util.compat and pandas.util.py3compat have been merged into pandas.compat. pandas.compat now includes many functions allowing 2/3 compatibility. It contains both list and iterator versions of range, filter, map and zip, plus other necessary elements for Python 3 compatibility. lmap, lzip, lrange and lfilter all produce lists instead of iterators, for compatibility with numpy, subscripting and pandas constructors.(GH4384, GH4375, GH4372)

  • Series.get with negative indexers now returns the same as [] (GH4390)

  • Changes to how Index and MultiIndex handle metadata (levels, labels, and names) (GH4039):

    # previously, you would have set levels or labels directly
    index.levels = [[1, 2, 3, 4], [1, 2, 4, 4]]
    
    # now, you use the set_levels or set_labels methods
    index = index.set_levels([[1, 2, 3, 4], [1, 2, 4, 4]])
    
    # similarly, for names, you can rename the object
    # but setting names is not deprecated
    index = index.set_names(["bob", "cranberry"])
    
    # and all methods take an inplace kwarg - but return None
    index.set_names(["bob", "cranberry"], inplace=True)
    
  • All division with NDFrame objects is now truedivision, regardless of the future import. This means that operating on pandas objects will by default use floating point division, and return a floating point dtype. You can use // and floordiv to do integer division.

    Integer division

    In [3]: arr = np.array([1, 2, 3, 4])
    
    In [4]: arr2 = np.array([5, 3, 2, 1])
    
    In [5]: arr / arr2
    Out[5]: array([0, 0, 1, 4])
    
    In [6]: Series(arr) // Series(arr2)
    Out[6]:
    0    0
    1    0
    2    1
    3    4
    dtype: int64
    

    True Division

    In [7]: pd.Series(arr) / pd.Series(arr2) # no future import required
    Out[7]:
    0    0.200000
    1    0.666667
    2    1.500000
    3    4.000000
    dtype: float64
    
  • Infer and downcast dtype if downcast='infer' is passed to fillna/ffill/bfill (GH4604)

  • __nonzero__ for all NDFrame objects, will now raise a ValueError, this reverts back to (GH1073, GH4633) behavior. See gotchas for a more detailed discussion.

    This prevents doing boolean comparison on entire pandas objects, which is inherently ambiguous. These all will raise a ValueError.

    if df:
       ....
    df1 and df2
    s1 and s2
    

    Added the .bool() method to NDFrame objects to facilitate evaluating of single-element boolean Series:

    In [1]: Series([True]).bool()
    Out[1]: True
    
    In [2]: Series([False]).bool()
    Out[2]: False
    
    In [3]: DataFrame([[True]]).bool()
    Out[3]: True
    
    In [4]: DataFrame([[False]]).bool()
    Out[4]: False
    
  • All non-Index NDFrames (Series, DataFrame, Panel, Panel4D, SparsePanel, etc.), now support the entire set of arithmetic operators and arithmetic flex methods (add, sub, mul, etc.). SparsePanel does not support pow or mod with non-scalars. (GH3765)

  • Series and DataFrame now have a mode() method to calculate the statistical mode(s) by axis/Series. (GH5367)

  • Chained assignment will now by default warn if the user is assigning to a copy. This can be changed with the option mode.chained_assignment, allowed options are raise/warn/None. See the docs.

    In [5]: dfc = DataFrame({'A':['aaa','bbb','ccc'],'B':[1,2,3]})
    
    In [6]: pd.set_option('chained_assignment','warn')
    

    The following warning / exception will show if this is attempted.

    In [7]: dfc.loc[0]['A'] = 1111
    
    Traceback (most recent call last)
       ...
    SettingWithCopyWarning:
       A value is trying to be set on a copy of a slice from a DataFrame.
       Try using .loc[row_index,col_indexer] = value instead
    

    Here is the correct method of assignment.

    In [8]: dfc.loc[0,'A'] = 11
    
    In [9]: dfc
    Out[9]: 
         A  B
    0   11  1
    1  bbb  2
    2  ccc  3
    
    [3 rows x 2 columns]
    
  • Panel.reindex has the following call signature Panel.reindex(items=None, major_axis=None, minor_axis=None, **kwargs)

    to conform with other NDFrame objects. See Internal Refactoring for more information.

  • Series.argmin and Series.argmax are now aliased to Series.idxmin and Series.idxmax. These return the index of the

    min or max element respectively. Prior to 0.13.0 these would return the position of the min / max element. (GH6214)

Prior Version Deprecations/Changes

These were announced changes in 0.12 or prior that are taking effect as of 0.13.0

  • Remove deprecated Factor (GH3650)
  • Remove deprecated set_printoptions/reset_printoptions (GH3046)
  • Remove deprecated _verbose_info (GH3215)
  • Remove deprecated read_clipboard/to_clipboard/ExcelFile/ExcelWriter from pandas.io.parsers (GH3717) These are available as functions in the main pandas namespace (e.g. pd.read_clipboard)
  • default for tupleize_cols is now False for both to_csv and read_csv. Fair warning in 0.12 (GH3604)
  • default for display.max_seq_len is now 100 rather then None. This activates truncated display (“…”) of long sequences in various places. (GH3391)

Deprecations

Deprecated in 0.13.0

  • deprecated iterkv, which will be removed in a future release (this was an alias of iteritems used to bypass 2to3’s changes). (GH4384, GH4375, GH4372)
  • deprecated the string method match, whose role is now performed more idiomatically by extract. In a future release, the default behavior of match will change to become analogous to contains, which returns a boolean indexer. (Their distinction is strictness: match relies on re.match while contains relies on re.search.) In this release, the deprecated behavior is the default, but the new behavior is available through the keyword argument as_indexer=True.

Indexing API Changes

Prior to 0.13, it was impossible to use a label indexer (.loc/.ix) to set a value that was not contained in the index of a particular axis. (GH2578). See the docs

In the Series case this is effectively an appending operation

In [10]: s = Series([1,2,3])

In [11]: s
Out[11]: 
0    1
1    2
2    3
Length: 3, dtype: int64

In [12]: s[5] = 5.

In [13]: s
Out[13]: 
0    1.0
1    2.0
2    3.0
5    5.0
Length: 4, dtype: float64
In [14]: dfi = DataFrame(np.arange(6).reshape(3,2),
   ....:                 columns=['A','B'])
   ....: 

In [15]: dfi
Out[15]: 
   A  B
0  0  1
1  2  3
2  4  5

[3 rows x 2 columns]

This would previously KeyError

In [16]: dfi.loc[:,'C'] = dfi.loc[:,'A']

In [17]: dfi
Out[17]: 
   A  B  C
0  0  1  0
1  2  3  2
2  4  5  4

[3 rows x 3 columns]

This is like an append operation.

In [18]: dfi.loc[3] = 5

In [19]: dfi
Out[19]: 
   A  B  C
0  0  1  0
1  2  3  2
2  4  5  4
3  5  5  5

[4 rows x 3 columns]

A Panel setting operation on an arbitrary axis aligns the input to the Panel

In [20]: p = pd.Panel(np.arange(16).reshape(2,4,2),
   ....:             items=['Item1','Item2'],
   ....:             major_axis=pd.date_range('2001/1/12',periods=4),
   ....:             minor_axis=['A','B'],dtype='float64')
   ....: 

In [21]: p
Out[21]: 
<class 'pandas.core.panel.Panel'>
Dimensions: 2 (items) x 4 (major_axis) x 2 (minor_axis)
Items axis: Item1 to Item2
Major_axis axis: 2001-01-12 00:00:00 to 2001-01-15 00:00:00
Minor_axis axis: A to B

In [22]: p.loc[:,:,'C'] = Series([30,32],index=p.items)

In [23]: p
Out[23]: 
<class 'pandas.core.panel.Panel'>
Dimensions: 2 (items) x 4 (major_axis) x 3 (minor_axis)
Items axis: Item1 to Item2
Major_axis axis: 2001-01-12 00:00:00 to 2001-01-15 00:00:00
Minor_axis axis: A to C

In [24]: p.loc[:,:,'C']
Out[24]: 
            Item1  Item2
2001-01-12   30.0   32.0
2001-01-13   30.0   32.0
2001-01-14   30.0   32.0
2001-01-15   30.0   32.0

[4 rows x 2 columns]

Float64Index API Change

  • Added a new index type, Float64Index. This will be automatically created when passing floating values in index creation. This enables a pure label-based slicing paradigm that makes [],ix,loc for scalar indexing and slicing work exactly the same. See the docs, (GH263)

    Construction is by default for floating type values.

    In [25]: index = Index([1.5, 2, 3, 4.5, 5])
    
    In [26]: index
    Out[26]: Float64Index([1.5, 2.0, 3.0, 4.5, 5.0], dtype='float64')
    
    In [27]: s = Series(range(5),index=index)
    
    In [28]: s
    Out[28]: 
    1.5    0
    2.0    1
    3.0    2
    4.5    3
    5.0    4
    Length: 5, dtype: int64
    

    Scalar selection for [],.ix,.loc will always be label based. An integer will match an equal float index (e.g. 3 is equivalent to 3.0)

    In [29]: s[3]
    Out[29]: 2
    
    In [30]: s.loc[3]
    Out[30]: 2
    

    The only positional indexing is via iloc

    In [31]: s.iloc[3]
    Out[31]: 3
    

    A scalar index that is not found will raise KeyError

    Slicing is ALWAYS on the values of the index, for [],ix,loc and ALWAYS positional with iloc

    In [32]: s[2:4]
    Out[32]: 
    2.0    1
    3.0    2
    Length: 2, dtype: int64
    
    In [33]: s.loc[2:4]
    Out[33]: 
    2.0    1
    3.0    2
    Length: 2, dtype: int64
    
    In [34]: s.iloc[2:4]
    Out[34]: 
    3.0    2
    4.5    3
    Length: 2, dtype: int64
    

    In float indexes, slicing using floats are allowed

    In [35]: s[2.1:4.6]
    Out[35]: 
    3.0    2
    4.5    3
    Length: 2, dtype: int64
    
    In [36]: s.loc[2.1:4.6]
    Out[36]: 
    3.0    2
    4.5    3
    Length: 2, dtype: int64
    
  • Indexing on other index types are preserved (and positional fallback for [],ix), with the exception, that floating point slicing on indexes on non Float64Index will now raise a TypeError.

    In [1]: Series(range(5))[3.5]
    TypeError: the label [3.5] is not a proper indexer for this index type (Int64Index)
    
    In [1]: Series(range(5))[3.5:4.5]
    TypeError: the slice start [3.5] is not a proper indexer for this index type (Int64Index)
    

    Using a scalar float indexer will be deprecated in a future version, but is allowed for now.

    In [3]: Series(range(5))[3.0]
    Out[3]: 3
    

HDFStore API Changes

  • Query Format Changes. A much more string-like query format is now supported. See the docs.

    In [37]: path = 'test.h5'
    
    In [38]: dfq = DataFrame(randn(10,4),
       ....:          columns=list('ABCD'),
       ....:          index=date_range('20130101',periods=10))
       ....: 
    
    In [39]: dfq.to_hdf(path,'dfq',format='table',data_columns=True)
    

    Use boolean expressions, with in-line function evaluation.

    In [40]: read_hdf(path,'dfq',
       ....:     where="index>Timestamp('20130104') & columns=['A', 'B']")
       ....: 
    Out[40]: 
                       A         B
    2013-01-05  1.057633 -0.791489
    2013-01-06  1.910759  0.787965
    2013-01-07  1.043945  2.107785
    2013-01-08  0.749185 -0.675521
    2013-01-09 -0.276646  1.924533
    2013-01-10  0.226363 -2.078618
    
    [6 rows x 2 columns]
    

    Use an inline column reference

    In [41]: read_hdf(path,'dfq',
       ....:     where="A>0 or C>0")
       ....: 
    Out[41]: 
                       A         B         C         D
    2013-01-01 -0.414505 -1.425795  0.209395 -0.592886
    2013-01-02 -1.473116 -0.896581  1.104352 -0.431550
    2013-01-03 -0.161137  0.889157  0.288377 -1.051539
    2013-01-04 -0.319561 -0.619993  0.156998 -0.571455
    2013-01-05  1.057633 -0.791489 -0.524627  0.071878
    2013-01-06  1.910759  0.787965  0.513082 -0.546416
    2013-01-07  1.043945  2.107785  1.459927  1.015405
    2013-01-08  0.749185 -0.675521  0.440266  0.688972
    2013-01-09 -0.276646  1.924533  0.411204  0.890765
    2013-01-10  0.226363 -2.078618 -0.387886 -0.087107
    
    [10 rows x 4 columns]
    
  • the format keyword now replaces the table keyword; allowed values are fixed(f) or table(t) the same defaults as prior < 0.13.0 remain, e.g. put implies fixed format and append implies table format. This default format can be set as an option by setting io.hdf.default_format.

    In [42]: path = 'test.h5'
    
    In [43]: df = pd.DataFrame(np.random.randn(10,2))
    
    In [44]: df.to_hdf(path,'df_table',format='table')
    
    In [45]: df.to_hdf(path,'df_table2',append=True)
    
    In [46]: df.to_hdf(path,'df_fixed')
    
    In [47]: with pd.HDFStore(path) as store:
       ....:    print(store)
       ....: 
    <class 'pandas.io.pytables.HDFStore'>
    File path: test.h5
    
  • Significant table writing performance improvements

  • handle a passed Series in table format (GH4330)

  • can now serialize a timedelta64[ns] dtype in a table (GH3577), See the docs.

  • added an is_open property to indicate if the underlying file handle is_open; a closed store will now report ‘CLOSED’ when viewing the store (rather than raising an error) (GH4409)

  • a close of a HDFStore now will close that instance of the HDFStore but will only close the actual file if the ref count (by PyTables) w.r.t. all of the open handles are 0. Essentially you have a local instance of HDFStore referenced by a variable. Once you close it, it will report closed. Other references (to the same file) will continue to operate until they themselves are closed. Performing an action on a closed file will raise ClosedFileError

    In [48]: path = 'test.h5'
    
    In [49]: df = DataFrame(randn(10,2))
    
    In [50]: store1 = HDFStore(path)
    
    In [51]: store2 = HDFStore(path)
    
    In [52]: store1.append('df',df)
    
    In [53]: store2.append('df2',df)
    
    In [54]: store1
    Out[54]: 
    <class 'pandas.io.pytables.HDFStore'>
    File path: test.h5
    
    In [55]: store2
    Out[55]: 
    <class 'pandas.io.pytables.HDFStore'>
    File path: test.h5
    
    In [56]: store1.close()
    
    In [57]: store2
    Out[57]: 
    <class 'pandas.io.pytables.HDFStore'>
    File path: test.h5
    
    In [58]: store2.close()
    
    In [59]: store2
    Out[59]: 
    <class 'pandas.io.pytables.HDFStore'>
    File path: test.h5
    
  • removed the _quiet attribute, replace by a DuplicateWarning if retrieving duplicate rows from a table (GH4367)

  • removed the warn argument from open. Instead a PossibleDataLossError exception will be raised if you try to use mode='w' with an OPEN file handle (GH4367)

  • allow a passed locations array or mask as a where condition (GH4467). See the docs for an example.

  • add the keyword dropna=True to append to change whether ALL nan rows are not written to the store (default is True, ALL nan rows are NOT written), also settable via the option io.hdf.dropna_table (GH4625)

  • pass thru store creation arguments; can be used to support in-memory stores

DataFrame repr Changes

The HTML and plain text representations of DataFrame now show a truncated view of the table once it exceeds a certain size, rather than switching to the short info view (GH4886, GH5550). This makes the representation more consistent as small DataFrames get larger.

Truncated HTML representation of a DataFrame

To get the info view, call DataFrame.info(). If you prefer the info view as the repr for large DataFrames, you can set this by running set_option('display.large_repr', 'info').

Enhancements

  • df.to_clipboard() learned a new excel keyword that let’s you paste df data directly into excel (enabled by default). (GH5070).

  • read_html now raises a URLError instead of catching and raising a ValueError (GH4303, GH4305)

  • Added a test for read_clipboard() and to_clipboard() (GH4282)

  • Clipboard functionality now works with PySide (GH4282)

  • Added a more informative error message when plot arguments contain overlapping color and style arguments (GH4402)

  • to_dict now takes records as a possible outtype. Returns an array of column-keyed dictionaries. (GH4936)

  • NaN handing in get_dummies (GH4446) with dummy_na

    # previously, nan was erroneously counted as 2 here
    # now it is not counted at all
    In [60]: get_dummies([1, 2, np.nan])
    Out[60]: 
       1.0  2.0
    0    1    0
    1    0    1
    2    0    0
    
    [3 rows x 2 columns]
    
    # unless requested
    In [61]: get_dummies([1, 2, np.nan], dummy_na=True)
    Out[61]: 
        1.0   2.0  NaN 
    0     1     0     0
    1     0     1     0
    2     0     0     1
    
    [3 rows x 3 columns]
    
  • timedelta64[ns] operations. See the docs.

    Warning

    Most of these operations require numpy >= 1.7

    Using the new top-level to_timedelta, you can convert a scalar or array from the standard timedelta format (produced by to_csv) into a timedelta type (np.timedelta64 in nanoseconds).

    In [62]: to_timedelta('1 days 06:05:01.00003')
    Out[62]: Timedelta('1 days 06:05:01.000030')
    
    In [63]: to_timedelta('15.5us')
    Out[63]: Timedelta('0 days 00:00:00.000015')
    
    In [64]: to_timedelta(['1 days 06:05:01.00003','15.5us','nan'])
    Out[64]: TimedeltaIndex(['1 days 06:05:01.000030', '0 days 00:00:00.000015', NaT], dtype='timedelta64[ns]', freq=None)
    
    In [65]: to_timedelta(np.arange(5),unit='s')
    Out[65]: TimedeltaIndex(['00:00:00', '00:00:01', '00:00:02', '00:00:03', '00:00:04'], dtype='timedelta64[ns]', freq=None)
    
    In [66]: to_timedelta(np.arange(5),unit='d')
    Out[66]: TimedeltaIndex(['0 days', '1 days', '2 days', '3 days', '4 days'], dtype='timedelta64[ns]', freq=None)
    

    A Series of dtype timedelta64[ns] can now be divided by another timedelta64[ns] object, or astyped to yield a float64 dtyped Series. This is frequency conversion. See the docs for the docs.

    In [67]: from datetime import timedelta
    
    In [68]: td = Series(date_range('20130101',periods=4))-Series(date_range('20121201',periods=4))
    
    In [69]: td[2] += np.timedelta64(timedelta(minutes=5,seconds=3))
    
    In [70]: td[3] = np.nan
    
    In [71]: td
    Out[71]: 
    0   31 days 00:00:00
    1   31 days 00:00:00
    2   31 days 00:05:03
    3                NaT
    Length: 4, dtype: timedelta64[ns]
    
    # to days
    In [72]: td / np.timedelta64(1,'D')
    Out[72]: 
    0    31.000000
    1    31.000000
    2    31.003507
    3          NaN
    Length: 4, dtype: float64
    
    In [73]: td.astype('timedelta64[D]')
    Out[73]: 
    0    31.0
    1    31.0
    2    31.0
    3     NaN
    Length: 4, dtype: float64
    
    # to seconds
    In [74]: td / np.timedelta64(1,'s')
    Out[74]: 
    0    2678400.0
    1    2678400.0
    2    2678703.0
    3          NaN
    Length: 4, dtype: float64
    
    In [75]: td.astype('timedelta64[s]')
    Out[75]: 
    0    2678400.0
    1    2678400.0
    2    2678703.0
    3          NaN
    Length: 4, dtype: float64
    

    Dividing or multiplying a timedelta64[ns] Series by an integer or integer Series

    In [76]: td * -1
    Out[76]: 
    0   -31 days +00:00:00
    1   -31 days +00:00:00
    2   -32 days +23:54:57
    3                  NaT
    Length: 4, dtype: timedelta64[ns]
    
    In [77]: td * Series([1,2,3,4])
    Out[77]: 
    0   31 days 00:00:00
    1   62 days 00:00:00
    2   93 days 00:15:09
    3                NaT
    Length: 4, dtype: timedelta64[ns]
    

    Absolute DateOffset objects can act equivalently to timedeltas

    In [78]: from pandas import offsets
    
    In [79]: td + offsets.Minute(5) + offsets.Milli(5)
    Out[79]: 
    0   31 days 00:05:00.005000
    1   31 days 00:05:00.005000
    2   31 days 00:10:03.005000
    3                       NaT
    Length: 4, dtype: timedelta64[ns]
    

    Fillna is now supported for timedeltas

    In [80]: td.fillna(0)
    Out[80]: 
    0   31 days 00:00:00
    1   31 days 00:00:00
    2   31 days 00:05:03
    3    0 days 00:00:00
    Length: 4, dtype: timedelta64[ns]
    
    In [81]: td.fillna(timedelta(days=1,seconds=5))
    Out[81]: 
    0   31 days 00:00:00
    1   31 days 00:00:00
    2   31 days 00:05:03
    3    1 days 00:00:05
    Length: 4, dtype: timedelta64[ns]
    

    You can do numeric reduction operations on timedeltas.

