Table Of Contents

Search

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

Release Notes

This is the list of changes to pandas between each release. For full details, see the commit logs at http://github.com/pydata/pandas

What is it

pandas is a Python package providing fast, flexible, and expressive data structures designed to make working with “relational” or “labeled” data both easy and intuitive. It aims to be the fundamental high-level building block for doing practical, real world data analysis in Python. Additionally, it has the broader goal of becoming the most powerful and flexible open source data analysis / manipulation tool available in any language.

Where to get it

pandas 0.15.0

Release date: (October 18, 2014)

This is a major release from 0.14.1 and includes a number of API changes, several new features, enhancements, and performance improvements along with a large number of bug fixes.

Highlights include:

  • Drop support for numpy < 1.7.0 (GH7711)
  • The Categorical type was integrated as a first-class pandas type, see here
  • New scalar type Timedelta, and a new index type TimedeltaIndex, see here
  • New DataFrame default display for df.info() to include memory usage, see Memory Usage
  • New datetimelike properties accessor .dt for Series, see Datetimelike Properties
  • Split indexing documentation into Indexing and Selecting Data and MultiIndex / Advanced Indexing
  • Split out string methods documentation into Working with Text Data
  • read_csv will now by default ignore blank lines when parsing, see here
  • API change in using Indexes in set operations, see here
  • Internal refactoring of the Index class to no longer sub-class ndarray, see Internal Refactoring
  • dropping support for PyTables less than version 3.0.0, and numexpr less than version 2.1 (GH7990)

See the v0.15.0 Whatsnew overview or the issue tracker on GitHub for an extensive list of all API changes, enhancements and bugs that have been fixed in 0.15.0.

Thanks

  • Aaron Schumacher
  • Adam Greenhall
  • Andy Hayden
  • Anthony O’Brien
  • Artemy Kolchinsky
  • behzad nouri
  • Benedikt Sauer
  • benjamin
  • Benjamin Thyreau
  • Ben Schiller
  • bjonen
  • BorisVerk
  • Chris Reynolds
  • Chris Stoafer
  • Dav Clark
  • dlovell
  • DSM
  • dsm054
  • FragLegs
  • German Gomez-Herrero
  • Hsiaoming Yang
  • Huan Li
  • hunterowens
  • Hyungtae Kim
  • immerrr
  • Isaac Slavitt
  • ischwabacher
  • Jacob Schaer
  • Jacob Wasserman
  • Jan Schulz
  • Jeff Tratner
  • Jesse Farnham
  • jmorris0x0
  • jnmclarty
  • Joe Bradish
  • Joerg Rittinger
  • John W. O’Brien
  • Joris Van den Bossche
  • jreback
  • Kevin Sheppard
  • klonuo
  • Kyle Meyer
  • lexual
  • Max Chang
  • mcjcode
  • Michael Mueller
  • Michael W Schatzow
  • Mike Kelly
  • Mortada Mehyar
  • mtrbean
  • Nathan Sanders
  • Nathan Typanski
  • onesandzeroes
  • Paul Masurel
  • Phillip Cloud
  • Pietro Battiston
  • RenzoBertocchi
  • rockg
  • Ross Petchler
  • seth-p
  • Shahul Hameed
  • Shashank Agarwal
  • sinhrks
  • someben
  • stahlous
  • stas-sl
  • Stephan Hoyer
  • thatneat
  • tom-alcorn
  • TomAugspurger
  • Tom Augspurger
  • Tony Lorenzo
  • unknown
  • unutbu
  • Wes Turner
  • Wilfred Hughes
  • Yevgeniy Grechka
  • Yoshiki Vázquez Baeza
  • zachcp

pandas 0.14.1

Release date: (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.

Highlights include:

  • New methods select_dtypes() to select columns based on the dtype and sem() to calculate the standard error of the mean.
  • Support for dateutil timezones (see docs).
  • Support for ignoring full line comments in the read_csv() text parser.
  • New documentation section on Options and Settings.
  • Lots of bug fixes.

See the v0.14.1 Whatsnew overview or the issue tracker on GitHub for an extensive list of all API changes, enhancements and bugs that have been fixed in 0.14.1.