    In [82]: td.mean()
    Out[82]: Timedelta('31 days 00:01:41')
    
    In [83]: td.quantile(.1)
    Out[83]: Timedelta('31 days 00:00:00')
    
  • plot(kind='kde') now accepts the optional parameters bw_method and ind, passed to scipy.stats.gaussian_kde() (for scipy >= 0.11.0) to set the bandwidth, and to gkde.evaluate() to specify the indices at which it is evaluated, respectively. See scipy docs. (GH4298)

  • DataFrame constructor now accepts a numpy masked record array (GH3478)

  • The new vectorized string method extract return regular expression matches more conveniently.

    In [84]: Series(['a1', 'b2', 'c3']).str.extract('[ab](\d)')
    Out[84]: 
         0
    0    1
    1    2
    2  NaN
    
    [3 rows x 1 columns]
    

    Elements that do not match return NaN. Extracting a regular expression with more than one group returns a DataFrame with one column per group.

    In [85]: Series(['a1', 'b2', 'c3']).str.extract('([ab])(\d)')
    Out[85]: 
         0    1
    0    a    1
    1    b    2
    2  NaN  NaN
    
    [3 rows x 2 columns]
    

    Elements that do not match return a row of NaN. Thus, a Series of messy strings can be converted into a like-indexed Series or DataFrame of cleaned-up or more useful strings, without necessitating get() to access tuples or re.match objects.

    Named groups like

    In [86]: Series(['a1', 'b2', 'c3']).str.extract(
       ....:         '(?P<letter>[ab])(?P<digit>\d)')
       ....: 
    Out[86]: 
      letter digit
    0      a     1
    1      b     2
    2    NaN   NaN
    
    [3 rows x 2 columns]
    

    and optional groups can also be used.

    In [87]: Series(['a1', 'b2', '3']).str.extract(
       ....:          '(?P<letter>[ab])?(?P<digit>\d)')
       ....: 
    Out[87]: 
      letter digit
    0      a     1
    1      b     2
    2    NaN     3
    
    [3 rows x 2 columns]
    
  • read_stata now accepts Stata 13 format (GH4291)

  • read_fwf now infers the column specifications from the first 100 rows of the file if the data has correctly separated and properly aligned columns using the delimiter provided to the function (GH4488).

  • support for nanosecond times as an offset

    Warning

    These operations require numpy >= 1.7

    Period conversions in the range of seconds and below were reworked and extended up to nanoseconds. Periods in the nanosecond range are now available.

    In [88]: date_range('2013-01-01', periods=5, freq='5N')
    Out[88]: 
    DatetimeIndex([          '2013-01-01 00:00:00',
                   '2013-01-01 00:00:00.000000005',
                   '2013-01-01 00:00:00.000000010',
                   '2013-01-01 00:00:00.000000015',
                   '2013-01-01 00:00:00.000000020'],
                  dtype='datetime64[ns]', freq='5N')
    

    or with frequency as offset

    In [89]: date_range('2013-01-01', periods=5, freq=pd.offsets.Nano(5))
    Out[89]: 
    DatetimeIndex([          '2013-01-01 00:00:00',
                   '2013-01-01 00:00:00.000000005',
                   '2013-01-01 00:00:00.000000010',
                   '2013-01-01 00:00:00.000000015',
                   '2013-01-01 00:00:00.000000020'],
                  dtype='datetime64[ns]', freq='5N')
    

    Timestamps can be modified in the nanosecond range

    In [90]: t = Timestamp('20130101 09:01:02')
    
    In [91]: t + pd.tseries.offsets.Nano(123)
    Out[91]: Timestamp('2013-01-01 09:01:02.000000123')
    
  • A new method, isin for DataFrames, which plays nicely with boolean indexing. The argument to isin, what we’re comparing the DataFrame to, can be a DataFrame, Series, dict, or array of values. See the docs for more.

    To get the rows where any of the conditions are met:

    In [92]: dfi = DataFrame({'A': [1, 2, 3, 4], 'B': ['a', 'b', 'f', 'n']})
    
    In [93]: dfi
    Out[93]: 
       A  B
    0  1  a
    1  2  b
    2  3  f
    3  4  n
    
    [4 rows x 2 columns]
    
    In [94]: other = DataFrame({'A': [1, 3, 3, 7], 'B': ['e', 'f', 'f', 'e']})
    
    In [95]: mask = dfi.isin(other)
    
    In [96]: mask
    Out[96]: 
           A      B
    0   True  False
    1  False  False
    2   True   True
    3  False  False
    
    [4 rows x 2 columns]
    
    In [97]: dfi[mask.any(1)]
    Out[97]: 
       A  B
    0  1  a
    2  3  f
    
    [2 rows x 2 columns]
    
  • Series now supports a to_frame method to convert it to a single-column DataFrame (GH5164)

  • All R datasets listed here http://stat.ethz.ch/R-manual/R-devel/library/datasets/html/00Index.html can now be loaded into Pandas objects

    # note that pandas.rpy was deprecated in v0.16.0
    import pandas.rpy.common as com
    com.load_data('Titanic')
    
  • tz_localize can infer a fall daylight savings transition based on the structure of the unlocalized data (GH4230), see the docs

  • DatetimeIndex is now in the API documentation, see the docs

  • json_normalize() is a new method to allow you to create a flat table from semi-structured JSON data. See the docs (GH1067)

  • Added PySide support for the qtpandas DataFrameModel and DataFrameWidget.

  • Python csv parser now supports usecols (GH4335)

  • Frequencies gained several new offsets:

    • LastWeekOfMonth (GH4637)
    • FY5253, and FY5253Quarter (GH4511)
  • DataFrame has a new interpolate method, similar to Series (GH4434, GH1892)

    In [98]: df = DataFrame({'A': [1, 2.1, np.nan, 4.7, 5.6, 6.8],
       ....:                 'B': [.25, np.nan, np.nan, 4, 12.2, 14.4]})
       ....: 
    
    In [99]: df.interpolate()
    Out[99]: 
         A      B
    0  1.0   0.25
    1  2.1   1.50
    2  3.4   2.75
    3  4.7   4.00
    4  5.6  12.20
    5  6.8  14.40
    
    [6 rows x 2 columns]
    

    Additionally, the method argument to interpolate has been expanded to include 'nearest', 'zero', 'slinear', 'quadratic', 'cubic', 'barycentric', 'krogh', 'piecewise_polynomial', 'pchip', 'polynomial', 'spline' The new methods require scipy. Consult the Scipy reference guide and documentation for more information about when the various methods are appropriate. See the docs.

    Interpolate now also accepts a limit keyword argument. This works similar to fillna’s limit:

    In [100]: ser = Series([1, 3, np.nan, np.nan, np.nan, 11])
    
    In [101]: ser.interpolate(limit=2)
    Out[101]: 
    0     1.0
    1     3.0
    2     5.0
    3     7.0
    4     NaN
    5    11.0
    Length: 6, dtype: float64
    
  • Added wide_to_long panel data convenience function. See the docs.

    In [102]: np.random.seed(123)
    
    In [103]: df = pd.DataFrame({"A1970" : {0 : "a", 1 : "b", 2 : "c"},
       .....:                    "A1980" : {0 : "d", 1 : "e", 2 : "f"},
       .....:                    "B1970" : {0 : 2.5, 1 : 1.2, 2 : .7},
       .....:                    "B1980" : {0 : 3.2, 1 : 1.3, 2 : .1},
       .....:                    "X"     : dict(zip(range(3), np.random.randn(3)))
       .....:                   })
       .....: 
    
    In [104]: df["id"] = df.index
    
    In [105]: df
    Out[105]: 
      A1970 A1980  B1970  B1980         X  id
    0     a     d    2.5    3.2 -1.085631   0
    1     b     e    1.2    1.3  0.997345   1
    2     c     f    0.7    0.1  0.282978   2
    
    [3 rows x 6 columns]
    
    In [106]: wide_to_long(df, ["A", "B"], i="id", j="year")
    Out[106]: 
                    X  A    B
    id year                  
    0  1970 -1.085631  a  2.5
    1  1970  0.997345  b  1.2
    2  1970  0.282978  c  0.7
    0  1980 -1.085631  d  3.2
    1  1980  0.997345  e  1.3
    2  1980  0.282978  f  0.1
    
    [6 rows x 3 columns]
    
  • to_csv now takes a date_format keyword argument that specifies how output datetime objects should be formatted. Datetimes encountered in the index, columns, and values will all have this formatting applied. (GH4313)
  • DataFrame.plot will scatter plot x versus y by passing kind='scatter' (GH2215)
  • Added support for Google Analytics v3 API segment IDs that also supports v2 IDs. (GH5271)

Experimental

  • The new eval() function implements expression evaluation using numexpr behind the scenes. This results in large speedups for complicated expressions involving large DataFrames/Series. For example,

    In [107]: nrows, ncols = 20000, 100
    
    In [108]: df1, df2, df3, df4 = [DataFrame(randn(nrows, ncols))
       .....:                       for _ in range(4)]
       .....: 
    
    # eval with NumExpr backend
    In [109]: %timeit pd.eval('df1 + df2 + df3 + df4')
    8.07 ms +- 212 us per loop (mean +- std. dev. of 7 runs, 100 loops each)
    
    # pure Python evaluation
    In [110]: %timeit df1 + df2 + df3 + df4
    11.4 ms +- 752 us per loop (mean +- std. dev. of 7 runs, 100 loops each)
    

    For more details, see the the docs

  • Similar to pandas.eval, DataFrame has a new DataFrame.eval method that evaluates an expression in the context of the DataFrame. For example,

    In [111]: df = DataFrame(randn(10, 2), columns=['a', 'b'])
    
    In [112]: df.eval('a + b')
    Out[112]: 
    0   -0.685204
    1    1.589745
    2    0.325441
    3   -1.784153
    4   -0.432893
    5    0.171850
    6    1.895919
    7    3.065587
    8   -0.092759
    9    1.391365
    Length: 10, dtype: float64
    
  • query() method has been added that allows you to select elements of a DataFrame using a natural query syntax nearly identical to Python syntax. For example,

    In [113]: n = 20
    
    In [114]: df = DataFrame(np.random.randint(n, size=(n, 3)), columns=['a', 'b', 'c'])
    
    In [115]: df.query('a < b < c')
    Out[115]: 
        a   b   c
    11  1   5   8
    15  8  16  19
    
    [2 rows x 3 columns]
    

    selects all the rows of df where a < b < c evaluates to True. For more details see the the docs.

  • pd.read_msgpack() and pd.to_msgpack() are now a supported method of serialization of arbitrary pandas (and python objects) in a lightweight portable binary format. See the docs

    Warning

    Since this is an EXPERIMENTAL LIBRARY, the storage format may not be stable until a future release.

    In [116]: df = DataFrame(np.random.rand(5,2),columns=list('AB'))
    
    In [117]: df.to_msgpack('foo.msg')
    
    In [118]: pd.read_msgpack('foo.msg')
    Out[118]: 
              A         B
    0  0.251082  0.017357
    1  0.347915  0.929879
    2  0.546233  0.203368
    3  0.064942  0.031722
    4  0.355309  0.524575
    
    [5 rows x 2 columns]
    
    In [119]: s = Series(np.random.rand(5),index=date_range('20130101',periods=5))
    
    In [120]: pd.to_msgpack('foo.msg', df, s)
    
    In [121]: pd.read_msgpack('foo.msg')
    Out[121]: 
    [          A         B
     0  0.251082  0.017357
     1  0.347915  0.929879
     2  0.546233  0.203368
     3  0.064942  0.031722
     4  0.355309  0.524575
     
     [5 rows x 2 columns], 2013-01-01    0.022321
     2013-01-02    0.227025
     2013-01-03    0.383282
     2013-01-04    0.193225
     2013-01-05    0.110977
     Freq: D, Length: 5, dtype: float64]
    

    You can pass iterator=True to iterator over the unpacked results

    In [122]: for o in pd.read_msgpack('foo.msg',iterator=True):
       .....:    print(o)
       .....: 
              A         B
    0  0.251082  0.017357
    1  0.347915  0.929879
    2  0.546233  0.203368
    3  0.064942  0.031722
    4  0.355309  0.524575
    
    [5 rows x 2 columns]
    2013-01-01    0.022321
    2013-01-02    0.227025
    2013-01-03    0.383282
    2013-01-04    0.193225
    2013-01-05    0.110977
    Freq: D, Length: 5, dtype: float64
    
  • pandas.io.gbq provides a simple way to extract from, and load data into, Google’s BigQuery Data Sets by way of pandas DataFrames. BigQuery is a high performance SQL-like database service, useful for performing ad-hoc queries against extremely large datasets. See the docs

    from pandas.io import gbq
    
    # A query to select the average monthly temperatures in the
    # in the year 2000 across the USA. The dataset,
    # publicata:samples.gsod, is available on all BigQuery accounts,
    # and is based on NOAA gsod data.
    
    query = """SELECT station_number as STATION,
    month as MONTH, AVG(mean_temp) as MEAN_TEMP
    FROM publicdata:samples.gsod
    WHERE YEAR = 2000
    GROUP BY STATION, MONTH
    ORDER BY STATION, MONTH ASC"""
    
    # Fetch the result set for this query
    
    # Your Google BigQuery Project ID
    # To find this, see your dashboard:
    # https://console.developers.google.com/iam-admin/projects?authuser=0
    projectid = xxxxxxxxx;
    
    df = gbq.read_gbq(query, project_id = projectid)
    
    # Use pandas to process and reshape the dataset
    
    df2 = df.pivot(index='STATION', columns='MONTH', values='MEAN_TEMP')
    df3 = pandas.concat([df2.min(), df2.mean(), df2.max()],
                        axis=1,keys=["Min Tem", "Mean Temp", "Max Temp"])
    

    The resulting DataFrame is:

    > df3
                Min Tem  Mean Temp    Max Temp
     MONTH
     1     -53.336667  39.827892   89.770968
     2     -49.837500  43.685219   93.437932
     3     -77.926087  48.708355   96.099998
     4     -82.892858  55.070087   97.317240
     5     -92.378261  61.428117  102.042856
     6     -77.703334  65.858888  102.900000
     7     -87.821428  68.169663  106.510714
     8     -89.431999  68.614215  105.500000
     9     -86.611112  63.436935  107.142856
     10    -78.209677  56.880838   92.103333
     11    -50.125000  48.861228   94.996428
     12    -50.332258  42.286879   94.396774
    

    Warning

    To use this module, you will need a BigQuery account. See <https://cloud.google.com/products/big-query> for details.

    As of 10/10/13, there is a bug in Google’s API preventing result sets from being larger than 100,000 rows. A patch is scheduled for the week of 10/14/13.

Internal Refactoring

In 0.13.0 there is a major refactor primarily to subclass Series from NDFrame, which is the base class currently for DataFrame and Panel, to unify methods and behaviors. Series formerly subclassed directly from ndarray. (GH4080, GH3862, GH816)

Warning

There are two potential incompatibilities from < 0.13.0

  • Using certain numpy functions would previously return a Series if passed a Series as an argument. This seems only to affect np.ones_like, np.empty_like, np.diff and np.where. These now return ndarrays.

    In [123]: s = Series([1,2,3,4])
    

    Numpy Usage

    In [124]: np.ones_like(s)
    Out[124]: array([1, 1, 1, 1])
    
    In [125]: np.diff(s)
    Out[125]: array([1, 1, 1])
    
    In [126]: np.where(s>1,s,np.nan)
    Out[126]: array([ nan,   2.,   3.,   4.])
    

    Pandonic Usage

    In [127]: Series(1,index=s.index)
    Out[127]: 
    0    1
    1    1
    2    1
    3    1
    Length: 4, dtype: int64
    
    In [128]: s.diff()
    Out[128]: 
    0    NaN
    1    1.0
    2    1.0
    3    1.0
    Length: 4, dtype: float64
    
    In [129]: s.where(s>1)
    Out[129]: 
    0    NaN
    1    2.0
    2    3.0
    3    4.0
    Length: 4, dtype: float64
    
  • Passing a Series directly to a cython function expecting an ndarray type will no long work directly, you must pass Series.values, See Enhancing Performance

  • Series(0.5) would previously return the scalar 0.5, instead this will return a 1-element Series

  • This change breaks rpy2<=2.3.8. an Issue has been opened against rpy2 and a workaround is detailed in GH5698. Thanks @JanSchulz.

  • Pickle compatibility is preserved for pickles created prior to 0.13. These must be unpickled with pd.read_pickle, see Pickling.

  • Refactor of series.py/frame.py/panel.py to move common code to generic.py

    • added _setup_axes to created generic NDFrame structures
    • moved methods
      • from_axes,_wrap_array,axes,ix,loc,iloc,shape,empty,swapaxes,transpose,pop
      • __iter__,keys,__contains__,__len__,__neg__,__invert__
      • convert_objects,as_blocks,as_matrix,values
      • __getstate__,__setstate__ (compat remains in frame/panel)
      • __getattr__,__setattr__
      • _indexed_same,reindex_like,align,where,mask
      • fillna,replace (Series replace is now consistent with DataFrame)
      • filter (also added axis argument to selectively filter on a different axis)
      • reindex,reindex_axis,take
      • truncate (moved to become part of NDFrame)
  • These are API changes which make Panel more consistent with DataFrame

    • swapaxes on a Panel with the same axes specified now return a copy
    • support attribute access for setting
    • filter supports the same API as the original DataFrame filter
  • Reindex called with no arguments will now return a copy of the input object

  • TimeSeries is now an alias for Series. the property is_time_series can be used to distinguish (if desired)

  • Refactor of Sparse objects to use BlockManager

    • Created a new block type in internals, SparseBlock, which can hold multi-dtypes and is non-consolidatable. SparseSeries and SparseDataFrame now inherit more methods from there hierarchy (Series/DataFrame), and no longer inherit from SparseArray (which instead is the object of the SparseBlock)
    • Sparse suite now supports integration with non-sparse data. Non-float sparse data is supportable (partially implemented)
    • Operations on sparse structures within DataFrames should preserve sparseness, merging type operations will convert to dense (and back to sparse), so might be somewhat inefficient
    • enable setitem on SparseSeries for boolean/integer/slices
    • SparsePanels implementation is unchanged (e.g. not using BlockManager, needs work)
  • added ftypes method to Series/DataFrame, similar to dtypes, but indicates if the underlying is sparse/dense (as well as the dtype)

  • All NDFrame objects can now use __finalize__() to specify various values to propagate to new objects from an existing one (e.g. name in Series will follow more automatically now)

  • Internal type checking is now done via a suite of generated classes, allowing isinstance(value, klass) without having to directly import the klass, courtesy of @jtratner

  • Bug in Series update where the parent frame is not updating its cache based on changes (GH4080) or types (GH3217), fillna (GH3386)

  • Indexing with dtype conversions fixed (GH4463, GH4204)

  • Refactor Series.reindex to core/generic.py (GH4604, GH4618), allow method= in reindexing on a Series to work

  • Series.copy no longer accepts the order parameter and is now consistent with NDFrame copy

  • Refactor rename methods to core/generic.py; fixes Series.rename for (GH4605), and adds rename with the same signature for Panel

  • Refactor clip methods to core/generic.py (GH4798)

  • Refactor of _get_numeric_data/_get_bool_data to core/generic.py, allowing Series/Panel functionality

  • Series (for index) / Panel (for items) now allow attribute access to its elements (GH1903)

    In [130]: s = Series([1,2,3],index=list('abc'))
    
    In [131]: s.b
    Out[131]: 2
    
    In [132]: s.a = 5
    
    In [133]: s
    Out[133]: 
    a    5
    b    2
    c    3
    Length: 3, dtype: int64
    

Bug Fixes

See V0.13.0 Bug Fixes for an extensive list of bugs that have been fixed in 0.13.0.

See the full release notes or issue tracker on GitHub for a complete list of all API changes, Enhancements and Bug Fixes.

v0.12.0 (July 24, 2013)

This is a major release from 0.11.0 and includes several new features and enhancements along with a large number of bug fixes.

Highlights include a consistent I/O API naming scheme, routines to read html, write multi-indexes to csv files, read & write STATA data files, read & write JSON format files, Python 3 support for HDFStore, filtering of groupby expressions via filter, and a revamped replace routine that accepts regular expressions.

API changes

  • The I/O API is now much more consistent with a set of top level reader functions accessed like pd.read_csv() that generally return a pandas object.

    • read_csv
    • read_excel
    • read_hdf
    • read_sql
    • read_json
    • read_html
    • read_stata
    • read_clipboard

    The corresponding writer functions are object methods that are accessed like df.to_csv()

    • to_csv
    • to_excel
    • to_hdf
    • to_sql
    • to_json
    • to_html
    • to_stata
    • to_clipboard
  • Fix modulo and integer division on Series,DataFrames to act similarly to float dtypes to return np.nan or np.inf as appropriate (GH3590). This correct a numpy bug that treats integer and float dtypes differently.

    In [1]: p = DataFrame({ 'first' : [4,5,8], 'second' : [0,0,3] })
    
    In [2]: p % 0
    Out[2]: 
       first  second
    0    NaN     NaN
    1    NaN     NaN
    2    NaN     NaN
    
    [3 rows x 2 columns]
    
    In [3]: p % p
    Out[3]: 
       first  second
    0    0.0     NaN
    1    0.0     NaN
    2    0.0     0.0
    
    [3 rows x 2 columns]
    
    In [4]: p / p
    Out[4]: 
       first  second
    0    1.0     NaN
    1    1.0     NaN
    2    1.0     1.0
    
    [3 rows x 2 columns]
    
    In [5]: p / 0
    Out[5]: 
       first  second
    0    inf     NaN
    1    inf     NaN
    2    inf     inf
    
    [3 rows x 2 columns]
    
  • Add squeeze keyword to groupby to allow reduction from DataFrame -> Series if groups are unique. This is a Regression from 0.10.1. We are reverting back to the prior behavior. This means groupby will return the same shaped objects whether the groups are unique or not. Revert this issue (GH2893) with (GH3596).

    In [6]: df2 = DataFrame([{"val1": 1, "val2" : 20}, {"val1":1, "val2": 19},
       ...:                  {"val1":1, "val2": 27}, {"val1":1, "val2": 12}])
       ...: 
    
    In [7]: def func(dataf):
       ...:     return dataf["val2"]  - dataf["val2"].mean()
       ...: 
    
    # squeezing the result frame to a series (because we have unique groups)
    In [8]: df2.groupby("val1", squeeze=True).apply(func)
    Out[8]: 
    0    0.5
    1   -0.5
    2    7.5
    3   -7.5
    Name: 1, Length: 4, dtype: float64
    
    # no squeezing (the default, and behavior in 0.10.1)
    In [9]: df2.groupby("val1").apply(func)
    Out[9]: 
    val2    0    1    2    3
    val1                    
    1     0.5 -0.5  7.5 -7.5
    
    [1 rows x 4 columns]
    
  • Raise on iloc when boolean indexing with a label based indexer mask e.g. a boolean Series, even with integer labels, will raise. Since iloc is purely positional based, the labels on the Series are not alignable (GH3631)

    This case is rarely used, and there are plently of alternatives. This preserves the iloc API to be purely positional based.

    In [10]: df = DataFrame(lrange(5), list('ABCDE'), columns=['a'])
    
    In [11]: mask = (df.a%2 == 0)
    
    In [12]: mask
    Out[12]: 
    A     True
    B    False
    C     True
    D    False
    E     True
    Name: a, Length: 5, dtype: bool
    
    # this is what you should use
    In [13]: df.loc[mask]
    Out[13]: 
       a
    A  0
    C  2
    E  4
    
    [3 rows x 1 columns]
    
    # this will work as well
    In [14]: df.iloc[mask.values]
    Out[14]: 
       a
    A  0
    C  2
    E  4
    
    [3 rows x 1 columns]
    

    df.iloc[mask] will raise a ValueError

  • The raise_on_error argument to plotting functions is removed. Instead, plotting functions raise a TypeError when the dtype of the object is object to remind you to avoid object arrays whenever possible and thus you should cast to an appropriate numeric dtype if you need to plot something.

  • Add colormap keyword to DataFrame plotting methods. Accepts either a matplotlib colormap object (ie, matplotlib.cm.jet) or a string name of such an object (ie, ‘jet’). The colormap is sampled to select the color for each column. Please see Colormaps for more information. (GH3860)

  • DataFrame.interpolate() is now deprecated. Please use DataFrame.fillna() and DataFrame.replace() instead. (GH3582, GH3675, GH3676)

  • the method and axis arguments of DataFrame.replace() are deprecated

  • DataFrame.replace ‘s infer_types parameter is removed and now performs conversion by default. (GH3907)

  • Add the keyword allow_duplicates to DataFrame.insert to allow a duplicate column to be inserted if True, default is False (same as prior to 0.12) (GH3679)

  • Implement __nonzero__ for NDFrame objects (GH3691, GH3696)

  • IO api

    • added top-level function read_excel to replace the following, The original API is deprecated and will be removed in a future version

      from pandas.io.parsers import ExcelFile
      xls = ExcelFile('path_to_file.xls')
      xls.parse('Sheet1', index_col=None, na_values=['NA'])
      

      With

      import pandas as pd
      pd.read_excel('path_to_file.xls', 'Sheet1', index_col=None, na_values=['NA'])
      
    • added top-level function read_sql that is equivalent to the following

      from pandas.io.sql import read_frame
      read_frame(....)
      
  • DataFrame.to_html and DataFrame.to_latex now accept a path for their first argument (GH3702)

  • Do not allow astypes on datetime64[ns] except to object, and timedelta64[ns] to object/int (GH3425)

  • The behavior of datetime64 dtypes has changed with respect to certain so-called reduction operations (GH3726). The following operations now raise a TypeError when performed on a Series and return an empty Series when performed on a DataFrame similar to performing these operations on, for example, a DataFrame of slice objects:

    • sum, prod, mean, std, var, skew, kurt, corr, and cov
  • read_html now defaults to None when reading, and falls back on bs4 + html5lib when lxml fails to parse. a list of parsers to try until success is also valid

  • The internal pandas class hierarchy has changed (slightly). The previous PandasObject now is called PandasContainer and a new PandasObject has become the baseclass for PandasContainer as well as Index, Categorical, GroupBy, SparseList, and SparseArray (+ their base classes). Currently, PandasObject provides string methods (from StringMixin). (GH4090, GH4092)

  • New StringMixin that, given a __unicode__ method, gets python 2 and python 3 compatible string methods (__str__, __bytes__, and __repr__). Plus string safety throughout. Now employed in many places throughout the pandas library. (GH4090, GH4092)

I/O Enhancements

  • pd.read_html() can now parse HTML strings, files or urls and return DataFrames, courtesy of @cpcloud. (GH3477, GH3605, GH3606, GH3616). It works with a single parser backend: BeautifulSoup4 + html5lib See the docs

    You can use pd.read_html() to read the output from DataFrame.to_html() like so

    In [15]: df = DataFrame({'a': range(3), 'b': list('abc')})
    
    In [16]: print(df)
       a  b
    0  0  a
    1  1  b
    2  2  c
    
    [3 rows x 2 columns]
    
    In [17]: html = df.to_html()
    
    In [18]: alist = pd.read_html(html, index_col=0)
    
    In [19]: print(df == alist[0])
          a     b
    0  True  True
    1  True  True
    2  True  True
    
    [3 rows x 2 columns]
    

    Note that alist here is a Python list so pd.read_html() and DataFrame.to_html() are not inverses.

    • pd.read_html() no longer performs hard conversion of date strings (GH3656).

    Warning

    You may have to install an older version of BeautifulSoup4, See the installation docs

  • Added module for reading and writing Stata files: pandas.io.stata (GH1512) accessible via read_stata top-level function for reading, and to_stata DataFrame method for writing, See the docs

  • Added module for reading and writing json format files: pandas.io.json accessible via read_json top-level function for reading, and to_json DataFrame method for writing, See the docs various issues (GH1226, GH3804, GH3876, GH3867, GH1305)

  • MultiIndex column support for reading and writing csv format files

    • The header option in read_csv now accepts a list of the rows from which to read the index.

    • The option, tupleize_cols can now be specified in both to_csv and read_csv, to provide compatibility for the pre 0.12 behavior of writing and reading MultIndex columns via a list of tuples. The default in 0.12 is to write lists of tuples and not interpret list of tuples as a MultiIndex column.

      Note: The default behavior in 0.12 remains unchanged from prior versions, but starting with 0.13, the default to write and read MultiIndex columns will be in the new format. (GH3571, GH1651, GH3141)

    • If an index_col is not specified (e.g. you don’t have an index, or wrote it with df.to_csv(..., index=False), then any names on the columns index will be lost.

      In [20]: from pandas.util.testing import makeCustomDataframe as mkdf
      
      In [21]: df = mkdf(5, 3, r_idx_nlevels=2, c_idx_nlevels=4)
      
      In [22]: df.to_csv('mi.csv')
      
      In [23]: print(open('mi.csv').read())
      C0,,C_l0_g0,C_l0_g1,C_l0_g2
      C1,,C_l1_g0,C_l1_g1,C_l1_g2
      C2,,C_l2_g0,C_l2_g1,C_l2_g2
      C3,,C_l3_g0,C_l3_g1,C_l3_g2
      R0,R1,,,
      R_l0_g0,R_l1_g0,R0C0,R0C1,R0C2
      R_l0_g1,R_l1_g1,R1C0,R1C1,R1C2
      R_l0_g2,R_l1_g2,R2C0,R2C1,R2C2
      R_l0_g3,R_l1_g3,R3C0,R3C1,R3C2
      R_l0_g4,R_l1_g4,R4C0,R4C1,R4C2
      
      
      In [24]: pd.read_csv('mi.csv', header=[0,1,2,3], index_col=[0,1])
      Out[24]: 
      C0              C_l0_g0 C_l0_g1 C_l0_g2
      C1              C_l1_g0 C_l1_g1 C_l1_g2
      C2              C_l2_g0 C_l2_g1 C_l2_g2
      C3              C_l3_g0 C_l3_g1 C_l3_g2
      R0      R1                             
      R_l0_g0 R_l1_g0    R0C0    R0C1    R0C2
      R_l0_g1 R_l1_g1    R1C0    R1C1    R1C2
      R_l0_g2 R_l1_g2    R2C0    R2C1    R2C2
      R_l0_g3 R_l1_g3    R3C0    R3C1    R3C2
      R_l0_g4 R_l1_g4    R4C0    R4C1    R4C2
      
      [5 rows x 3 columns]
      
  • Support for HDFStore (via PyTables 3.0.0) on Python3

  • Iterator support via read_hdf that automatically opens and closes the store when iteration is finished. This is only for tables

    In [25]: path = 'store_iterator.h5'
    
    In [26]: DataFrame(randn(10,2)).to_hdf(path,'df',table=True)
    
    In [27]: for df in read_hdf(path,'df', chunksize=3):
       ....:    print df
       ....:
              0         1
    0  0.713216 -0.778461
    1 -0.661062  0.862877
    2  0.344342  0.149565
              0         1
    3 -0.626968 -0.875772
    4 -0.930687 -0.218983
    5  0.949965 -0.442354
              0         1
    6 -0.402985  1.111358
    7 -0.241527 -0.670477
    8  0.049355  0.632633
              0         1
    9 -1.502767 -1.225492
    
  • read_csv will now throw a more informative error message when a file contains no columns, e.g., all newline characters

Other Enhancements

  • DataFrame.replace() now allows regular expressions on contained Series with object dtype. See the examples section in the regular docs Replacing via String Expression

    For example you can do

    In [25]: df = DataFrame({'a': list('ab..'), 'b': [1, 2, 3, 4]})
    
    In [26]: df.replace(regex=r'\s*\.\s*', value=np.nan)
    Out[26]: 
         a  b
    0    a  1
    1    b  2
    2  NaN  3
    3  NaN  4
    
    [4 rows x 2 columns]
    

    to replace all occurrences of the string '.' with zero or more instances of surrounding whitespace with NaN.

    Regular string replacement still works as expected. For example, you can do

    In [27]: df.replace('.', np.nan)
    Out[27]: 
         a  b
    0    a  1
    1    b  2
    2  NaN  3
    3  NaN  4
    
    [4 rows x 2 columns]
    

    to replace all occurrences of the string '.' with NaN.

  • pd.melt() now accepts the optional parameters var_name and value_name to specify custom column names of the returned DataFrame.

  • pd.set_option() now allows N option, value pairs (GH3667).

    Let’s say that we had an option 'a.b' and another option 'b.c'. We can set them at the same time:

    In [28]: pd.get_option('a.b')
    Out[28]: 2
    
    In [29]: pd.get_option('b.c')
    Out[29]: 3
    
    In [30]: pd.set_option('a.b', 1, 'b.c', 4)
    
    In [31]: pd.get_option('a.b')
    Out[31]: 1
    
    In [32]: pd.get_option('b.c')
    Out[32]: 4
    
  • The filter method for group objects returns a subset of the original object. Suppose we want to take only elements that belong to groups with a group sum greater than 2.

    In [33]: sf = Series([1, 1, 2, 3, 3, 3])
    
    In [34]: sf.groupby(sf).filter(lambda x: x.sum() > 2)
    Out[34]: 
    3    3
    4    3
    5    3
    Length: 3, dtype: int64
    

    The argument of filter must a function that, applied to the group as a whole, returns True or False.

    Another useful operation is filtering out elements that belong to groups with only a couple members.

    In [35]: dff = DataFrame({'A': np.arange(8), 'B': list('aabbbbcc')})
    
    In [36]: dff.groupby('B').filter(lambda x: len(x) > 2)
    Out[36]: 
       A  B
    2  2  b
    3  3  b
    4  4  b
    5  5  b
    
    [4 rows x 2 columns]
    

    Alternatively, instead of dropping the offending groups, we can return a like-indexed objects where the groups that do not pass the filter are filled with NaNs.

    In [37]: dff.groupby('B').filter(lambda x: len(x) > 2, dropna=False)
    Out[37]: 
         A    B
    0  NaN  NaN
    1  NaN  NaN
    2  2.0    b
    3  3.0    b
    4  4.0    b
    5  5.0    b
    6  NaN  NaN
    7  NaN  NaN
    
    [8 rows x 2 columns]
    
  • Series and DataFrame hist methods now take a figsize argument (GH3834)

  • DatetimeIndexes no longer try to convert mixed-integer indexes during join operations (GH3877)

  • Timestamp.min and Timestamp.max now represent valid Timestamp instances instead of the default datetime.min and datetime.max (respectively), thanks @SleepingPills

  • read_html now raises when no tables are found and BeautifulSoup==4.2.0 is detected (GH4214)

Experimental Features

  • Added experimental CustomBusinessDay class to support DateOffsets with custom holiday calendars and custom weekmasks. (GH2301)

    Note

    This uses the numpy.busdaycalendar API introduced in Numpy 1.7 and therefore requires Numpy 1.7.0 or newer.

    In [38]: from pandas.tseries.offsets import CustomBusinessDay
    
    In [39]: from datetime import datetime
    
    # As an interesting example, let's look at Egypt where
    # a Friday-Saturday weekend is observed.
    In [40]: weekmask_egypt = 'Sun Mon Tue Wed Thu'
    
    # They also observe International Workers' Day so let's
    # add that for a couple of years
    In [41]: holidays = ['2012-05-01', datetime(2013, 5, 1), np.datetime64('2014-05-01')]
    
    In [42]: bday_egypt = CustomBusinessDay(holidays=holidays, weekmask=weekmask_egypt)
    
    In [43]: dt = datetime(2013, 4, 30)
    
    In [44]: print(dt + 2 * bday_egypt)
    2013-05-05 00:00:00
    
    In [45]: dts = date_range(dt, periods=5, freq=bday_egypt)
    
    In [46]: print(Series(dts.weekday, dts).map(Series('Mon Tue Wed Thu Fri Sat Sun'.split())))
    2013-04-30    Tue
    2013-05-02    Thu
    2013-05-05    Sun
    2013-05-06    Mon
    2013-05-07    Tue
    Freq: C, Length: 5, dtype: object
    

Bug Fixes

  • Plotting functions now raise a TypeError before trying to plot anything if the associated objects have have a dtype of object (GH1818, GH3572, GH3911, GH3912), but they will try to convert object arrays to numeric arrays if possible so that you can still plot, for example, an object array with floats. This happens before any drawing takes place which elimnates any spurious plots from showing up.

  • fillna methods now raise a TypeError if the value parameter is a list or tuple.

  • Series.str now supports iteration (GH3638). You can iterate over the individual elements of each string in the Series. Each iteration yields yields a Series with either a single character at each index of the original Series or NaN. For example,

    In [47]: strs = 'go', 'bow', 'joe', 'slow'
    
    In [48]: ds = Series(strs)
    
    In [49]: for s in ds.str:
       ....:     print(s)
       ....: 
    0    g
    1    b
    2    j
    3    s
    Length: 4, dtype: object
    0    o
    1    o
    2    o
    3    l
    Length: 4, dtype: object
    0    NaN
    1      w
    2      e
    3      o
    Length: 4, dtype: object
    0    NaN
    1    NaN
    2    NaN
    3      w
    Length: 4, dtype: object
    
    In [50]: s
    Out[50]: 
    0    NaN
    1    NaN
    2    NaN
    3      w
    Length: 4, dtype: object
    
    In [51]: s.dropna().values.item() == 'w'
    Out[51]: True
    

    The last element yielded by the iterator will be a Series containing the last element of the longest string in the Series with all other elements being NaN. Here since 'slow' is the longest string and there are no other strings with the same length 'w' is the only non-null string in the yielded Series.

  • HDFStore

    • will retain index attributes (freq,tz,name) on recreation (GH3499)
    • will warn with a AttributeConflictWarning if you are attempting to append an index with a different frequency than the existing, or attempting to append an index with a different name than the existing
    • support datelike columns with a timezone as data_columns (GH2852)
  • Non-unique index support clarified (GH3468).