Thanks

  • Andrew Rosenfeld
  • Andy Hayden
  • Benjamin Adams
  • Benjamin M. Gross
  • Brian Quistorff
  • Brian Wignall
  • bwignall
  • clham
  • Daniel Waeber
  • David Bew
  • David Stephens
  • DSM
  • dsm054
  • helger
  • immerrr
  • Jacob Schaer
  • jaimefrio
  • Jan Schulz
  • John David Reaver
  • John W. O’Brien
  • Joris Van den Bossche
  • jreback
  • Julien Danjou
  • Kevin Sheppard
  • K.-Michael Aye
  • Kyle Meyer
  • lexual
  • Matthew Brett
  • Matt Wittmann
  • Michael Mueller
  • Mortada Mehyar
  • onesandzeroes
  • Phillip Cloud
  • Rob Levy
  • rockg
  • sanguineturtle
  • Schaer, Jacob C
  • seth-p
  • sinhrks
  • Stephan Hoyer
  • Thomas Kluyver
  • Todd Jennings
  • TomAugspurger
  • unknown
  • yelite

pandas 0.14.0

Release date: (May 31, 2014)

This is a major release from 0.13.1 and includes a number of API changes, several new features, enhancements, and performance improvements along with a large number of bug fixes.

Highlights include:

  • Officially support Python 3.4
  • SQL interfaces updated to use sqlalchemy, see here.
  • Display interface changes, see here
  • MultiIndexing using Slicers, see here.
  • Ability to join a singly-indexed DataFrame with a multi-indexed DataFrame, see here
  • More consistency in groupby results and more flexible groupby specifications, see here
  • Holiday calendars are now supported in CustomBusinessDay, see here
  • Several improvements in plotting functions, including: hexbin, area and pie plots, see here.
  • Performance doc section on I/O operations, see here

See the v0.14.0 Whatsnew overview or the issue tracker on GitHub for an extensive list of all API changes, enhancements and bugs that have been fixed in 0.14.0.

Thanks

  • Acanthostega
  • Adam Marcus
  • agijsberts
  • akittredge
  • Alex Gaudio
  • Alex Rothberg
  • AllenDowney
  • Andrew Rosenfeld
  • Andy Hayden
  • ankostis
  • anomrake
  • Antoine Mazières
  • anton-d
  • bashtage
  • Benedikt Sauer
  • benjamin
  • Brad Buran
  • bwignall
  • cgohlke
  • chebee7i
  • Christopher Whelan
  • Clark Fitzgerald
  • clham
  • Dale Jung
  • Dan Allan
  • Dan Birken
  • danielballan
  • Daniel Waeber
  • David Jung
  • David Stephens
  • Douglas McNeil
  • DSM
  • Garrett Drapala
  • Gouthaman Balaraman
  • Guillaume Poulin
  • hshimizu77
  • hugo
  • immerrr
  • ischwabacher
  • Jacob Howard
  • Jacob Schaer
  • jaimefrio
  • Jason Sexauer
  • Jeff Reback
  • Jeffrey Starr
  • Jeff Tratner
  • John David Reaver
  • John McNamara
  • John W. O’Brien
  • Jonathan Chambers
  • Joris Van den Bossche
  • jreback
  • jsexauer
  • Julia Evans
  • Júlio
  • Katie Atkinson
  • kdiether
  • Kelsey Jordahl
  • Kevin Sheppard
  • K.-Michael Aye
  • Matthias Kuhn
  • Matt Wittmann
  • Max Grender-Jones
  • Michael E. Gruen
  • michaelws
  • mikebailey
  • Mike Kelly
  • Nipun Batra
  • Noah Spies
  • ojdo
  • onesandzeroes
  • Patrick O’Keeffe
  • phaebz
  • Phillip Cloud
  • Pietro Battiston
  • PKEuS
  • Randy Carnevale
  • ribonoous
  • Robert Gibboni
  • rockg
  • sinhrks
  • Skipper Seabold
  • SplashDance
  • Stephan Hoyer
  • Tim Cera
  • Tobias Brandt
  • Todd Jennings
  • TomAugspurger
  • Tom Augspurger
  • unutbu
  • westurner
  • Yaroslav Halchenko
  • y-p
  • zach powers

pandas 0.13.1

Release date: (February 3, 2014)