    • Fix assigning a new index to a duplicate index in a DataFrame would fail (GH3468)
    • Fix construction of a DataFrame with a duplicate index
    • ref_locs support to allow duplicative indices across dtypes, allows iget support to always find the index (even across dtypes) (GH2194)
    • applymap on a DataFrame with a non-unique index now works (removed warning) (GH2786), and fix (GH3230)
    • Fix to_csv to handle non-unique columns (GH3495)
    • Duplicate indexes with getitem will return items in the correct order (GH3455, GH3457) and handle missing elements like unique indices (GH3561)
    • Duplicate indexes with and empty DataFrame.from_records will return a correct frame (GH3562)
    • Concat to produce a non-unique columns when duplicates are across dtypes is fixed (GH3602)
    • Allow insert/delete to non-unique columns (GH3679)
    • Non-unique indexing with a slice via loc and friends fixed (GH3659)
    • Allow insert/delete to non-unique columns (GH3679)
    • Extend reindex to correctly deal with non-unique indices (GH3679)
    • DataFrame.itertuples() now works with frames with duplicate column names (GH3873)
    • Bug in non-unique indexing via iloc (GH4017); added takeable argument to reindex for location-based taking
    • Allow non-unique indexing in series via .ix/.loc and __getitem__ (GH4246)
    • Fixed non-unique indexing memory allocation issue with .ix/.loc (GH4280)
  • DataFrame.from_records did not accept empty recarrays (GH3682)

  • read_html now correctly skips tests (GH3741)

  • Fixed a bug where DataFrame.replace with a compiled regular expression in the to_replace argument wasn’t working (GH3907)

  • Improved network test decorator to catch IOError (and therefore URLError as well). Added with_connectivity_check decorator to allow explicitly checking a website as a proxy for seeing if there is network connectivity. Plus, new optional_args decorator factory for decorators. (GH3910, GH3914)

  • Fixed testing issue where too many sockets where open thus leading to a connection reset issue (GH3982, GH3985, GH4028, GH4054)

  • Fixed failing tests in test_yahoo, test_google where symbols were not retrieved but were being accessed (GH3982, GH3985, GH4028, GH4054)

  • Series.hist will now take the figure from the current environment if one is not passed

  • Fixed bug where a 1xN DataFrame would barf on a 1xN mask (GH4071)

  • Fixed running of tox under python3 where the pickle import was getting rewritten in an incompatible way (GH4062, GH4063)

  • Fixed bug where sharex and sharey were not being passed to grouped_hist (GH4089)

  • Fixed bug in DataFrame.replace where a nested dict wasn’t being iterated over when regex=False (GH4115)

  • Fixed bug in the parsing of microseconds when using the format argument in to_datetime (GH4152)

  • Fixed bug in PandasAutoDateLocator where invert_xaxis triggered incorrectly MilliSecondLocator (GH3990)

  • Fixed bug in plotting that wasn’t raising on invalid colormap for matplotlib 1.1.1 (GH4215)

  • Fixed the legend displaying in DataFrame.plot(kind='kde') (GH4216)

  • Fixed bug where Index slices weren’t carrying the name attribute (GH4226)

  • Fixed bug in initializing DatetimeIndex with an array of strings in a certain time zone (GH4229)

  • Fixed bug where html5lib wasn’t being properly skipped (GH4265)

  • Fixed bug where get_data_famafrench wasn’t using the correct file edges (GH4281)

See the full release notes or issue tracker on GitHub for a complete list.

v0.11.0 (April 22, 2013)

This is a major release from 0.10.1 and includes many new features and enhancements along with a large number of bug fixes. The methods of Selecting Data have had quite a number of additions, and Dtype support is now full-fledged. There are also a number of important API changes that long-time pandas users should pay close attention to.

There is a new section in the documentation, 10 Minutes to Pandas, primarily geared to new users.

There is a new section in the documentation, Cookbook, a collection of useful recipes in pandas (and that we want contributions!).

There are several libraries that are now Recommended Dependencies

Selection Choices

Starting in 0.11.0, object selection has had a number of user-requested additions in order to support more explicit location based indexing. Pandas now supports three types of multi-axis indexing.

  • .loc is strictly label based, will raise KeyError when the items are not found, allowed inputs are:

    • A single label, e.g. 5 or 'a', (note that 5 is interpreted as a label of the index. This use is not an integer position along the index)
    • A list or array of labels ['a', 'b', 'c']
    • A slice object with labels 'a':'f', (note that contrary to usual python slices, both the start and the stop are included!)
    • A boolean array

    See more at Selection by Label

  • .iloc is strictly integer position based (from 0 to length-1 of the axis), will raise IndexError when the requested indicies are out of bounds. Allowed inputs are:

    • An integer e.g. 5
    • A list or array of integers [4, 3, 0]
    • A slice object with ints 1:7
    • A boolean array

    See more at Selection by Position

  • .ix supports mixed integer and label based access. It is primarily label based, but will fallback to integer positional access. .ix is the most general and will support any of the inputs to .loc and .iloc, as well as support for floating point label schemes. .ix is especially useful when dealing with mixed positional and label based hierarchial indexes.

    As using integer slices with .ix have different behavior depending on whether the slice is interpreted as position based or label based, it’s usually better to be explicit and use .iloc or .loc.

    See more at Advanced Indexing and Advanced Hierarchical.

Selection Deprecations

Starting in version 0.11.0, these methods may be deprecated in future versions.

  • irow
  • icol
  • iget_value

See the section Selection by Position for substitutes.

Dtypes

Numeric dtypes will propagate and can coexist in DataFrames. If a dtype is passed (either directly via the dtype keyword, a passed ndarray, or a passed Series, then it will be preserved in DataFrame operations. Furthermore, different numeric dtypes will NOT be combined. The following example will give you a taste.

In [1]: df1 = DataFrame(randn(8, 1), columns = ['A'], dtype = 'float32')

In [2]: df1
Out[2]: 
          A
0  1.392665
1 -0.123497
2 -0.402761
3 -0.246604
4 -0.288433
5 -0.763434
6  2.069526
7 -1.203569

[8 rows x 1 columns]

In [3]: df1.dtypes
Out[3]: 
A    float32
Length: 1, dtype: object

In [4]: df2 = DataFrame(dict( A = Series(randn(8),dtype='float16'),
   ...:                       B = Series(randn(8)),
   ...:                       C = Series(randn(8),dtype='uint8') ))
   ...: 

In [5]: df2
Out[5]: 
          A         B    C
0  0.591797 -0.038605    0
1  0.841309 -0.460478    1
2 -0.500977 -0.310458    0
3 -0.816406  0.866493  254
4 -0.207031  0.245972    0
5 -0.664062  0.319442    1
6  0.580566  1.378512    1
7 -0.965820  0.292502  255

[8 rows x 3 columns]

In [6]: df2.dtypes
Out[6]: 
A    float16
B    float64
C      uint8
Length: 3, dtype: object

# here you get some upcasting
In [7]: df3 = df1.reindex_like(df2).fillna(value=0.0) + df2

In [8]: df3
Out[8]: 
          A         B      C
0  1.984462 -0.038605    0.0
1  0.717812 -0.460478    1.0
2 -0.903737 -0.310458    0.0
3 -1.063011  0.866493  254.0
4 -0.495465  0.245972    0.0
5 -1.427497  0.319442    1.0
6  2.650092  1.378512    1.0
7 -2.169390  0.292502  255.0

[8 rows x 3 columns]

In [9]: df3.dtypes
Out[9]: 
A    float32
B    float64
C    float64
Length: 3, dtype: object

Dtype Conversion

This is lower-common-denominator upcasting, meaning you get the dtype which can accommodate all of the types

In [10]: df3.values.dtype
Out[10]: dtype('float64')

Conversion

In [11]: df3.astype('float32').dtypes
Out[11]: 
A    float32
B    float32
C    float32
Length: 3, dtype: object

Mixed Conversion

In [12]: df3['D'] = '1.'

In [13]: df3['E'] = '1'

In [14]: df3.convert_objects(convert_numeric=True).dtypes
Out[14]: 
A    float32
B    float64
C    float64
D    float64
E      int64
Length: 5, dtype: object

# same, but specific dtype conversion
In [15]: df3['D'] = df3['D'].astype('float16')

In [16]: df3['E'] = df3['E'].astype('int32')

In [17]: df3.dtypes
Out[17]: 
A    float32
B    float64
C    float64
D    float16
E      int32
Length: 5, dtype: object

Forcing Date coercion (and setting NaT when not datelike)

In [18]: from datetime import datetime

In [19]: s = Series([datetime(2001,1,1,0,0), 'foo', 1.0, 1,
   ....:             Timestamp('20010104'), '20010105'],dtype='O')
   ....: 

In [20]: s.convert_objects(convert_dates='coerce')
Out[20]: 
0   2001-01-01
1          NaT
2          NaT
3          NaT
4   2001-01-04
5   2001-01-05
Length: 6, dtype: datetime64[ns]

Dtype Gotchas

Platform Gotchas

Starting in 0.11.0, construction of DataFrame/Series will use default dtypes of int64 and float64, regardless of platform. This is not an apparent change from earlier versions of pandas. If you specify dtypes, they WILL be respected, however (GH2837)

The following will all result in int64 dtypes

In [21]: DataFrame([1,2],columns=['a']).dtypes
Out[21]: 
a    int64
Length: 1, dtype: object

In [22]: DataFrame({'a' : [1,2] }).dtypes
Out[22]: 
a    int64
Length: 1, dtype: object

In [23]: DataFrame({'a' : 1 }, index=range(2)).dtypes
Out[23]: 
a    int64
Length: 1, dtype: object

Keep in mind that DataFrame(np.array([1,2])) WILL result in int32 on 32-bit platforms!

Upcasting Gotchas

Performing indexing operations on integer type data can easily upcast the data. The dtype of the input data will be preserved in cases where nans are not introduced.

In [24]: dfi = df3.astype('int32')

In [25]: dfi['D'] = dfi['D'].astype('int64')

In [26]: dfi
Out[26]: 
   A  B    C  D  E
0  1  0    0  1  1
1  0  0    1  1  1
2  0  0    0  1  1
3 -1  0  254  1  1
4  0  0    0  1  1
5 -1  0    1  1  1
6  2  1    1  1  1
7 -2  0  255  1  1

[8 rows x 5 columns]

In [27]: dfi.dtypes
Out[27]: 
A    int32
B    int32
C    int32
D    int64
E    int32
Length: 5, dtype: object

In [28]: casted = dfi[dfi>0]

In [29]: casted
Out[29]: 
     A    B      C  D  E
0  1.0  NaN    NaN  1  1
1  NaN  NaN    1.0  1  1
2  NaN  NaN    NaN  1  1
3  NaN  NaN  254.0  1  1
4  NaN  NaN    NaN  1  1
5  NaN  NaN    1.0  1  1
6  2.0  1.0    1.0  1  1
7  NaN  NaN  255.0  1  1

[8 rows x 5 columns]

In [30]: casted.dtypes
Out[30]: 
A    float64
B    float64
C    float64
D      int64
E      int32
Length: 5, dtype: object

While float dtypes are unchanged.

In [31]: df4 = df3.copy()

In [32]: df4['A'] = df4['A'].astype('float32')

In [33]: df4.dtypes
Out[33]: 
A    float32
B    float64
C    float64
D    float16
E      int32
Length: 5, dtype: object

In [34]: casted = df4[df4>0]

In [35]: casted
Out[35]: 
          A         B      C    D  E
0  1.984462       NaN    NaN  1.0  1
1  0.717812       NaN    1.0  1.0  1
2       NaN       NaN    NaN  1.0  1
3       NaN  0.866493  254.0  1.0  1
4       NaN  0.245972    NaN  1.0  1
5       NaN  0.319442    1.0  1.0  1
6  2.650092  1.378512    1.0  1.0  1
7       NaN  0.292502  255.0  1.0  1

[8 rows x 5 columns]

In [36]: casted.dtypes
Out[36]: 
A    float32
B    float64
C    float64
D    float16
E      int32
Length: 5, dtype: object

Datetimes Conversion

Datetime64[ns] columns in a DataFrame (or a Series) allow the use of np.nan to indicate a nan value, in addition to the traditional NaT, or not-a-time. This allows convenient nan setting in a generic way. Furthermore datetime64[ns] columns are created by default, when passed datetimelike objects (this change was introduced in 0.10.1) (GH2809, GH2810)

In [37]: df = DataFrame(randn(6,2),date_range('20010102',periods=6),columns=['A','B'])

In [38]: df['timestamp'] = Timestamp('20010103')

In [39]: df
Out[39]: 
                   A         B  timestamp
2001-01-02  1.023958  0.660103 2001-01-03
2001-01-03  1.236475 -2.170629 2001-01-03
2001-01-04 -0.270630 -1.685677 2001-01-03
2001-01-05 -0.440747 -0.115070 2001-01-03
2001-01-06 -0.632102 -0.585977 2001-01-03
2001-01-07 -1.444787 -0.201135 2001-01-03

[6 rows x 3 columns]

# datetime64[ns] out of the box
In [40]: df.get_dtype_counts()
Out[40]: 
float64           2
datetime64[ns]    1
Length: 2, dtype: int64

# use the traditional nan, which is mapped to NaT internally
In [41]: df.loc[df.index[2:4], ['A','timestamp']] = np.nan

In [42]: df
Out[42]: 
                   A         B  timestamp
2001-01-02  1.023958  0.660103 2001-01-03
2001-01-03  1.236475 -2.170629 2001-01-03
2001-01-04       NaN -1.685677        NaT
2001-01-05       NaN -0.115070        NaT
2001-01-06 -0.632102 -0.585977 2001-01-03
2001-01-07 -1.444787 -0.201135 2001-01-03

[6 rows x 3 columns]

Astype conversion on datetime64[ns] to object, implicitly converts NaT to np.nan

In [43]: import datetime

In [44]: s = Series([datetime.datetime(2001, 1, 2, 0, 0) for i in range(3)])

In [45]: s.dtype
Out[45]: dtype('<M8[ns]')

In [46]: s[1] = np.nan

In [47]: s
Out[47]: 
0   2001-01-02
1          NaT
2   2001-01-02
Length: 3, dtype: datetime64[ns]

In [48]: s.dtype
Out[48]: dtype('<M8[ns]')

In [49]: s = s.astype('O')

In [50]: s
Out[50]: 
0    2001-01-02 00:00:00
1                    NaT
2    2001-01-02 00:00:00
Length: 3, dtype: object

In [51]: s.dtype
Out[51]: dtype('O')

API changes

  • Added to_series() method to indicies, to facilitate the creation of indexers (GH3275)
  • HDFStore
    • added the method select_column to select a single column from a table as a Series.
    • deprecated the unique method, can be replicated by select_column(key,column).unique()
    • min_itemsize parameter to append will now automatically create data_columns for passed keys

Enhancements

  • Improved performance of df.to_csv() by up to 10x in some cases. (GH3059)

  • Numexpr is now a Recommended Dependencies, to accelerate certain types of numerical and boolean operations

  • Bottleneck is now a Recommended Dependencies, to accelerate certain types of nan operations

  • HDFStore

    • support read_hdf/to_hdf API similar to read_csv/to_csv

      In [52]: df = DataFrame(dict(A=lrange(5), B=lrange(5)))
      
      In [53]: df.to_hdf('store.h5','table',append=True)
      
      In [54]: read_hdf('store.h5', 'table', where = ['index>2'])
      Out[54]: 
         A  B
      3  3  3
      4  4  4
      
      [2 rows x 2 columns]
      
    • provide dotted attribute access to get from stores, e.g. store.df == store['df']

    • new keywords iterator=boolean, and chunksize=number_in_a_chunk are provided to support iteration on select and select_as_multiple (GH3076)

  • You can now select timestamps from an unordered timeseries similarly to an ordered timeseries (GH2437)

  • You can now select with a string from a DataFrame with a datelike index, in a similar way to a Series (GH3070)

    In [55]: idx = date_range("2001-10-1", periods=5, freq='M')
    
    In [56]: ts = Series(np.random.rand(len(idx)),index=idx)
    
    In [57]: ts['2001']
    Out[57]: 
    2001-10-31    0.663256
    2001-11-30    0.079126
    2001-12-31    0.587699
    Freq: M, Length: 3, dtype: float64
    
    In [58]: df = DataFrame(dict(A = ts))
    
    In [59]: df['2001']
    Out[59]: 
                       A
    2001-10-31  0.663256
    2001-11-30  0.079126
    2001-12-31  0.587699
    
    [3 rows x 1 columns]
    
  • Squeeze to possibly remove length 1 dimensions from an object.

    In [60]: p = Panel(randn(3,4,4),items=['ItemA','ItemB','ItemC'],
       ....:                    major_axis=date_range('20010102',periods=4),
       ....:                    minor_axis=['A','B','C','D'])
       ....: 
    
    In [61]: p
    Out[61]: 
    <class 'pandas.core.panel.Panel'>
    Dimensions: 3 (items) x 4 (major_axis) x 4 (minor_axis)
    Items axis: ItemA to ItemC
    Major_axis axis: 2001-01-02 00:00:00 to 2001-01-05 00:00:00
    Minor_axis axis: A to D
    
    In [62]: p.reindex(items=['ItemA']).squeeze()
    Out[62]: 
                       A         B         C         D
    2001-01-02 -1.203403  0.425882 -0.436045 -0.982462
    2001-01-03  0.348090 -0.969649  0.121731  0.202798
    2001-01-04  1.215695 -0.218549 -0.631381 -0.337116
    2001-01-05  0.404238  0.907213 -0.865657  0.483186
    
    [4 rows x 4 columns]
    
    In [63]: p.reindex(items=['ItemA'],minor=['B']).squeeze()
    Out[63]: 
    2001-01-02    0.425882
    2001-01-03   -0.969649
    2001-01-04   -0.218549
    2001-01-05    0.907213
    Freq: D, Name: B, Length: 4, dtype: float64
    
  • In pd.io.data.Options,

    • Fix bug when trying to fetch data for the current month when already past expiry.
    • Now using lxml to scrape html instead of BeautifulSoup (lxml was faster).
    • New instance variables for calls and puts are automatically created when a method that creates them is called. This works for current month where the instance variables are simply calls and puts. Also works for future expiry months and save the instance variable as callsMMYY or putsMMYY, where MMYY are, respectively, the month and year of the option’s expiry.
    • Options.get_near_stock_price now allows the user to specify the month for which to get relevant options data.
    • Options.get_forward_data now has optional kwargs near and above_below. This allows the user to specify if they would like to only return forward looking data for options near the current stock price. This just obtains the data from Options.get_near_stock_price instead of Options.get_xxx_data() (GH2758).
  • Cursor coordinate information is now displayed in time-series plots.

  • added option display.max_seq_items to control the number of elements printed per sequence pprinting it. (GH2979)

  • added option display.chop_threshold to control display of small numerical values. (GH2739)

  • added option display.max_info_rows to prevent verbose_info from being calculated for frames above 1M rows (configurable). (GH2807, GH2918)

  • value_counts() now accepts a “normalize” argument, for normalized histograms. (GH2710).

  • DataFrame.from_records now accepts not only dicts but any instance of the collections.Mapping ABC.

  • added option display.mpl_style providing a sleeker visual style for plots. Based on https://gist.github.com/huyng/816622 (GH3075).

  • Treat boolean values as integers (values 1 and 0) for numeric operations. (GH2641)

  • to_html() now accepts an optional “escape” argument to control reserved HTML character escaping (enabled by default) and escapes &, in addition to < and >. (GH2919)

See the full release notes or issue tracker on GitHub for a complete list.

v0.10.1 (January 22, 2013)

This is a minor release from 0.10.0 and includes new features, enhancements, and bug fixes. In particular, there is substantial new HDFStore functionality contributed by Jeff Reback.

An undesired API breakage with functions taking the inplace option has been reverted and deprecation warnings added.

API changes

  • Functions taking an inplace option return the calling object as before. A deprecation message has been added
  • Groupby aggregations Max/Min no longer exclude non-numeric data (GH2700)
  • Resampling an empty DataFrame now returns an empty DataFrame instead of raising an exception (GH2640)
  • The file reader will now raise an exception when NA values are found in an explicitly specified integer column instead of converting the column to float (GH2631)
  • DatetimeIndex.unique now returns a DatetimeIndex with the same name and
  • timezone instead of an array (GH2563)

New features

  • MySQL support for database (contribution from Dan Allan)

HDFStore

You may need to upgrade your existing data files. Please visit the compatibility section in the main docs.

You can designate (and index) certain columns that you want to be able to perform queries on a table, by passing a list to data_columns

In [1]: store = HDFStore('store.h5')

In [2]: df = DataFrame(randn(8, 3), index=date_range('1/1/2000', periods=8),
   ...:            columns=['A', 'B', 'C'])
   ...: 

In [3]: df['string'] = 'foo'

In [4]: df.loc[df.index[4:6], 'string'] = np.nan

In [5]: df.loc[df.index[7:9], 'string'] = 'bar'

In [6]: df['string2'] = 'cool'

In [7]: df
Out[7]: 
                   A         B         C string string2
2000-01-01  1.885136 -0.183873  2.550850    foo    cool
2000-01-02  0.180759 -1.117089  0.061462    foo    cool
2000-01-03 -0.294467 -0.591411 -0.876691    foo    cool
2000-01-04  3.127110  1.451130  0.045152    foo    cool
2000-01-05 -0.242846  1.195819  1.533294    NaN    cool
2000-01-06  0.820521 -0.281201  1.651561    NaN    cool
2000-01-07 -0.034086  0.252394 -0.498772    foo    cool
2000-01-08 -2.290958 -1.601262 -0.256718    bar    cool

[8 rows x 5 columns]

# on-disk operations
In [8]: store.append('df', df, data_columns = ['B','C','string','string2'])

In [9]: store.select('df', "B>0 and string=='foo'")
Out[9]: 
                   A         B         C string string2
2000-01-04  3.127110  1.451130  0.045152    foo    cool
2000-01-07 -0.034086  0.252394 -0.498772    foo    cool

[2 rows x 5 columns]

# this is in-memory version of this type of selection
In [10]: df[(df.B > 0) & (df.string == 'foo')]
Out[10]: 
                   A         B         C string string2
2000-01-04  3.127110  1.451130  0.045152    foo    cool
2000-01-07 -0.034086  0.252394 -0.498772    foo    cool

[2 rows x 5 columns]

Retrieving unique values in an indexable or data column.

# note that this is deprecated as of 0.14.0
# can be replicated by: store.select_column('df','index').unique()
store.unique('df','index')
store.unique('df','string')

You can now store datetime64 in data columns

In [11]: df_mixed               = df.copy()

In [12]: df_mixed['datetime64'] = Timestamp('20010102')

In [13]: df_mixed.loc[df_mixed.index[3:4], ['A','B']] = np.nan

In [14]: store.append('df_mixed', df_mixed)

In [15]: df_mixed1 = store.select('df_mixed')

In [16]: df_mixed1
Out[16]: 
                   A         B         C string string2 datetime64
2000-01-01  1.885136 -0.183873  2.550850    foo    cool 2001-01-02
2000-01-02  0.180759 -1.117089  0.061462    foo    cool 2001-01-02
2000-01-03 -0.294467 -0.591411 -0.876691    foo    cool 2001-01-02
2000-01-04       NaN       NaN  0.045152    foo    cool 2001-01-02
2000-01-05 -0.242846  1.195819  1.533294    NaN    cool 2001-01-02
2000-01-06  0.820521 -0.281201  1.651561    NaN    cool 2001-01-02
2000-01-07 -0.034086  0.252394 -0.498772    foo    cool 2001-01-02
2000-01-08 -2.290958 -1.601262 -0.256718    bar    cool 2001-01-02

[8 rows x 6 columns]

In [17]: df_mixed1.get_dtype_counts()
Out[17]: 
float64           3
object            2
datetime64[ns]    1
Length: 3, dtype: int64

You can pass columns keyword to select to filter a list of the return columns, this is equivalent to passing a Term('columns',list_of_columns_to_filter)

In [18]: store.select('df',columns = ['A','B'])
Out[18]: 
                   A         B
2000-01-01  1.885136 -0.183873
2000-01-02  0.180759 -1.117089
2000-01-03 -0.294467 -0.591411
2000-01-04  3.127110  1.451130
2000-01-05 -0.242846  1.195819
2000-01-06  0.820521 -0.281201
2000-01-07 -0.034086  0.252394
2000-01-08 -2.290958 -1.601262

[8 rows x 2 columns]

HDFStore now serializes multi-index dataframes when appending tables.

In [19]: index = MultiIndex(levels=[['foo', 'bar', 'baz', 'qux'],
   ....:                            ['one', 'two', 'three']],
   ....:                    labels=[[0, 0, 0, 1, 1, 2, 2, 3, 3, 3],
   ....:                            [0, 1, 2, 0, 1, 1, 2, 0, 1, 2]],
   ....:                    names=['foo', 'bar'])
   ....: 

In [20]: df = DataFrame(np.random.randn(10, 3), index=index,
   ....:                columns=['A', 'B', 'C'])
   ....: 

In [21]: df
Out[21]: 
                  A         B         C
foo bar                                
foo one    0.239369  0.174122 -1.131794
    two   -1.948006  0.980347 -0.674429
    three -0.361633 -0.761218  1.768215
bar one    0.152288 -0.862613 -0.210968
    two   -0.859278  1.498195  0.462413
baz two   -0.647604  1.511487 -0.727189
    three -0.342928 -0.007364  1.427674
qux one    0.104020  2.052171 -1.230963
    two   -0.019240 -1.713238  0.838912
    three -0.637855  0.215109 -1.515362

[10 rows x 3 columns]

In [22]: store.append('mi',df)

In [23]: store.select('mi')
Out[23]: 
                  A         B         C
foo bar                                
foo one    0.239369  0.174122 -1.131794
    two   -1.948006  0.980347 -0.674429
    three -0.361633 -0.761218  1.768215
bar one    0.152288 -0.862613 -0.210968
    two   -0.859278  1.498195  0.462413
baz two   -0.647604  1.511487 -0.727189
    three -0.342928 -0.007364  1.427674
qux one    0.104020  2.052171 -1.230963
    two   -0.019240 -1.713238  0.838912
    three -0.637855  0.215109 -1.515362

[10 rows x 3 columns]

# the levels are automatically included as data columns
In [24]: store.select('mi', "foo='bar'")
Out[24]: 
                A         B         C
foo bar                              
bar one  0.152288 -0.862613 -0.210968
    two -0.859278  1.498195  0.462413

[2 rows x 3 columns]

Multi-table creation via append_to_multiple and selection via select_as_multiple can create/select from multiple tables and return a combined result, by using where on a selector table.

In [25]: df_mt = DataFrame(randn(8, 6), index=date_range('1/1/2000', periods=8),
   ....:                                columns=['A', 'B', 'C', 'D', 'E', 'F'])
   ....: 

In [26]: df_mt['foo'] = 'bar'

# you can also create the tables individually
In [27]: store.append_to_multiple({ 'df1_mt' : ['A','B'], 'df2_mt' : None }, df_mt, selector = 'df1_mt')

In [28]: store
Out[28]: 
<class 'pandas.io.pytables.HDFStore'>
File path: store.h5

# indiviual tables were created
In [29]: store.select('df1_mt')
Out[29]: 
                   A         B
2000-01-01  1.586924 -0.447974
2000-01-02 -0.102206  0.870302
2000-01-03  1.249874  1.458210
2000-01-04 -0.616293  0.150468
2000-01-05 -0.431163  0.016640
2000-01-06  0.800353 -0.451572
2000-01-07  1.239198  0.185437
2000-01-08 -0.040863  0.290110

[8 rows x 2 columns]

In [30]: store.select('df2_mt')
Out[30]: 
                   C         D         E         F  foo
2000-01-01 -1.573998  0.630925 -0.071659 -1.277640  bar
2000-01-02  1.275280 -1.199212  1.060780  1.673018  bar
2000-01-03 -0.710542  0.825392  1.557329  1.993441  bar
2000-01-04  0.132104  0.580923 -0.128750  1.445964  bar
2000-01-05  0.904578 -1.645852 -0.688741  0.228006  bar
2000-01-06  0.831767  0.228760  0.932498 -2.200069  bar
2000-01-07 -0.540770 -0.370038  1.298390  1.662964  bar
2000-01-08 -0.096145  1.717830 -0.462446 -0.112019  bar

[8 rows x 5 columns]

# as a multiple
In [31]: store.select_as_multiple(['df1_mt','df2_mt'], where = [ 'A>0','B>0' ], selector = 'df1_mt')
Out[31]: 
                   A         B         C         D         E         F  foo
2000-01-03  1.249874  1.458210 -0.710542  0.825392  1.557329  1.993441  bar
2000-01-07  1.239198  0.185437 -0.540770 -0.370038  1.298390  1.662964  bar

[2 rows x 7 columns]

Enhancements

  • HDFStore now can read native PyTables table format tables
  • You can pass nan_rep = 'my_nan_rep' to append, to change the default nan representation on disk (which converts to/from np.nan), this defaults to nan.
  • You can pass index to append. This defaults to True. This will automagically create indicies on the indexables and data columns of the table
  • You can pass chunksize=an integer to append, to change the writing chunksize (default is 50000). This will significantly lower your memory usage on writing.
  • You can pass expectedrows=an integer to the first append, to set the TOTAL number of expectedrows that PyTables will expected. This will optimize read/write performance.
  • Select now supports passing start and stop to provide selection space limiting in selection.
  • Greatly improved ISO8601 (e.g., yyyy-mm-dd) date parsing for file parsers (GH2698)
  • Allow DataFrame.merge to handle combinatorial sizes too large for 64-bit integer (GH2690)
  • Series now has unary negation (-series) and inversion (~series) operators (GH2686)
  • DataFrame.plot now includes a logx parameter to change the x-axis to log scale (GH2327)
  • Series arithmetic operators can now handle constant and ndarray input (GH2574)
  • ExcelFile now takes a kind argument to specify the file type (GH2613)
  • A faster implementation for Series.str methods (GH2602)

Bug Fixes

  • HDFStore tables can now store float32 types correctly (cannot be mixed with float64 however)
  • Fixed Google Analytics prefix when specifying request segment (GH2713).
  • Function to reset Google Analytics token store so users can recover from improperly setup client secrets (GH2687).
  • Fixed groupby bug resulting in segfault when passing in MultiIndex (GH2706)
  • Fixed bug where passing a Series with datetime64 values into to_datetime results in bogus output values (GH2699)
  • Fixed bug in pattern in HDFStore expressions when pattern is not a valid regex (GH2694)
  • Fixed performance issues while aggregating boolean data (GH2692)
  • When given a boolean mask key and a Series of new values, Series __setitem__ will now align the incoming values with the original Series (GH2686)
  • Fixed MemoryError caused by performing counting sort on sorting MultiIndex levels with a very large number of combinatorial values (GH2684)
  • Fixed bug that causes plotting to fail when the index is a DatetimeIndex with a fixed-offset timezone (GH2683)
  • Corrected businessday subtraction logic when the offset is more than 5 bdays and the starting date is on a weekend (GH2680)
  • Fixed C file parser behavior when the file has more columns than data (GH2668)
  • Fixed file reader bug that misaligned columns with data in the presence of an implicit column and a specified usecols value
  • DataFrames with numerical or datetime indices are now sorted prior to plotting (GH2609)
  • Fixed DataFrame.from_records error when passed columns, index, but empty records (GH2633)
  • Several bug fixed for Series operations when dtype is datetime64 (GH2689, GH2629, GH2626)

See the full release notes or issue tracker on GitHub for a complete list.

v0.10.0 (December 17, 2012)

This is a major release from 0.9.1 and includes many new features and enhancements along with a large number of bug fixes. There are also a number of important API changes that long-time pandas users should pay close attention to.

File parsing new features

The delimited file parsing engine (the guts of read_csv and read_table) has been rewritten from the ground up and now uses a fraction the amount of memory while parsing, while being 40% or more faster in most use cases (in some cases much faster).

There are also many new features:

  • Much-improved Unicode handling via the encoding option.
  • Column filtering (usecols)
  • Dtype specification (dtype argument)
  • Ability to specify strings to be recognized as True/False
  • Ability to yield NumPy record arrays (as_recarray)
  • High performance delim_whitespace option
  • Decimal format (e.g. European format) specification
  • Easier CSV dialect options: escapechar, lineterminator, quotechar, etc.
  • More robust handling of many exceptional kinds of files observed in the wild

API changes

Deprecated DataFrame BINOP TimeSeries special case behavior

The default behavior of binary operations between a DataFrame and a Series has always been to align on the DataFrame’s columns and broadcast down the rows, except in the special case that the DataFrame contains time series. Since there are now method for each binary operator enabling you to specify how you want to broadcast, we are phasing out this special case (Zen of Python: Special cases aren’t special enough to break the rules). Here’s what I’m talking about:

In [1]: import pandas as pd

In [2]: df = pd.DataFrame(np.random.randn(6, 4),
   ...:                   index=pd.date_range('1/1/2000', periods=6))
   ...: 

In [3]: df
Out[3]: 
                   0         1         2         3
2000-01-01 -0.134024 -0.205969  1.348944 -1.198246
2000-01-02 -1.626124  0.982041  0.059493 -0.460111
2000-01-03 -1.565401 -0.025706  0.942864  2.502156
2000-01-04 -0.302741  0.261551 -0.066342  0.897097
2000-01-05  0.268766 -1.225092  0.582752 -1.490764
2000-01-06 -0.639757 -0.952750 -0.892402  0.505987

[6 rows x 4 columns]

# deprecated now
In [4]: df - df[0]
Out[4]: 
            2000-01-01 00:00:00  2000-01-02 00:00:00  2000-01-03 00:00:00  2000-01-04 00:00:00 ...   0   1   2   3
2000-01-01                  NaN                  NaN                  NaN                  NaN ... NaN NaN NaN NaN
2000-01-02                  NaN                  NaN                  NaN                  NaN ... NaN NaN NaN NaN
2000-01-03                  NaN                  NaN                  NaN                  NaN ... NaN NaN NaN NaN
2000-01-04                  NaN                  NaN                  NaN                  NaN ... NaN NaN NaN NaN
2000-01-05                  NaN                  NaN                  NaN                  NaN ... NaN NaN NaN NaN
2000-01-06                  NaN                  NaN                  NaN                  NaN ... NaN NaN NaN NaN

[6 rows x 10 columns]

# Change your code to
In [5]: df.sub(df[0], axis=0) # align on axis 0 (rows)
Out[5]: 
              0         1         2         3
2000-01-01  0.0 -0.071946  1.482967 -1.064223
2000-01-02  0.0  2.608165  1.685618  1.166013
2000-01-03  0.0  1.539695  2.508265  4.067556
2000-01-04  0.0  0.564293  0.236399  1.199839
2000-01-05  0.0 -1.493857  0.313986 -1.759530
2000-01-06  0.0 -0.312993 -0.252645  1.145744

[6 rows x 4 columns]

You will get a deprecation warning in the 0.10.x series, and the deprecated functionality will be removed in 0.11 or later.

Altered resample default behavior

The default time series resample binning behavior of daily D and higher frequencies has been changed to closed='left', label='left'. Lower nfrequencies are unaffected. The prior defaults were causing a great deal of confusion for users, especially resampling data to daily frequency (which labeled the aggregated group with the end of the interval: the next day).

In [1]: dates = pd.date_range('1/1/2000', '1/5/2000', freq='4h')

In [2]: series = Series(np.arange(len(dates)), index=dates)

In [3]: series
Out[3]:
2000-01-01 00:00:00     0
2000-01-01 04:00:00     1
2000-01-01 08:00:00     2
2000-01-01 12:00:00     3
2000-01-01 16:00:00     4
2000-01-01 20:00:00     5
2000-01-02 00:00:00     6
2000-01-02 04:00:00     7
2000-01-02 08:00:00     8
2000-01-02 12:00:00     9
2000-01-02 16:00:00    10
2000-01-02 20:00:00    11
2000-01-03 00:00:00    12
2000-01-03 04:00:00    13
2000-01-03 08:00:00    14
2000-01-03 12:00:00    15
2000-01-03 16:00:00    16
2000-01-03 20:00:00    17
2000-01-04 00:00:00    18
2000-01-04 04:00:00    19
2000-01-04 08:00:00    20
2000-01-04 12:00:00    21
2000-01-04 16:00:00    22
2000-01-04 20:00:00    23
2000-01-05 00:00:00    24
Freq: 4H, dtype: int64

In [4]: series.resample('D', how='sum')
Out[4]:
2000-01-01     15
2000-01-02     51
2000-01-03     87
2000-01-04    123
2000-01-05     24
Freq: D, dtype: int64

In [5]: # old behavior
In [6]: series.resample('D', how='sum', closed='right', label='right')
Out[6]:
2000-01-01      0
2000-01-02     21
2000-01-03     57
2000-01-04     93
2000-01-05    129
Freq: D, dtype: int64
  • Infinity and negative infinity are no longer treated as NA by isnull and notnull. That they ever were was a relic of early pandas. This behavior can be re-enabled globally by the mode.use_inf_as_null option:
In [6]: s = pd.Series([1.5, np.inf, 3.4, -np.inf])

In [7]: pd.isnull(s)
Out[7]:
0    False
1    False
2    False
3    False
Length: 4, dtype: bool

In [8]: s.fillna(0)
Out[8]:
0    1.500000
1         inf
2    3.400000
3        -inf
Length: 4, dtype: float64

In [9]: pd.set_option('use_inf_as_null', True)

In [10]: pd.isnull(s)
Out[10]:
0    False
1     True
2    False
3     True
Length: 4, dtype: bool

In [11]: s.fillna(0)
Out[11]:
0    1.5
1    0.0
2    3.4
3    0.0
Length: 4, dtype: float64

In [12]: pd.reset_option('use_inf_as_null')
  • Methods with the inplace option now all return None instead of the calling object. E.g. code written like df = df.fillna(0, inplace=True) may stop working. To fix, simply delete the unnecessary variable assignment.
  • pandas.merge no longer sorts the group keys (sort=False) by default. This was done for performance reasons: the group-key sorting is often one of the more expensive parts of the computation and is often unnecessary.
  • The default column names for a file with no header have been changed to the integers 0 through N - 1. This is to create consistency with the DataFrame constructor with no columns specified. The v0.9.0 behavior (names X0, X1, …) can be reproduced by specifying prefix='X':
In [6]: data= 'a,b,c\n1,Yes,2\n3,No,4'

In [7]: print(data)
a,b,c
1,Yes,2
3,No,4

In [8]: pd.read_csv(StringIO(data), header=None)
Out[8]: 
   0    1  2
0  a    b  c
1  1  Yes  2
2  3   No  4

[3 rows x 3 columns]

In [9]: pd.read_csv(StringIO(data), header=None, prefix='X')
Out[9]: 
  X0   X1 X2
0  a    b  c
1  1  Yes  2
2  3   No  4

[3 rows x 3 columns]
  • Values like 'Yes' and 'No' are not interpreted as boolean by default, though this can be controlled by new true_values and false_values arguments:
In [10]: print(data)
a,b,c
1,Yes,2
3,No,4

In [11]: pd.read_csv(StringIO(data))
Out[11]: 
   a    b  c
0  1  Yes  2
1  3   No  4

[2 rows x 3 columns]

In [12]: pd.read_csv(StringIO(data), true_values=['Yes'], false_values=['No'])
Out[12]: 
   a      b  c
0  1   True  2
1  3  False  4

[2 rows x 3 columns]
  • The file parsers will not recognize non-string values arising from a converter function as NA if passed in the na_values argument. It’s better to do post-processing using the replace function instead.
  • Calling fillna on Series or DataFrame with no arguments is no longer valid code. You must either specify a fill value or an interpolation method:
In [13]: s = Series([np.nan, 1., 2., np.nan, 4])

In [14]: s
Out[14]: 
0    NaN
1    1.0
2    2.0
3    NaN
4    4.0
Length: 5, dtype: float64

In [15]: s.fillna(0)
Out[15]: 
0    0.0
1    1.0
2    2.0
3    0.0
4    4.0
Length: 5, dtype: float64

In [16]: s.fillna(method='pad')
Out[16]: 
0    NaN
1    1.0
2    2.0
3    2.0
4    4.0
Length: 5, dtype: float64

Convenience methods ffill and bfill have been added:

In [17]: s.ffill()
Out[17]: 
0    NaN
1    1.0
2    2.0
3    2.0
4    4.0
Length: 5, dtype: float64
  • Series.apply will now operate on a returned value from the applied function, that is itself a series, and possibly upcast the result to a DataFrame

    In [18]: def f(x):
       ....:     return Series([ x, x**2 ], index = ['x', 'x^2'])
       ....: 
    
    In [19]: s = Series(np.random.rand(5))
    
    In [20]: s
    Out[20]: 
    0    0.717478
    1    0.815199
    2    0.452478
    3    0.848385
    4    0.235477
    Length: 5, dtype: float64
    
    In [21]: s.apply(f)
    Out[21]: 
              x       x^2
    0  0.717478  0.514775
    1  0.815199  0.664550
    2  0.452478  0.204737
    3  0.848385  0.719757
    4  0.235477  0.055449
    
    [5 rows x 2 columns]
    
  • New API functions for working with pandas options (GH2097):

    • get_option / set_option - get/set the value of an option. Partial names are accepted. - reset_option - reset one or more options to their default value. Partial names are accepted. - describe_option - print a description of one or more options. When called with no arguments. print all registered options.