New Features

  • Added date_format and datetime_format attribute to ExcelWriter. (GH4133)

API Changes

  • Series.sort will raise a ValueError (rather than a TypeError) on sorting an object that is a view of another (GH5856, GH5853)
  • Raise/Warn SettingWithCopyError (according to the option chained_assignment in more cases, when detecting chained assignment, related (GH5938, GH6025)
  • DataFrame.head(0) returns self instead of empty frame (GH5846)
  • autocorrelation_plot now accepts **kwargs. (GH5623)
  • convert_objects now accepts a convert_timedeltas='coerce' argument to allow forced dtype conversion of timedeltas (GH5458,:issue:5689)
  • Add -NaN and -nan to the default set of NA values (GH5952). See NA Values.
  • NDFrame now has an equals method. (GH5283)
  • DataFrame.apply will use the reduce argument to determine whether a Series or a DataFrame should be returned when the DataFrame is empty (GH6007).

Experimental Features

Improvements to existing features

  • perf improvements in Series datetime/timedelta binary operations (GH5801)
  • option_context context manager now available as top-level API (GH5752)
  • df.info() view now display dtype info per column (GH5682)
  • df.info() now honors option max_info_rows, disable null counts for large frames (GH5974)
  • perf improvements in 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)
  • support dtypes property on Series/Panel/Panel4D
  • extend Panel.apply to allow arbitrary functions (rather than only ufuncs) (GH1148) allow multiple axes to be used to operate on slabs of a Panel
  • The ArrayFormatter for datetime and timedelta64 now intelligently limit precision based on the values in the array (GH3401)
  • pd.show_versions() is now available for convenience when reporting issues.
  • perf improvements to Series.str.extract (GH5944)
  • perf improvements in dtypes/ftypes methods (GH5968)
  • perf improvements in indexing with object dtypes (GH5968)
  • improved dtype inference for timedelta like passed to constructors (GH5458, GH5689)
  • escape special characters when writing to latex (:issue: 5374)
  • perf improvements in DataFrame.apply (GH6013)
  • 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,:issue:6021)
  • add ability to recognize ‘%p’ format code (am/pm) to date parsers when the specific format is supplied (GH5361)
  • Fix performance regression in JSON IO (GH5765)
  • performance regression in Index construction from Series (GH6150)