    Note: set_printoptions/ reset_printoptions are now deprecated (but functioning), the print options now live under “display.XYZ”. For example:

    In [22]: get_option("display.max_rows")
    Out[22]: 15
    
  • to_string() methods now always return unicode strings (GH2224).

New features

Wide DataFrame Printing

Instead of printing the summary information, pandas now splits the string representation across multiple rows by default:

In [23]: wide_frame = DataFrame(randn(5, 16))

In [24]: wide_frame
Out[24]: 
         0         1         2         3         4     ...           11        12        13        14        15
0 -0.681624  0.191356  1.180274 -0.834179  0.703043    ...    -0.941235  0.863067 -0.336232 -0.976847  0.033862
1  0.441522 -0.316864 -0.017062  1.570114 -0.360875    ...    -0.906840  1.014601 -0.475108 -0.358944  1.262942
2 -0.412451 -0.462580  0.422194  0.288403 -0.487393    ...     0.246354 -0.727728 -0.094414 -0.276854  0.158399
3 -0.277255  1.331263  0.585174 -0.568825 -0.719412    ...    -1.425911 -0.548829  0.774225  0.740501  1.510263
4 -1.642511  0.432560  1.218080 -0.564705 -0.581790    ...    -0.471786  0.314510 -0.059986 -2.069319 -1.115104

[5 rows x 16 columns]

The old behavior of printing out summary information can be achieved via the ‘expand_frame_repr’ print option:

In [25]: pd.set_option('expand_frame_repr', False)

In [26]: wide_frame
Out[26]: 
         0         1         2         3         4         5         6         7         8         9         10        11        12        13        14        15
0 -0.681624  0.191356  1.180274 -0.834179  0.703043  0.166568 -0.583599 -1.201796 -1.422811 -0.882554  1.209871 -0.941235  0.863067 -0.336232 -0.976847  0.033862
1  0.441522 -0.316864 -0.017062  1.570114 -0.360875 -0.880096  0.235532  0.207232 -1.983857 -1.702547 -1.621234 -0.906840  1.014601 -0.475108 -0.358944  1.262942
2 -0.412451 -0.462580  0.422194  0.288403 -0.487393 -0.777639  0.055865  1.383381  0.085638  0.246392  0.965887  0.246354 -0.727728 -0.094414 -0.276854  0.158399
3 -0.277255  1.331263  0.585174 -0.568825 -0.719412  1.191340 -0.456362  0.089931  0.776079  0.752889 -1.195795 -1.425911 -0.548829  0.774225  0.740501  1.510263
4 -1.642511  0.432560  1.218080 -0.564705 -0.581790  0.286071  0.048725  1.002440  1.276582  0.054399  0.241963 -0.471786  0.314510 -0.059986 -2.069319 -1.115104

[5 rows x 16 columns]

The width of each line can be changed via ‘line_width’ (80 by default):

In [27]: pd.set_option('line_width', 40)
---------------------------------------------------------------------------
OptionError                               Traceback (most recent call last)
<ipython-input-27-b8740c4a0a1b> in <module>()
----> 1 pd.set_option('line_width', 40)

~/sandbox/pandas-release/pandas-docs/pandas/core/config.py in __call__(self, *args, **kwds)
    225 
    226     def __call__(self, *args, **kwds):
--> 227         return self.__func__(*args, **kwds)
    228 
    229     @property

~/sandbox/pandas-release/pandas-docs/pandas/core/config.py in _set_option(*args, **kwargs)
    117 
    118     for k, v in zip(args[::2], args[1::2]):
--> 119         key = _get_single_key(k, silent)
    120 
    121         o = _get_registered_option(key)

~/sandbox/pandas-release/pandas-docs/pandas/core/config.py in _get_single_key(pat, silent)
     81         if not silent:
     82             _warn_if_deprecated(pat)
---> 83         raise OptionError('No such keys(s): {pat!r}'.format(pat=pat))
     84     if len(keys) > 1:
     85         raise OptionError('Pattern matched multiple keys')

OptionError: "No such keys(s): 'line_width'"

In [28]: wide_frame
Out[28]: 
         0         1         2         3         4     ...           11        12        13        14        15
0 -0.681624  0.191356  1.180274 -0.834179  0.703043    ...    -0.941235  0.863067 -0.336232 -0.976847  0.033862
1  0.441522 -0.316864 -0.017062  1.570114 -0.360875    ...    -0.906840  1.014601 -0.475108 -0.358944  1.262942
2 -0.412451 -0.462580  0.422194  0.288403 -0.487393    ...     0.246354 -0.727728 -0.094414 -0.276854  0.158399
3 -0.277255  1.331263  0.585174 -0.568825 -0.719412    ...    -1.425911 -0.548829  0.774225  0.740501  1.510263
4 -1.642511  0.432560  1.218080 -0.564705 -0.581790    ...    -0.471786  0.314510 -0.059986 -2.069319 -1.115104

[5 rows x 16 columns]

Updated PyTables Support

Docs for PyTables Table format & several enhancements to the api. Here is a taste of what to expect.

In [29]: store = HDFStore('store.h5')

In [30]: df = DataFrame(randn(8, 3), index=date_range('1/1/2000', periods=8),
   ....:            columns=['A', 'B', 'C'])
   ....: 

In [31]: df
Out[31]: 
                   A         B         C
2000-01-01 -0.369325 -1.502617 -0.376280
2000-01-02  0.511936 -0.116412 -0.625256
2000-01-03 -0.550627  1.261433 -0.552429
2000-01-04  1.695803 -1.025917 -0.910942
2000-01-05  0.426805 -0.131749  0.432600
2000-01-06  0.044671 -0.341265  1.844536
2000-01-07 -2.036047  0.000830 -0.955697
2000-01-08 -0.898872 -0.725411  0.059904

[8 rows x 3 columns]

# appending data frames
In [32]: df1 = df[0:4]

In [33]: df2 = df[4:]

In [34]: store.append('df', df1)

In [35]: store.append('df', df2)

In [36]: store
Out[36]: 
<class 'pandas.io.pytables.HDFStore'>
File path: store.h5

# selecting the entire store
In [37]: store.select('df')
Out[37]: 
                   A         B         C
2000-01-01 -0.369325 -1.502617 -0.376280
2000-01-02  0.511936 -0.116412 -0.625256
2000-01-03 -0.550627  1.261433 -0.552429
2000-01-04  1.695803 -1.025917 -0.910942
2000-01-05  0.426805 -0.131749  0.432600
2000-01-06  0.044671 -0.341265  1.844536
2000-01-07 -2.036047  0.000830 -0.955697
2000-01-08 -0.898872 -0.725411  0.059904

[8 rows x 3 columns]
In [38]: wp = Panel(randn(2, 5, 4), items=['Item1', 'Item2'],
   ....:        major_axis=date_range('1/1/2000', periods=5),
   ....:        minor_axis=['A', 'B', 'C', 'D'])
   ....: 

In [39]: wp
Out[39]: 
<class 'pandas.core.panel.Panel'>
Dimensions: 2 (items) x 5 (major_axis) x 4 (minor_axis)
Items axis: Item1 to Item2
Major_axis axis: 2000-01-01 00:00:00 to 2000-01-05 00:00:00
Minor_axis axis: A to D

# storing a panel
In [40]: store.append('wp',wp)

# selecting via A QUERY
In [41]: store.select('wp', "major_axis>20000102 and minor_axis=['A','B']")
Out[41]: 
<class 'pandas.core.panel.Panel'>
Dimensions: 2 (items) x 3 (major_axis) x 2 (minor_axis)
Items axis: Item1 to Item2
Major_axis axis: 2000-01-03 00:00:00 to 2000-01-05 00:00:00
Minor_axis axis: A to B

# removing data from tables
In [42]: store.remove('wp', "major_axis>20000103")
Out[42]: 8

In [43]: store.select('wp')
Out[43]: 
<class 'pandas.core.panel.Panel'>
Dimensions: 2 (items) x 3 (major_axis) x 4 (minor_axis)
Items axis: Item1 to Item2
Major_axis axis: 2000-01-01 00:00:00 to 2000-01-03 00:00:00
Minor_axis axis: A to D

# deleting a store
In [44]: del store['df']

In [45]: store
Out[45]: 
<class 'pandas.io.pytables.HDFStore'>
File path: store.h5

Enhancements

  • added ability to hierarchical keys

    In [46]: store.put('foo/bar/bah', df)
    
    In [47]: store.append('food/orange', df)
    
    In [48]: store.append('food/apple',  df)
    
    In [49]: store
    Out[49]: 
    <class 'pandas.io.pytables.HDFStore'>
    File path: store.h5
    
    # remove all nodes under this level
    In [50]: store.remove('food')
    
    In [51]: store
    Out[51]: 
    <class 'pandas.io.pytables.HDFStore'>
    File path: store.h5
    
  • added mixed-dtype support!

    In [52]: df['string'] = 'string'
    
    In [53]: df['int']    = 1
    
    In [54]: store.append('df',df)
    
    In [55]: df1 = store.select('df')
    
    In [56]: df1
    Out[56]: 
                       A         B         C  string  int
    2000-01-01 -0.369325 -1.502617 -0.376280  string    1
    2000-01-02  0.511936 -0.116412 -0.625256  string    1
    2000-01-03 -0.550627  1.261433 -0.552429  string    1
    2000-01-04  1.695803 -1.025917 -0.910942  string    1
    2000-01-05  0.426805 -0.131749  0.432600  string    1
    2000-01-06  0.044671 -0.341265  1.844536  string    1
    2000-01-07 -2.036047  0.000830 -0.955697  string    1
    2000-01-08 -0.898872 -0.725411  0.059904  string    1
    
    [8 rows x 5 columns]
    
    In [57]: df1.get_dtype_counts()
    Out[57]: 
    float64    3
    object     1
    int64      1
    Length: 3, dtype: int64
    
  • performance improvements on table writing

  • support for arbitrarily indexed dimensions

  • SparseSeries now has a density property (GH2384)

  • enable Series.str.strip/lstrip/rstrip methods to take an input argument to strip arbitrary characters (GH2411)

  • implement value_vars in melt to limit values to certain columns and add melt to pandas namespace (GH2412)

Bug Fixes

  • added Term method of specifying where conditions (GH1996).
  • del store['df'] now call store.remove('df') for store deletion
  • deleting of consecutive rows is much faster than before
  • min_itemsize parameter can be specified in table creation to force a minimum size for indexing columns (the previous implementation would set the column size based on the first append)
  • indexing support via create_table_index (requires PyTables >= 2.3) (GH698).
  • appending on a store would fail if the table was not first created via put
  • fixed issue with missing attributes after loading a pickled dataframe (GH2431)
  • minor change to select and remove: require a table ONLY if where is also provided (and not None)

Compatibility

0.10 of HDFStore is backwards compatible for reading tables created in a prior version of pandas, however, query terms using the prior (undocumented) methodology are unsupported. You must read in the entire file and write it out using the new format to take advantage of the updates.

N Dimensional Panels (Experimental)

Adding experimental support for Panel4D and factory functions to create n-dimensional named panels. Here is a taste of what to expect.

In [58]: p4d = Panel4D(randn(2, 2, 5, 4),
  ....:       labels=['Label1','Label2'],
  ....:       items=['Item1', 'Item2'],
  ....:       major_axis=date_range('1/1/2000', periods=5),
  ....:       minor_axis=['A', 'B', 'C', 'D'])
  ....:

In [59]: p4d
Out[59]:
<class 'pandas.core.panelnd.Panel4D'>
Dimensions: 2 (labels) x 2 (items) x 5 (major_axis) x 4 (minor_axis)
Labels axis: Label1 to Label2
Items axis: Item1 to Item2
Major_axis axis: 2000-01-01 00:00:00 to 2000-01-05 00:00:00
Minor_axis axis: A to D

See the full release notes or issue tracker on GitHub for a complete list.

v0.9.1 (November 14, 2012)

This is a bugfix release from 0.9.0 and includes several new features and enhancements along with a large number of bug fixes. The new features include by-column sort order for DataFrame and Series, improved NA handling for the rank method, masking functions for DataFrame, and intraday time-series filtering for DataFrame.

New features

  • Series.sort, DataFrame.sort, and DataFrame.sort_index can now be specified in a per-column manner to support multiple sort orders (GH928)

    In [2]:  df = DataFrame(np.random.randint(0, 2, (6, 3)), columns=['A', 'B', 'C'])
    
    In [3]: df.sort(['A', 'B'], ascending=[1, 0])
    
    Out[3]:
       A  B  C
    3  0  1  1
    4  0  1  1
    2  0  0  1
    0  1  0  0
    1  1  0  0
    5  1  0  0
    
  • DataFrame.rank now supports additional argument values for the na_option parameter so missing values can be assigned either the largest or the smallest rank (GH1508, GH2159)

    In [1]: df = DataFrame(np.random.randn(6, 3), columns=['A', 'B', 'C'])
    
    In [2]: df.loc[2:4] = np.nan
    
    In [3]: df.rank()
    Out[3]: 
         A    B    C
    0  3.0  1.0  3.0
    1  1.0  3.0  2.0
    2  NaN  NaN  NaN
    3  NaN  NaN  NaN
    4  NaN  NaN  NaN
    5  2.0  2.0  1.0
    
    [6 rows x 3 columns]
    
    In [4]: df.rank(na_option='top')
    Out[4]: 
         A    B    C
    0  6.0  4.0  6.0
    1  4.0  6.0  5.0
    2  2.0  2.0  2.0
    3  2.0  2.0  2.0
    4  2.0  2.0  2.0
    5  5.0  5.0  4.0
    
    [6 rows x 3 columns]
    
    In [5]: df.rank(na_option='bottom')
    Out[5]: 
         A    B    C
    0  3.0  1.0  3.0
    1  1.0  3.0  2.0
    2  5.0  5.0  5.0
    3  5.0  5.0  5.0
    4  5.0  5.0  5.0
    5  2.0  2.0  1.0
    
    [6 rows x 3 columns]
    
  • DataFrame has new where and mask methods to select values according to a given boolean mask (GH2109, GH2151)

    DataFrame currently supports slicing via a boolean vector the same length as the DataFrame (inside the []). The returned DataFrame has the same number of columns as the original, but is sliced on its index.

    In [6]: df = DataFrame(np.random.randn(5, 3), columns = ['A','B','C'])
    
    In [7]: df
    Out[7]: 
              A         B         C
    0 -1.101581 -1.187831  0.630693
    1  2.369983  0.333769 -0.870464
    2  1.118760 -0.224382  0.642489
    3  0.961751 -1.848369  0.440883
    4  1.235390  1.615529 -0.303272
    
    [5 rows x 3 columns]
    
    In [8]: df[df['A'] > 0]
    Out[8]: 
              A         B         C
    1  2.369983  0.333769 -0.870464
    2  1.118760 -0.224382  0.642489
    3  0.961751 -1.848369  0.440883
    4  1.235390  1.615529 -0.303272
    
    [4 rows x 3 columns]
    

    If a DataFrame is sliced with a DataFrame based boolean condition (with the same size as the original DataFrame), then a DataFrame the same size (index and columns) as the original is returned, with elements that do not meet the boolean condition as NaN. This is accomplished via the new method DataFrame.where. In addition, where takes an optional other argument for replacement.

    In [9]: df[df>0]
    Out[9]: 
              A         B         C
    0       NaN       NaN  0.630693
    1  2.369983  0.333769       NaN
    2  1.118760       NaN  0.642489
    3  0.961751       NaN  0.440883
    4  1.235390  1.615529       NaN
    
    [5 rows x 3 columns]
    
    In [10]: df.where(df>0)
    Out[10]: 
              A         B         C
    0       NaN       NaN  0.630693
    1  2.369983  0.333769       NaN
    2  1.118760       NaN  0.642489
    3  0.961751       NaN  0.440883
    4  1.235390  1.615529       NaN
    
    [5 rows x 3 columns]
    
    In [11]: df.where(df>0,-df)
    Out[11]: 
              A         B         C
    0  1.101581  1.187831  0.630693
    1  2.369983  0.333769  0.870464
    2  1.118760  0.224382  0.642489
    3  0.961751  1.848369  0.440883
    4  1.235390  1.615529  0.303272
    
    [5 rows x 3 columns]
    

    Furthermore, where now aligns the input boolean condition (ndarray or DataFrame), such that partial selection with setting is possible. This is analogous to partial setting via .ix (but on the contents rather than the axis labels)

    In [12]: df2 = df.copy()
    
    In [13]: df2[ df2[1:4] > 0 ] = 3
    
    In [14]: df2
    Out[14]: 
              A         B         C
    0 -1.101581 -1.187831  0.630693
    1  3.000000  3.000000 -0.870464
    2  3.000000 -0.224382  3.000000
    3  3.000000 -1.848369  3.000000
    4  1.235390  1.615529 -0.303272
    
    [5 rows x 3 columns]
    

    DataFrame.mask is the inverse boolean operation of where.

    In [15]: df.mask(df<=0)
    Out[15]: 
              A         B         C
    0       NaN       NaN  0.630693
    1  2.369983  0.333769       NaN
    2  1.118760       NaN  0.642489
    3  0.961751       NaN  0.440883
    4  1.235390  1.615529       NaN
    
    [5 rows x 3 columns]
    
  • Enable referencing of Excel columns by their column names (GH1936)

    In [16]: xl = ExcelFile('data/test.xls')
    
    In [17]: xl.parse('Sheet1', index_col=0, parse_dates=True,
       ....:          parse_cols='A:D')
       ....: 
    Out[17]: 
                       A         B         C         D
    2000-01-03  0.980269  3.685731 -0.364217 -1.159738
    2000-01-04  1.047916 -0.041232 -0.161812  0.212549
    2000-01-05  0.498581  0.731168 -0.537677  1.346270
    2000-01-06  1.120202  1.567621  0.003641  0.675253
    2000-01-07 -0.487094  0.571455 -1.611639  0.103469
    2000-01-10  0.836649  0.246462  0.588543  1.062782
    2000-01-11 -0.157161  1.340307  1.195778 -1.097007
    
    [7 rows x 4 columns]
    
  • Added option to disable pandas-style tick locators and formatters using series.plot(x_compat=True) or pandas.plot_params[‘x_compat’] = True (GH2205)

  • Existing TimeSeries methods at_time and between_time were added to DataFrame (GH2149)

  • DataFrame.dot can now accept ndarrays (GH2042)

  • DataFrame.drop now supports non-unique indexes (GH2101)

  • Panel.shift now supports negative periods (GH2164)

  • DataFrame now support unary ~ operator (GH2110)

API changes

  • Upsampling data with a PeriodIndex will result in a higher frequency TimeSeries that spans the original time window

    In [1]: prng = period_range('2012Q1', periods=2, freq='Q')
    
    In [2]: s = Series(np.random.randn(len(prng)), prng)
    
    In [4]: s.resample('M')
    Out[4]:
    2012-01   -1.471992
    2012-02         NaN
    2012-03         NaN
    2012-04   -0.493593
    2012-05         NaN
    2012-06         NaN
    Freq: M, dtype: float64
    
  • Period.end_time now returns the last nanosecond in the time interval (GH2124, GH2125, GH1764)

    In [18]: p = Period('2012')
    
    In [19]: p.end_time
    Out[19]: Timestamp('2012-12-31 23:59:59.999999999')
    
  • File parsers no longer coerce to float or bool for columns that have custom converters specified (GH2184)

    In [20]: data = 'A,B,C\n00001,001,5\n00002,002,6'
    
    In [21]: read_csv(StringIO(data), converters={'A' : lambda x: x.strip()})
    Out[21]: 
           A  B  C
    0  00001  1  5
    1  00002  2  6
    
    [2 rows x 3 columns]
    

See the full release notes or issue tracker on GitHub for a complete list.

v0.9.0 (October 7, 2012)

This is a major release from 0.8.1 and includes several new features and enhancements along with a large number of bug fixes. New features include vectorized unicode encoding/decoding for Series.str, to_latex method to DataFrame, more flexible parsing of boolean values, and enabling the download of options data from Yahoo! Finance.