Bug Fixes

  • Bug in io.wb.get_countries not including all countries (GH6008)
  • Bug in Series replace with timestamp dict (GH5797)
  • read_csv/read_table now respects the prefix kwarg (GH5732).
  • Bug in selection with missing values via .ix from a duplicate indexed DataFrame failing (GH5835)
  • Fix issue of boolean comparison on empty DataFrames (GH5808)
  • Bug in isnull handling NaT in an object array (GH5443)
  • Bug in to_datetime when passed a np.nan or integer datelike and a format string (GH5863)
  • Bug in groupby dtype conversion with datetimelike (GH5869)
  • Regression in handling of empty Series as indexers to Series (GH5877)
  • Bug in internal caching, related to (GH5727)
  • Testing bug in reading JSON/msgpack from a non-filepath on windows under py3 (GH5874)
  • Bug when assigning to .ix[tuple(...)] (GH5896)
  • Bug in fully reindexing a Panel (GH5905)
  • Bug in idxmin/max with object dtypes (GH5914)
  • Bug in BusinessDay when adding n days to a date not on offset when n>5 and n%5==0 (GH5890)
  • Bug in assigning to chained series with a series via ix (GH5928)
  • Bug in creating an empty DataFrame, copying, then assigning (GH5932)
  • Bug in DataFrame.tail with empty frame (GH5846)
  • Bug in propagating metadata on resample (GH5862)
  • Fixed string-representation of NaT to be “NaT” (GH5708)
  • Fixed string-representation for Timestamp to show nanoseconds if present (GH5912)
  • pd.match not returning passed sentinel
  • Panel.to_frame() no longer fails when major_axis is a MultiIndex (GH5402).
  • Bug in pd.read_msgpack with inferring a DateTimeIndex frequency incorrectly (GH5947)
  • Fixed to_datetime for array with both Tz-aware datetimes and NaT‘s (GH5961)
  • Bug in rolling skew/kurtosis when passed a Series with bad data (GH5749)
  • Bug in scipy interpolate methods with a datetime index (GH5975)
  • Bug in NaT comparison if a mixed datetime/np.datetime64 with NaT were passed (GH5968)
  • Fixed bug with pd.concat losing dtype information if all inputs are empty (GH5742)
  • Recent changes in IPython cause warnings to be emitted when using previous versions of pandas in QTConsole, now fixed. If you’re using an older version and need to suppress the warnings, see (GH5922).
  • Bug in merging timedelta dtypes (GH5695)
  • Bug in plotting.scatter_matrix function. Wrong alignment among diagonal and off-diagonal plots, see (GH5497).
  • Regression in Series with a multi-index via ix (GH6018)
  • Bug in Series.xs with a multi-index (GH6018)
  • Bug in Series construction of mixed type with datelike and an integer (which should result in object type and not automatic conversion) (GH6028)
  • Possible segfault when chained indexing with an object array under numpy 1.7.1 (GH6026, GH6056)
  • Bug in setting using fancy indexing a single element with a non-scalar (e.g. a list), (GH6043)
  • to_sql did not respect if_exists (GH4110 GH4304)
  • Regression in .get(None) indexing from 0.12 (GH5652)
  • Subtle iloc indexing bug, surfaced in (GH6059)
  • Bug with insert of strings into DatetimeIndex (GH5818)
  • Fixed unicode bug in to_html/HTML repr (GH6098)
  • Fixed missing arg validation in get_options_data (GH6105)
  • Bug in assignment with duplicate columns in a frame where the locations are a slice (e.g. next to each other) (GH6120)
  • Bug in propogating _ref_locs during construction of a DataFrame with dups index/columns (GH6121)
  • Bug in DataFrame.apply when using mixed datelike reductions (GH6125)
  • Bug in DataFrame.append when appending a row with different columns (GH6129)
  • Bug in DataFrame construction with recarray and non-ns datetime dtype (GH6140)
  • Bug in .loc setitem indexing with a dataframe on rhs, multiple item setting, and a datetimelike (GH6152)
  • Fixed a bug in query/eval during lexicographic string comparisons (GH6155).
  • Fixed a bug in query where the index of a single-element Series was being thrown away (GH6148).
  • Bug in HDFStore on appending a dataframe with multi-indexed columns to an existing table (GH6167)
  • Consistency with dtypes in setting an empty DataFrame (GH6171)
  • Bug in selecting on a multi-index HDFStore even in the presence of under specified column spec (GH6169)
  • Bug in nanops.var with ddof=1 and 1 elements would sometimes return inf rather than nan on some platforms (GH6136)
  • Bug in Series and DataFrame bar plots ignoring the use_index keyword (GH6209)
  • Bug in groupby with mixed str/int under python3 fixed; argsort was failing (GH6212)

pandas 0.13.0

Release date: January 3, 2014

New Features

  • 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 indicies at which it is evaluated, respectively. See scipy docs. (GH4298)
  • Added isin method to DataFrame (GH4211)
  • df.to_clipboard() learned a new excel keyword that let’s you paste df data directly into excel (enabled by default). (GH5070).
  • Clipboard functionality now works with PySide (GH4282)
  • New extract string method returns regex matches more conveniently (GH4685)
  • Auto-detect field widths in read_fwf when unspecified (GH4488)
  • to_csv() now outputs datetime objects according to a specified format string via the date_format keyword (GH4313)
  • Added LastWeekOfMonth DateOffset (GH4637)
  • Added cumcount groupby method (GH4646)
  • Added FY5253, and FY5253Quarter DateOffsets (GH4511)
  • Added mode() method to Series and DataFrame to get the statistical mode(s) of a column/series. (GH5367)

Experimental Features

  • The new eval() function implements expression evaluation using numexpr behind the scenes. This results in large speedups for complicated expressions involving large DataFrames/Series.
  • DataFrame has a new eval() that evaluates an expression in the context of the DataFrame; allows inline expression assignment
  • A query() method has been added that allows you to select elements of a DataFrame using a natural query syntax nearly identical to Python syntax.
  • pd.eval and friends now evaluate operations involving datetime64 objects in Python space because numexpr cannot handle NaT values (GH4897).
  • Add msgpack support via pd.read_msgpack() and pd.to_msgpack() / df.to_msgpack() for serialization of arbitrary pandas (and python objects) in a lightweight portable binary format (GH686, GH5506)
  • Added PySide support for the qtpandas DataFrameModel and DataFrameWidget.
  • Added pandas.io.gbq for reading from (and writing to) Google BigQuery into a DataFrame. (GH4140)

Improvements to existing features

  • read_html now raises a URLError instead of catching and raising a ValueError (GH4303, GH4305)
  • read_excel now supports an integer in its sheetname argument giving the index of the sheet to read in (GH4301).
  • get_dummies works with NaN (GH4446)
  • Added a test for read_clipboard() and to_clipboard() (GH4282)
  • Added bins argument to value_counts (GH3945), also sort and ascending, now available in Series method as well as top-level function.
  • Text parser now treats anything that reads like inf (“inf”, “Inf”, “-Inf”, “iNf”, etc.) to infinity. (GH4220, GH4219), affecting read_table, read_csv, etc.
  • Added a more informative error message when plot arguments contain overlapping color and style arguments (GH4402)
  • Significant table writing performance improvements in HDFStore
  • JSON date serialization now performed in low-level C code.
  • JSON support for encoding datetime.time
  • Expanded JSON docs, more info about orient options and the use of the numpy param when decoding.
  • Add drop_level argument to xs (GH4180)
  • Can now resample a DataFrame with ohlc (GH2320)
  • Index.copy() and MultiIndex.copy() now accept keyword arguments to change attributes (i.e., names, levels, labels) (GH4039)
  • Add rename and set_names methods to Index as well as set_names, set_levels, set_labels to MultiIndex. (GH4039) with improved validation for all (GH4039, GH4794)
  • A Series of dtype timedelta64[ns] can now be divided/multiplied by an integer series (GH4521)
  • A Series of dtype timedelta64[ns] can now be divided by another timedelta64[ns] object to yield a float64 dtyped Series. This is frequency conversion; astyping is also supported.
  • Timedelta64 support fillna/ffill/bfill with an integer interpreted as seconds, or a timedelta (GH3371)
  • Box numeric ops on timedelta Series (GH4984)
  • Datetime64 support ffill/bfill
  • Performance improvements with __getitem__ on DataFrames with when the key is a column
  • Support for using a DatetimeIndex/PeriodsIndex directly in a datelike calculation e.g. s-s.index (GH4629)
  • Better/cleaned up exceptions in core/common, io/excel and core/format (GH4721, GH3954), as well as cleaned up test cases in tests/test_frame, tests/test_multilevel (GH4732).
  • Performance improvement of timeseries plotting with PeriodIndex and added test to vbench (GH4705 and GH4722)
  • Add axis and level keywords to where, so that the other argument can now be an alignable pandas object.
  • to_datetime with a format of ‘%Y%m%d’ now parses much faster
  • It’s now easier to hook new Excel writers into pandas (just subclass ExcelWriter and register your engine). You can specify an engine in to_excel or in ExcelWriter. You can also specify which writers you want to use by default with config options io.excel.xlsx.writer and io.excel.xls.writer. (GH4745, GH4750)