New features

  • Add encode and decode for unicode handling to vectorized string processing methods in Series.str (GH1706)
  • Add DataFrame.to_latex method (GH1735)
  • Add convenient expanding window equivalents of all rolling_* ops (GH1785)
  • Add Options class to pandas.io.data for fetching options data from Yahoo! Finance (GH1748, GH1739)
  • More flexible parsing of boolean values (Yes, No, TRUE, FALSE, etc) (GH1691, GH1295)
  • Add level parameter to Series.reset_index
  • TimeSeries.between_time can now select times across midnight (GH1871)
  • Series constructor can now handle generator as input (GH1679)
  • DataFrame.dropna can now take multiple axes (tuple/list) as input (GH924)
  • Enable skip_footer parameter in ExcelFile.parse (GH1843)

API changes

  • The default column names when header=None and no columns names passed to functions like read_csv has changed to be more Pythonic and amenable to attribute access:
In [1]: data = '0,0,1\n1,1,0\n0,1,0'

In [2]: df = read_csv(StringIO(data), header=None)

In [3]: df
Out[3]: 
   0  1  2
0  0  0  1
1  1  1  0
2  0  1  0

[3 rows x 3 columns]
  • Creating a Series from another Series, passing an index, will cause reindexing to happen inside rather than treating the Series like an ndarray. Technically improper usages like Series(df[col1], index=df[col2]) that worked before “by accident” (this was never intended) will lead to all NA Series in some cases. To be perfectly clear:
In [4]: s1 = Series([1, 2, 3])

In [5]: s1
Out[5]: 
0    1
1    2
2    3
Length: 3, dtype: int64

In [6]: s2 = Series(s1, index=['foo', 'bar', 'baz'])

In [7]: s2
Out[7]: 
foo   NaN
bar   NaN
baz   NaN
Length: 3, dtype: float64
  • Deprecated day_of_year API removed from PeriodIndex, use dayofyear (GH1723)
  • Don’t modify NumPy suppress printoption to True at import time
  • The internal HDF5 data arrangement for DataFrames has been transposed. Legacy files will still be readable by HDFStore (GH1834, GH1824)
  • Legacy cruft removed: pandas.stats.misc.quantileTS
  • Use ISO8601 format for Period repr: monthly, daily, and on down (GH1776)
  • Empty DataFrame columns are now created as object dtype. This will prevent a class of TypeErrors that was occurring in code where the dtype of a column would depend on the presence of data or not (e.g. a SQL query having results) (GH1783)
  • Setting parts of DataFrame/Panel using ix now aligns input Series/DataFrame (GH1630)
  • first and last methods in GroupBy no longer drop non-numeric columns (GH1809)
  • Resolved inconsistencies in specifying custom NA values in text parser. na_values of type dict no longer override default NAs unless keep_default_na is set to false explicitly (GH1657)
  • DataFrame.dot will not do data alignment, and also work with Series (GH1915)

See the full release notes or issue tracker on GitHub for a complete list.

v0.8.1 (July 22, 2012)

This release includes a few new features, performance enhancements, and over 30 bug fixes from 0.8.0. New features include notably NA friendly string processing functionality and a series of new plot types and options.

New features

Performance improvements

  • Improved implementation of rolling min and max (thanks to Bottleneck !)
  • Add accelerated 'median' GroupBy option (GH1358)
  • Significantly improve the performance of parsing ISO8601-format date strings with DatetimeIndex or to_datetime (GH1571)
  • Improve the performance of GroupBy on single-key aggregations and use with Categorical types
  • Significant datetime parsing performance improvements

v0.8.0 (June 29, 2012)

This is a major release from 0.7.3 and includes extensive work on the time series handling and processing infrastructure as well as a great deal of new functionality throughout the library. It includes over 700 commits from more than 20 distinct authors. Most pandas 0.7.3 and earlier users should not experience any issues upgrading, but due to the migration to the NumPy datetime64 dtype, there may be a number of bugs and incompatibilities lurking. Lingering incompatibilities will be fixed ASAP in a 0.8.1 release if necessary. See the full release notes or issue tracker on GitHub for a complete list.

Support for non-unique indexes

All objects can now work with non-unique indexes. Data alignment / join operations work according to SQL join semantics (including, if application, index duplication in many-to-many joins)

NumPy datetime64 dtype and 1.6 dependency

Time series data are now represented using NumPy’s datetime64 dtype; thus, pandas 0.8.0 now requires at least NumPy 1.6. It has been tested and verified to work with the development version (1.7+) of NumPy as well which includes some significant user-facing API changes. NumPy 1.6 also has a number of bugs having to do with nanosecond resolution data, so I recommend that you steer clear of NumPy 1.6’s datetime64 API functions (though limited as they are) and only interact with this data using the interface that pandas provides.

See the end of the 0.8.0 section for a “porting” guide listing potential issues for users migrating legacy codebases from pandas 0.7 or earlier to 0.8.0.

Bug fixes to the 0.7.x series for legacy NumPy < 1.6 users will be provided as they arise. There will be no more further development in 0.7.x beyond bug fixes.

Time series changes and improvements

Note

With this release, legacy scikits.timeseries users should be able to port their code to use pandas.

Note

See documentation for overview of pandas timeseries API.

  • New datetime64 representation speeds up join operations and data alignment, reduces memory usage, and improve serialization / deserialization performance significantly over datetime.datetime
  • High performance and flexible resample method for converting from high-to-low and low-to-high frequency. Supports interpolation, user-defined aggregation functions, and control over how the intervals and result labeling are defined. A suite of high performance Cython/C-based resampling functions (including Open-High-Low-Close) have also been implemented.
  • Revamp of frequency aliases and support for frequency shortcuts like ‘15min’, or ‘1h30min’
  • New DatetimeIndex class supports both fixed frequency and irregular time series. Replaces now deprecated DateRange class
  • New PeriodIndex and Period classes for representing time spans and performing calendar logic, including the 12 fiscal quarterly frequencies <timeseries.quarterly>. This is a partial port of, and a substantial enhancement to, elements of the scikits.timeseries codebase. Support for conversion between PeriodIndex and DatetimeIndex
  • New Timestamp data type subclasses datetime.datetime, providing the same interface while enabling working with nanosecond-resolution data. Also provides easy time zone conversions.
  • Enhanced support for time zones. Add tz_convert and tz_lcoalize methods to TimeSeries and DataFrame. All timestamps are stored as UTC; Timestamps from DatetimeIndex objects with time zone set will be localized to localtime. Time zone conversions are therefore essentially free. User needs to know very little about pytz library now; only time zone names as as strings are required. Time zone-aware timestamps are equal if and only if their UTC timestamps match. Operations between time zone-aware time series with different time zones will result in a UTC-indexed time series.
  • Time series string indexing conveniences / shortcuts: slice years, year and month, and index values with strings
  • Enhanced time series plotting; adaptation of scikits.timeseries matplotlib-based plotting code
  • New date_range, bdate_range, and period_range factory functions
  • Robust frequency inference function infer_freq and inferred_freq property of DatetimeIndex, with option to infer frequency on construction of DatetimeIndex
  • to_datetime function efficiently parses array of strings to DatetimeIndex. DatetimeIndex will parse array or list of strings to datetime64
  • Optimized support for datetime64-dtype data in Series and DataFrame columns
  • New NaT (Not-a-Time) type to represent NA in timestamp arrays
  • Optimize Series.asof for looking up “as of” values for arrays of timestamps
  • Milli, Micro, Nano date offset objects
  • Can index time series with datetime.time objects to select all data at particular time of day (TimeSeries.at_time) or between two times (TimeSeries.between_time)
  • Add tshift method for leading/lagging using the frequency (if any) of the index, as opposed to a naive lead/lag using shift

Other new features

  • New cut and qcut functions (like R’s cut function) for computing a categorical variable from a continuous variable by binning values either into value-based (cut) or quantile-based (qcut) bins
  • Rename Factor to Categorical and add a number of usability features
  • Add limit argument to fillna/reindex
  • More flexible multiple function application in GroupBy, and can pass list (name, function) tuples to get result in particular order with given names
  • Add flexible replace method for efficiently substituting values
  • Enhanced read_csv/read_table for reading time series data and converting multiple columns to dates
  • Add comments option to parser functions: read_csv, etc.
  • Add dayfirst option to parser functions for parsing international DD/MM/YYYY dates
  • Allow the user to specify the CSV reader dialect to control quoting etc.
  • Handling thousands separators in read_csv to improve integer parsing.
  • Enable unstacking of multiple levels in one shot. Alleviate pivot_table bugs (empty columns being introduced)
  • Move to klib-based hash tables for indexing; better performance and less memory usage than Python’s dict
  • Add first, last, min, max, and prod optimized GroupBy functions
  • New ordered_merge function
  • Add flexible comparison instance methods eq, ne, lt, gt, etc. to DataFrame, Series
  • Improve scatter_matrix plotting function and add histogram or kernel density estimates to diagonal
  • Add ‘kde’ plot option for density plots
  • Support for converting DataFrame to R data.frame through rpy2
  • Improved support for complex numbers in Series and DataFrame
  • Add pct_change method to all data structures
  • Add max_colwidth configuration option for DataFrame console output
  • Interpolate Series values using index values
  • Can select multiple columns from GroupBy
  • Add update methods to Series/DataFrame for updating values in place
  • Add any and all method to DataFrame

New plotting methods

Series.plot now supports a secondary_y option:

In [1]: plt.figure()
Out[1]: <Figure size 640x480 with 0 Axes>

In [2]: fx['FR'].plot(style='g')
Out[2]: <matplotlib.axes._subplots.AxesSubplot at 0x1c376acfd0>

In [3]: fx['IT'].plot(style='k--', secondary_y=True)
Out[3]: <matplotlib.axes._subplots.AxesSubplot at 0x1c3781cb38>

Vytautas Jancauskas, the 2012 GSOC participant, has added many new plot types. For example, 'kde' is a new option:

In [4]: s = Series(np.concatenate((np.random.randn(1000),
   ...:                            np.random.randn(1000) * 0.5 + 3)))
   ...: 

In [5]: plt.figure()
Out[5]: <Figure size 640x480 with 0 Axes>

In [6]: s.hist(normed=True, alpha=0.2)
Out[6]: <matplotlib.axes._subplots.AxesSubplot at 0x1c37c3a588>

In [7]: s.plot(kind='kde')
Out[7]: <matplotlib.axes._subplots.AxesSubplot at 0x1c37c3a588>

See the plotting page for much more.

Other API changes

  • Deprecation of offset, time_rule, and timeRule arguments names in time series functions. Warnings will be printed until pandas 0.9 or 1.0.

Potential porting issues for pandas <= 0.7.3 users

The major change that may affect you in pandas 0.8.0 is that time series indexes use NumPy’s datetime64 data type instead of dtype=object arrays of Python’s built-in datetime.datetime objects. DateRange has been replaced by DatetimeIndex but otherwise behaved identically. But, if you have code that converts DateRange or Index objects that used to contain datetime.datetime values to plain NumPy arrays, you may have bugs lurking with code using scalar values because you are handing control over to NumPy:

In [8]: import datetime

In [9]: rng = date_range('1/1/2000', periods=10)

In [10]: rng[5]
Out[10]: Timestamp('2000-01-06 00:00:00', freq='D')

In [11]: isinstance(rng[5], datetime.datetime)
Out[11]: True

In [12]: rng_asarray = np.asarray(rng)

In [13]: scalar_val = rng_asarray[5]

In [14]: type(scalar_val)
Out[14]: numpy.datetime64

pandas’s Timestamp object is a subclass of datetime.datetime that has nanosecond support (the nanosecond field store the nanosecond value between 0 and 999). It should substitute directly into any code that used datetime.datetime values before. Thus, I recommend not casting DatetimeIndex to regular NumPy arrays.

If you have code that requires an array of datetime.datetime objects, you have a couple of options. First, the astype(object) method of DatetimeIndex produces an array of Timestamp objects:

In [15]: stamp_array = rng.astype(object)

In [16]: stamp_array
Out[16]: 
Index([2000-01-01 00:00:00, 2000-01-02 00:00:00, 2000-01-03 00:00:00,
       2000-01-04 00:00:00, 2000-01-05 00:00:00, 2000-01-06 00:00:00,
       2000-01-07 00:00:00, 2000-01-08 00:00:00, 2000-01-09 00:00:00,
       2000-01-10 00:00:00],
      dtype='object')

In [17]: stamp_array[5]
Out[17]: Timestamp('2000-01-06 00:00:00', freq='D')

To get an array of proper datetime.datetime objects, use the to_pydatetime method:

In [18]: dt_array = rng.to_pydatetime()

In [19]: dt_array
Out[19]: 
array([datetime.datetime(2000, 1, 1, 0, 0),
       datetime.datetime(2000, 1, 2, 0, 0),
       datetime.datetime(2000, 1, 3, 0, 0),
       datetime.datetime(2000, 1, 4, 0, 0),
       datetime.datetime(2000, 1, 5, 0, 0),
       datetime.datetime(2000, 1, 6, 0, 0),
       datetime.datetime(2000, 1, 7, 0, 0),
       datetime.datetime(2000, 1, 8, 0, 0),
       datetime.datetime(2000, 1, 9, 0, 0),
       datetime.datetime(2000, 1, 10, 0, 0)], dtype=object)

In [20]: dt_array[5]
Out[20]: datetime.datetime(2000, 1, 6, 0, 0)

matplotlib knows how to handle datetime.datetime but not Timestamp objects. While I recommend that you plot time series using TimeSeries.plot, you can either use to_pydatetime or register a converter for the Timestamp type. See matplotlib documentation for more on this.

Warning

There are bugs in the user-facing API with the nanosecond datetime64 unit in NumPy 1.6. In particular, the string version of the array shows garbage values, and conversion to dtype=object is similarly broken.

In [21]: rng = date_range('1/1/2000', periods=10)

In [22]: rng
Out[22]: 
DatetimeIndex(['2000-01-01', '2000-01-02', '2000-01-03', '2000-01-04',
               '2000-01-05', '2000-01-06', '2000-01-07', '2000-01-08',
               '2000-01-09', '2000-01-10'],
              dtype='datetime64[ns]', freq='D')

In [23]: np.asarray(rng)
Out[23]: 
array(['2000-01-01T00:00:00.000000000', '2000-01-02T00:00:00.000000000',
       '2000-01-03T00:00:00.000000000', '2000-01-04T00:00:00.000000000',
       '2000-01-05T00:00:00.000000000', '2000-01-06T00:00:00.000000000',
       '2000-01-07T00:00:00.000000000', '2000-01-08T00:00:00.000000000',
       '2000-01-09T00:00:00.000000000', '2000-01-10T00:00:00.000000000'], dtype='datetime64[ns]')

In [24]: converted = np.asarray(rng, dtype=object)

In [25]: converted[5]
Out[25]: 947116800000000000

Trust me: don’t panic. If you are using NumPy 1.6 and restrict your interaction with datetime64 values to pandas’s API you will be just fine. There is nothing wrong with the data-type (a 64-bit integer internally); all of the important data processing happens in pandas and is heavily tested. I strongly recommend that you do not work directly with datetime64 arrays in NumPy 1.6 and only use the pandas API.

Support for non-unique indexes: In the latter case, you may have code inside a try:... catch: block that failed due to the index not being unique. In many cases it will no longer fail (some method like append still check for uniqueness unless disabled). However, all is not lost: you can inspect index.is_unique and raise an exception explicitly if it is False or go to a different code branch.

v.0.7.3 (April 12, 2012)

This is a minor release from 0.7.2 and fixes many minor bugs and adds a number of nice new features. There are also a couple of API changes to note; these should not affect very many users, and we are inclined to call them “bug fixes” even though they do constitute a change in behavior. See the full release notes or issue tracker on GitHub for a complete list.

New features

from pandas.tools.plotting import scatter_matrix
scatter_matrix(df, alpha=0.2)
_images/scatter_matrix_kde.png
  • Add stacked argument to Series and DataFrame’s plot method for stacked bar plots.
df.plot(kind='bar', stacked=True)
_images/bar_plot_stacked_ex.png
df.plot(kind='barh', stacked=True)
_images/barh_plot_stacked_ex.png
  • Add log x and y scaling options to DataFrame.plot and Series.plot
  • Add kurt methods to Series and DataFrame for computing kurtosis

NA Boolean Comparison API Change

Reverted some changes to how NA values (represented typically as NaN or None) are handled in non-numeric Series:

In [1]: series = Series(['Steve', np.nan, 'Joe'])

In [2]: series == 'Steve'
Out[2]: 
0     True
1    False
2    False
Length: 3, dtype: bool

In [3]: series != 'Steve'
Out[3]: 
0    False
1     True
2     True
Length: 3, dtype: bool

In comparisons, NA / NaN will always come through as False except with != which is True. Be very careful with boolean arithmetic, especially negation, in the presence of NA data. You may wish to add an explicit NA filter into boolean array operations if you are worried about this:

In [4]: mask = series == 'Steve'

In [5]: series[mask & series.notnull()]
Out[5]: 
0    Steve
Length: 1, dtype: object

While propagating NA in comparisons may seem like the right behavior to some users (and you could argue on purely technical grounds that this is the right thing to do), the evaluation was made that propagating NA everywhere, including in numerical arrays, would cause a large amount of problems for users. Thus, a “practicality beats purity” approach was taken. This issue may be revisited at some point in the future.

Other API Changes

When calling apply on a grouped Series, the return value will also be a Series, to be more consistent with the groupby behavior with DataFrame:

In [6]: df = DataFrame({'A' : ['foo', 'bar', 'foo', 'bar',
   ...:                     'foo', 'bar', 'foo', 'foo'],
   ...:                 'B' : ['one', 'one', 'two', 'three',
   ...:                        'two', 'two', 'one', 'three'],
   ...:                 'C' : np.random.randn(8), 'D' : np.random.randn(8)})
   ...: 

In [7]: df
Out[7]: 
     A      B         C         D
0  foo    one -0.841015  0.459840
1  bar    one  0.114219 -0.253040
2  foo    two -0.405617 -0.261128
3  bar  three  1.240678  0.406604
4  foo    two -0.122828 -1.022256
5  bar    two  1.525196 -0.882785
6  foo    one  0.520047  1.793331
7  foo  three  0.163834 -0.429688

[8 rows x 4 columns]

In [8]: grouped = df.groupby('A')['C']

In [9]: grouped.describe()
Out[9]: 
     count      mean       std       min       25%       50%       75%       max
A                                                                               
bar    3.0  0.960031  0.746181  0.114219  0.677448  1.240678  1.382937  1.525196
foo    5.0 -0.137116  0.522064 -0.841015 -0.405617 -0.122828  0.163834  0.520047

[2 rows x 8 columns]

In [10]: grouped.apply(lambda x: x.sort_values()[-2:]) # top 2 values
Out[10]: 
A     
bar  3    1.240678
     5    1.525196
foo  7    0.163834
     6    0.520047
Name: C, Length: 4, dtype: float64

v.0.7.2 (March 16, 2012)

This release targets bugs in 0.7.1, and adds a few minor features.