  • Panel.to_excel() now accepts keyword arguments that will be passed to its DataFrame‘s to_excel() methods. (GH4750)
  • Added XlsxWriter as an optional ExcelWriter engine. This is about 5x faster than the default openpyxl xlsx writer and is equivalent in speed to the xlwt xls writer module. (GH4542)
  • allow DataFrame constructor to accept more list-like objects, e.g. list of collections.Sequence and array.Array objects (GH3783, GH4297, GH4851), thanks @lgautier
  • DataFrame constructor now accepts a numpy masked record array (GH3478), thanks @jnothman
  • __getitem__ with tuple key (e.g., [:, 2]) on Series without MultiIndex raises ValueError (GH4759, GH4837)
  • read_json now raises a (more informative) ValueError when the dict contains a bad key and orient='split' (GH4730, GH4838)
  • read_stata now accepts Stata 13 format (GH4291)
  • ExcelWriter and ExcelFile can be used as contextmanagers. (GH3441, GH4933)
  • pandas is now tested with two different versions of statsmodels (0.4.3 and 0.5.0) (GH4981).
  • Better string representations of MultiIndex (including ability to roundtrip via repr). (GH3347, GH4935)
  • Both ExcelFile and read_excel to accept an xlrd.Book for the io (formerly path_or_buf) argument; this requires engine to be set. (GH4961).
  • concat now gives a more informative error message when passed objects that cannot be concatenated (GH4608).
  • Add halflife option to exponentially weighted moving functions (PR GH4998)
  • to_dict now takes records as a possible outtype. Returns an array of column-keyed dictionaries. (GH4936)
  • tz_localize can infer a fall daylight savings transition based on the structure of unlocalized data (GH4230)
  • DatetimeIndex is now in the API documentation
  • Improve support for converting R datasets to pandas objects (more informative index for timeseries and numeric, support for factors, dist, and high-dimensional arrays).
  • read_html() now supports the parse_dates, tupleize_cols and thousands parameters (GH4770).
  • json_normalize() is a new method to allow you to create a flat table from semi-structured JSON data. See the docs (GH1067)
  • DataFrame.from_records() will now accept generators (GH4910)
  • DataFrame.interpolate() and Series.interpolate() have been expanded to include interpolation methods from scipy. (GH4434, GH1892)
  • Series now supports a to_frame method to convert it to a single-column DataFrame (GH5164)
  • DatetimeIndex (and date_range) can now be constructed in a left- or right-open fashion using the closed parameter (GH4579)
  • Python csv parser now supports usecols (GH4335)
  • Added support for Google Analytics v3 API segment IDs that also supports v2 IDs. (GH5271)
  • NDFrame.drop() now accepts names as well as integers for the axis argument. (GH5354)
  • Added short docstrings to a few methods that were missing them + fixed the docstrings for Panel flex methods. (GH5336)
  • NDFrame.drop(), NDFrame.dropna(), and .drop_duplicates() all accept inplace as a keyword argument; however, this only means that the wrapper is updated inplace, a copy is still made internally. (GH1960, GH5247, GH5628, and related GH2325 [still not closed])
  • Fixed bug in tools.plotting.andrews_curvres so that lines are drawn grouped by color as expected.
  • read_excel() now tries to convert integral floats (like 1.0) to int by default. (GH5394)
  • Excel writers now have a default option merge_cells in to_excel() to merge cells in MultiIndex and Hierarchical Rows. Note: using this option it is no longer possible to round trip Excel files with merged MultiIndex and Hierarchical Rows. Set the merge_cells to False to restore the previous behaviour. (GH5254)
  • The FRED DataReader now accepts multiple series (:issue`3413`)
  • StataWriter adjusts variable names to Stata’s limitations (GH5709)