New features

  • Add additional tie-breaking methods in DataFrame.rank (GH874)
  • Add ascending parameter to rank in Series, DataFrame (GH875)
  • Add coerce_float option to DataFrame.from_records (GH893)
  • Add sort_columns parameter to allow unsorted plots (GH918)
  • Enable column access via attributes on GroupBy (GH882)
  • Can pass dict of values to DataFrame.fillna (GH661)
  • Can select multiple hierarchical groups by passing list of values in .ix (GH134)
  • Add axis option to DataFrame.fillna (GH174)
  • Add level keyword to drop for dropping values from a level (GH159)

Performance improvements

  • Use khash for Series.value_counts, add raw function to algorithms.py (GH861)
  • Intercept __builtin__.sum in groupby (GH885)

v.0.7.1 (February 29, 2012)

This release includes a few new features and addresses over a dozen bugs in 0.7.0.

New features

  • Add to_clipboard function to pandas namespace for writing objects to the system clipboard (GH774)
  • Add itertuples method to DataFrame for iterating through the rows of a dataframe as tuples (GH818)
  • Add ability to pass fill_value and method to DataFrame and Series align method (GH806, GH807)
  • Add fill_value option to reindex, align methods (GH784)
  • Enable concat to produce DataFrame from Series (GH787)
  • Add between method to Series (GH802)
  • Add HTML representation hook to DataFrame for the IPython HTML notebook (GH773)
  • Support for reading Excel 2007 XML documents using openpyxl

Performance improvements

  • Improve performance and memory usage of fillna on DataFrame
  • Can concatenate a list of Series along axis=1 to obtain a DataFrame (GH787)

v.0.7.0 (February 9, 2012)

New features

  • New unified merge function for efficiently performing full gamut of database / relational-algebra operations. Refactored existing join methods to use the new infrastructure, resulting in substantial performance gains (GH220, GH249, GH267)
  • New unified concatenation function for concatenating Series, DataFrame or Panel objects along an axis. Can form union or intersection of the other axes. Improves performance of Series.append and DataFrame.append (GH468, GH479, GH273)
  • Can pass multiple DataFrames to DataFrame.append to concatenate (stack) and multiple Series to Series.append too
  • Can pass list of dicts (e.g., a list of JSON objects) to DataFrame constructor (GH526)
  • You can now set multiple columns in a DataFrame via __getitem__, useful for transformation (GH342)
  • Handle differently-indexed output values in DataFrame.apply (GH498)
In [1]: df = DataFrame(randn(10, 4))

In [2]: df.apply(lambda x: x.describe())
Out[2]: 
               0          1          2          3
count  10.000000  10.000000  10.000000  10.000000
mean    0.424980   0.115056   0.452000  -0.103829
std     0.898046   0.712034   0.867316   0.830197
min    -1.337024  -0.779344  -1.206466  -1.022360
25%    -0.000996  -0.344789  -0.120217  -0.744313
50%     0.860419   0.098422   0.540168  -0.387554
75%     1.094236   0.375137   1.076331   0.674952
max     1.207725   1.601703   1.663859   1.096187

[8 rows x 4 columns]
  • Add reorder_levels method to Series and DataFrame (GH534)
  • Add dict-like get function to DataFrame and Panel (GH521)
  • Add DataFrame.iterrows method for efficiently iterating through the rows of a DataFrame
  • Add DataFrame.to_panel with code adapted from LongPanel.to_long
  • Add reindex_axis method added to DataFrame
  • Add level option to binary arithmetic functions on DataFrame and Series
  • Add level option to the reindex and align methods on Series and DataFrame for broadcasting values across a level (GH542, GH552, others)
  • Add attribute-based item access to Panel and add IPython completion (GH563)
  • Add logy option to Series.plot for log-scaling on the Y axis
  • Add index and header options to DataFrame.to_string
  • Can pass multiple DataFrames to DataFrame.join to join on index (GH115)
  • Can pass multiple Panels to Panel.join (GH115)
  • Added justify argument to DataFrame.to_string to allow different alignment of column headers
  • Add sort option to GroupBy to allow disabling sorting of the group keys for potential speedups (GH595)
  • Can pass MaskedArray to Series constructor (GH563)
  • Add Panel item access via attributes and IPython completion (GH554)
  • Implement DataFrame.lookup, fancy-indexing analogue for retrieving values given a sequence of row and column labels (GH338)
  • Can pass a list of functions to aggregate with groupby on a DataFrame, yielding an aggregated result with hierarchical columns (GH166)
  • Can call cummin and cummax on Series and DataFrame to get cumulative minimum and maximum, respectively (GH647)
  • value_range added as utility function to get min and max of a dataframe (GH288)
  • Added encoding argument to read_csv, read_table, to_csv and from_csv for non-ascii text (GH717)
  • Added abs method to pandas objects
  • Added crosstab function for easily computing frequency tables
  • Added isin method to index objects
  • Added level argument to xs method of DataFrame.

API Changes to integer indexing

One of the potentially riskiest API changes in 0.7.0, but also one of the most important, was a complete review of how integer indexes are handled with regard to label-based indexing. Here is an example:

In [3]: s = Series(randn(10), index=range(0, 20, 2))

In [4]: s
Out[4]: 
0     0.446246
2    -0.500268
4     0.814725
6    -0.312744
8     1.098892
10    1.306330
12   -0.366970
14   -0.030890
16    1.608095
18   -0.023287
Length: 10, dtype: float64

In [5]: s[0]
Out[5]: 0.44624598505731339

In [6]: s[2]
Out[6]: -0.500268093241102

In [7]: s[4]
Out[7]: 0.8147247587659604

This is all exactly identical to the behavior before. However, if you ask for a key not contained in the Series, in versions 0.6.1 and prior, Series would fall back on a location-based lookup. This now raises a KeyError:

In [2]: s[1]
KeyError: 1

This change also has the same impact on DataFrame:

In [3]: df = DataFrame(randn(8, 4), index=range(0, 16, 2))

In [4]: df
    0        1       2       3
0   0.88427  0.3363 -0.1787  0.03162
2   0.14451 -0.1415  0.2504  0.58374
4  -1.44779 -0.9186 -1.4996  0.27163
6  -0.26598 -2.4184 -0.2658  0.11503
8  -0.58776  0.3144 -0.8566  0.61941
10  0.10940 -0.7175 -1.0108  0.47990
12 -1.16919 -0.3087 -0.6049 -0.43544
14 -0.07337  0.3410  0.0424 -0.16037

In [5]: df.ix[3]
KeyError: 3

In order to support purely integer-based indexing, the following methods have been added:

Method Description
Series.iget_value(i) Retrieve value stored at location i
Series.iget(i) Alias for iget_value
DataFrame.irow(i) Retrieve the i-th row
DataFrame.icol(j) Retrieve the j-th column
DataFrame.iget_value(i, j) Retrieve the value at row i and column j

API tweaks regarding label-based slicing

Label-based slicing using ix now requires that the index be sorted (monotonic) unless both the start and endpoint are contained in the index:

In [1]: s = Series(randn(6), index=list('gmkaec'))

In [2]: s
Out[2]:
g   -1.182230
m   -0.276183
k   -0.243550
a    1.628992
e    0.073308
c   -0.539890
dtype: float64

Then this is OK:

In [3]: s.ix['k':'e']
Out[3]:
k   -0.243550
a    1.628992
e    0.073308
dtype: float64

But this is not:

In [12]: s.ix['b':'h']
KeyError 'b'

If the index had been sorted, the “range selection” would have been possible:

In [4]: s2 = s.sort_index()

In [5]: s2
Out[5]:
a    1.628992
c   -0.539890
e    0.073308
g   -1.182230
k   -0.243550
m   -0.276183
dtype: float64

In [6]: s2.ix['b':'h']
Out[6]:
c   -0.539890
e    0.073308
g   -1.182230
dtype: float64

Changes to Series [] operator

As as notational convenience, you can pass a sequence of labels or a label slice to a Series when getting and setting values via [] (i.e. the __getitem__ and __setitem__ methods). The behavior will be the same as passing similar input to ix except in the case of integer indexing:

In [8]: s = Series(randn(6), index=list('acegkm'))

In [9]: s
Out[9]: 
a   -0.800734
c   -0.229737
e   -0.781940
g    0.756053
k    2.613373
m   -0.159310
Length: 6, dtype: float64

In [10]: s[['m', 'a', 'c', 'e']]
Out[10]: 
m   -0.159310
a   -0.800734
c   -0.229737
e   -0.781940
Length: 4, dtype: float64

In [11]: s['b':'l']
Out[11]: 
c   -0.229737
e   -0.781940
g    0.756053
k    2.613373
Length: 4, dtype: float64

In [12]: s['c':'k']
Out[12]: 
c   -0.229737
e   -0.781940
g    0.756053
k    2.613373
Length: 4, dtype: float64

In the case of integer indexes, the behavior will be exactly as before (shadowing ndarray):

In [13]: s = Series(randn(6), index=range(0, 12, 2))

In [14]: s[[4, 0, 2]]
Out[14]: 
4    0.022862
0   -0.321246
2   -0.707337
Length: 3, dtype: float64

In [15]: s[1:5]
Out[15]: 
2   -0.707337
4    0.022862
6    0.306713
8   -0.162222
Length: 4, dtype: float64

If you wish to do indexing with sequences and slicing on an integer index with label semantics, use ix.

Other API Changes

  • The deprecated LongPanel class has been completely removed
  • If Series.sort is called on a column of a DataFrame, an exception will now be raised. Before it was possible to accidentally mutate a DataFrame’s column by doing df[col].sort() instead of the side-effect free method df[col].order() (GH316)
  • Miscellaneous renames and deprecations which will (harmlessly) raise FutureWarning
  • drop added as an optional parameter to DataFrame.reset_index (GH699)

Performance improvements

  • Cythonized GroupBy aggregations no longer presort the data, thus achieving a significant speedup (GH93). GroupBy aggregations with Python functions significantly sped up by clever manipulation of the ndarray data type in Cython (GH496).
  • Better error message in DataFrame constructor when passed column labels don’t match data (GH497)
  • Substantially improve performance of multi-GroupBy aggregation when a Python function is passed, reuse ndarray object in Cython (GH496)
  • Can store objects indexed by tuples and floats in HDFStore (GH492)
  • Don’t print length by default in Series.to_string, add length option (GH489)
  • Improve Cython code for multi-groupby to aggregate without having to sort the data (GH93)
  • Improve MultiIndex reindexing speed by storing tuples in the MultiIndex, test for backwards unpickling compatibility
  • Improve column reindexing performance by using specialized Cython take function
  • Further performance tweaking of Series.__getitem__ for standard use cases
  • Avoid Index dict creation in some cases (i.e. when getting slices, etc.), regression from prior versions
  • Friendlier error message in setup.py if NumPy not installed
  • Use common set of NA-handling operations (sum, mean, etc.) in Panel class also (GH536)
  • Default name assignment when calling reset_index on DataFrame with a regular (non-hierarchical) index (GH476)
  • Use Cythonized groupers when possible in Series/DataFrame stat ops with level parameter passed (GH545)
  • Ported skiplist data structure to C to speed up rolling_median by about 5-10x in most typical use cases (GH374)

v.0.6.1 (December 13, 2011)

New features

Performance improvements

  • Improve memory usage of DataFrame.describe (do not copy data unnecessarily) (PR #425)
  • Optimize scalar value lookups in the general case by 25% or more in Series and DataFrame
  • Fix performance regression in cross-sectional count in DataFrame, affecting DataFrame.dropna speed
  • Column deletion in DataFrame copies no data (computes views on blocks) (GH #158)

v.0.6.0 (November 25, 2011)

New Features

  • Added melt function to pandas.core.reshape
  • Added level parameter to group by level in Series and DataFrame descriptive statistics (GH313)
  • Added head and tail methods to Series, analogous to to DataFrame (GH296)
  • Added Series.isin function which checks if each value is contained in a passed sequence (GH289)
  • Added float_format option to Series.to_string
  • Added skip_footer (GH291) and converters (GH343) options to read_csv and read_table
  • Added drop_duplicates and duplicated functions for removing duplicate DataFrame rows and checking for duplicate rows, respectively (GH319)
  • Implemented operators ‘&’, ‘|’, ‘^’, ‘-‘ on DataFrame (GH347)
  • Added Series.mad, mean absolute deviation
  • Added QuarterEnd DateOffset (GH321)
  • Added dot to DataFrame (GH65)
  • Added orient option to Panel.from_dict (GH359, GH301)
  • Added orient option to DataFrame.from_dict
  • Added passing list of tuples or list of lists to DataFrame.from_records (GH357)
  • Added multiple levels to groupby (GH103)
  • Allow multiple columns in by argument of DataFrame.sort_index (GH92, GH362)
  • Added fast get_value and put_value methods to DataFrame (GH360)
  • Added cov instance methods to Series and DataFrame (GH194, GH362)
  • Added kind='bar' option to DataFrame.plot (GH348)
  • Added idxmin and idxmax to Series and DataFrame (GH286)
  • Added read_clipboard function to parse DataFrame from clipboard (GH300)
  • Added nunique function to Series for counting unique elements (GH297)
  • Made DataFrame constructor use Series name if no columns passed (GH373)
  • Support regular expressions in read_table/read_csv (GH364)
  • Added DataFrame.to_html for writing DataFrame to HTML (GH387)
  • Added support for MaskedArray data in DataFrame, masked values converted to NaN (GH396)
  • Added DataFrame.boxplot function (GH368)
  • Can pass extra args, kwds to DataFrame.apply (GH376)
  • Implement DataFrame.join with vector on argument (GH312)
  • Added legend boolean flag to DataFrame.plot (GH324)
  • Can pass multiple levels to stack and unstack (GH370)
  • Can pass multiple values columns to pivot_table (GH381)
  • Use Series name in GroupBy for result index (GH363)
  • Added raw option to DataFrame.apply for performance if only need ndarray (GH309)
  • Added proper, tested weighted least squares to standard and panel OLS (GH303)

Performance Enhancements

  • VBENCH Cythonized cache_readonly, resulting in substantial micro-performance enhancements throughout the codebase (GH361)
  • VBENCH Special Cython matrix iterator for applying arbitrary reduction operations with 3-5x better performance than np.apply_along_axis (GH309)
  • VBENCH Improved performance of MultiIndex.from_tuples
  • VBENCH Special Cython matrix iterator for applying arbitrary reduction operations
  • VBENCH + DOCUMENT Add raw option to DataFrame.apply for getting better performance when
  • VBENCH Faster cythonized count by level in Series and DataFrame (GH341)
  • VBENCH? Significant GroupBy performance enhancement with multiple keys with many “empty” combinations
  • VBENCH New Cython vectorized function map_infer speeds up Series.apply and Series.map significantly when passed elementwise Python function, motivated by (GH355)
  • VBENCH Significantly improved performance of Series.order, which also makes np.unique called on a Series faster (GH327)
  • VBENCH Vastly improved performance of GroupBy on axes with a MultiIndex (GH299)

v.0.5.0 (October 24, 2011)

New Features

  • Added DataFrame.align method with standard join options
  • Added parse_dates option to read_csv and read_table methods to optionally try to parse dates in the index columns
  • Added nrows, chunksize, and iterator arguments to read_csv and read_table. The last two return a new TextParser class capable of lazily iterating through chunks of a flat file (GH242)
  • Added ability to join on multiple columns in DataFrame.join (GH214)
  • Added private _get_duplicates function to Index for identifying duplicate values more easily (ENH5c)
  • Added column attribute access to DataFrame.
  • Added Python tab completion hook for DataFrame columns. (GH233, GH230)
  • Implemented Series.describe for Series containing objects (GH241)
  • Added inner join option to DataFrame.join when joining on key(s) (GH248)
  • Implemented selecting DataFrame columns by passing a list to __getitem__ (GH253)
  • Implemented & and | to intersect / union Index objects, respectively (GH261)
  • Added pivot_table convenience function to pandas namespace (GH234)
  • Implemented Panel.rename_axis function (GH243)
  • DataFrame will show index level names in console output (GH334)
  • Implemented Panel.take
  • Added set_eng_float_format for alternate DataFrame floating point string formatting (ENH61)
  • Added convenience set_index function for creating a DataFrame index from its existing columns
  • Implemented groupby hierarchical index level name (GH223)
  • Added support for different delimiters in DataFrame.to_csv (GH244)
  • TODO: DOCS ABOUT TAKE METHODS

Performance Enhancements

  • VBENCH Major performance improvements in file parsing functions read_csv and read_table
  • VBENCH Added Cython function for converting tuples to ndarray very fast. Speeds up many MultiIndex-related operations
  • VBENCH Refactored merging / joining code into a tidy class and disabled unnecessary computations in the float/object case, thus getting about 10% better performance (GH211)
  • VBENCH Improved speed of DataFrame.xs on mixed-type DataFrame objects by about 5x, regression from 0.3.0 (GH215)
  • VBENCH With new DataFrame.align method, speeding up binary operations between differently-indexed DataFrame objects by 10-25%.
  • VBENCH Significantly sped up conversion of nested dict into DataFrame (GH212)
  • VBENCH Significantly speed up DataFrame __repr__ and count on large mixed-type DataFrame objects

v.0.4.3 through v0.4.1 (September 25 - October 9, 2011)

New Features

  • Added Python 3 support using 2to3 (GH200)
  • Added name attribute to Series, now prints as part of Series.__repr__
  • Added instance methods isnull and notnull to Series (GH209, GH203)
  • Added Series.align method for aligning two series with choice of join method (ENH56)
  • Added method get_level_values to MultiIndex (GH188)
  • Set values in mixed-type DataFrame objects via .ix indexing attribute (GH135)
  • Added new DataFrame methods get_dtype_counts and property dtypes (ENHdc)
  • Added ignore_index option to DataFrame.append to stack DataFrames (ENH1b)
  • read_csv tries to sniff delimiters using csv.Sniffer (GH146)
  • read_csv can read multiple columns into a MultiIndex; DataFrame’s to_csv method writes out a corresponding MultiIndex (GH151)
  • DataFrame.rename has a new copy parameter to rename a DataFrame in place (ENHed)
  • Enable unstacking by name (GH142)
  • Enable sortlevel to work by level (GH141)

Performance Enhancements

  • Altered binary operations on differently-indexed SparseSeries objects to use the integer-based (dense) alignment logic which is faster with a larger number of blocks (GH205)
  • Wrote faster Cython data alignment / merging routines resulting in substantial speed increases
  • Improved performance of isnull and notnull, a regression from v0.3.0 (GH187)
  • Refactored code related to DataFrame.join so that intermediate aligned copies of the data in each DataFrame argument do not need to be created. Substantial performance increases result (GH176)
  • Substantially improved performance of generic Index.intersection and Index.union
  • Implemented BlockManager.take resulting in significantly faster take performance on mixed-type DataFrame objects (GH104)
  • Improved performance of Series.sort_index
  • Significant groupby performance enhancement: removed unnecessary integrity checks in DataFrame internals that were slowing down slicing operations to retrieve groups
  • Optimized _ensure_index function resulting in performance savings in type-checking Index objects
  • Wrote fast time series merging / joining methods in Cython. Will be integrated later into DataFrame.join and related functions
Scroll To Top