API Changes

  • DataFrame.reindex() and forward/backward filling now raises ValueError if either index is not monotonic (GH4483, GH4484).
  • 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)
  • deprecated iterkv, which will be removed in a future release (was just an alias of iteritems used to get around 2to3‘s changes). (GH4384, GH4375, GH4372)
  • Series.get with negative indexers now returns the same as [] (GH4390)
  • allow ix/loc for Series/DataFrame/Panel to set on any axis even when the single-key is not currently contained in the index for that axis (GH2578, GH5226, GH5632, GH5720, GH5744, GH5756)
  • Default export for to_clipboard is now csv with a sep of t for compat (GH3368)
  • at now will enlarge the object inplace (and return the same) (GH2578)
  • DataFrame.plot will scatter plot x versus y by passing kind='scatter' (GH2215)
  • HDFStore
    • append_to_multiple automatically synchronizes writing rows to multiple tables and adds a dropna kwarg (GH4698)
    • handle a passed Series in table format (GH4330)
    • 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
    • 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)
    • 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)
    • the format keyword now replaces the table keyword; allowed values are fixed(f)|table(t) the Storer format has been renamed to Fixed
    • a column multi-index will be recreated properly (GH4710); raise on trying to use a multi-index with data_columns on the same axis
    • select_as_coordinates will now return an Int64Index of the resultant selection set
    • support timedelta64[ns] as a serialization type (GH3577)
    • store datetime.date objects as ordinals rather then timetuples to avoid timezone issues (GH2852), thanks @tavistmorph and @numpand
    • numexpr 2.2.2 fixes incompatibility in PyTables 2.4 (GH4908)
    • flush now accepts an fsync parameter, which defaults to False (GH5364)
    • unicode indices not supported on table formats (GH5386)
    • pass thru store creation arguments; can be used to support in-memory stores
  • JSON
    • added date_unit parameter to specify resolution of timestamps. Options are seconds, milliseconds, microseconds and nanoseconds. (GH4362, GH4498).
    • added default_handler parameter to allow a callable to be passed which will be responsible for handling otherwise unserialiable objects. (GH5138)
  • Index and MultiIndex changes (GH4039):
    • Setting levels and labels directly on MultiIndex is now deprecated. Instead, you can use the set_levels() and set_labels() methods.
    • levels, labels and names properties no longer return lists, but instead return containers that do not allow setting of items (‘mostly immutable’)
    • levels, labels and names are validated upon setting and are either copied or shallow-copied.
    • inplace setting of levels or labels now correctly invalidates the cached properties. (GH5238).
    • __deepcopy__ now returns a shallow copy (currently: a view) of the data - allowing metadata changes.
    • MultiIndex.astype() now only allows np.object_-like dtypes and now returns a MultiIndex rather than an Index. (GH4039)
    • Added is_ method to Index that allows fast equality comparison of views (similar to np.may_share_memory but no false positives, and changes on levels and labels setting on MultiIndex). (GH4859 , GH4909)
    • Aliased __iadd__ to __add__. (GH4996)
    • Added is_ method to Index that allows fast equality comparison of views (similar to np.may_share_memory but no false positives, and changes on levels and labels setting on MultiIndex). (GH4859, GH4909)
  • 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. Add .bool() method to NDFrame objects to facilitate evaluating of single-element boolean Series
  • DataFrame.update() no longer raises a DataConflictError, it now will raise a ValueError instead (if necessary) (GH4732)
  • Series.isin() and DataFrame.isin() now raise a TypeError when passed a string (GH4763). Pass a list of one element (containing the string) instead.
  • Remove undocumented/unused kind keyword argument from read_excel, and ExcelFile. (GH4713, GH4712)
  • The method argument of NDFrame.replace() is valid again, so that a a list can be passed to to_replace (GH4743).
  • provide automatic dtype conversions on _reduce operations (GH3371)
  • exclude non-numerics if mixed types with datelike in _reduce operations (GH3371)
  • default for tupleize_cols is now False for both to_csv and read_csv. Fair warning in 0.12 (GH3604)
  • moved timedeltas support to pandas.tseries.timedeltas.py; add timedeltas string parsing, add top-level to_timedelta function
  • NDFrame now is compatible with Python’s toplevel abs() function (GH4821).
  • raise a TypeError on invalid comparison ops on Series/DataFrame (e.g. integer/datetime) (GH4968)
  • 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. 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 raise a TypeError, e.g. Series(range(5))[3.5:4.5] (GH263,:issue:5375)
  • Make Categorical repr nicer (GH4368)
  • Remove deprecated Factor (GH3650)
  • Remove deprecated set_printoptions/reset_printoptions (:issue:3046)
  • Remove deprecated _verbose_info (GH3215)
  • Begin removing methods that don’t make sense on GroupBy objects (GH4887).
  • Remove deprecated read_clipboard/to_clipboard/ExcelFile/ExcelWriter from pandas.io.parsers (GH3717)
  • 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)
  • Arithmetic func factories are now passed real names (suitable for using with super) (GH5240)
  • Provide numpy compatibility with 1.7 for a calling convention like np.prod(pandas_object) as numpy call with additional keyword args (GH4435)
  • Provide __dir__ method (and local context) for tab completion / remove ipython completers code (GH4501)
  • Support non-unique axes in a Panel via indexing operations (GH4960)
  • .truncate will raise a ValueError if invalid before and afters dates are given (GH5242)
  • Timestamp now supports now/today/utcnow class methods (GH5339)
  • default for display.max_seq_len is now 100 rather then None. This activates truncated display (”...”) of long sequences in various places. (GH3391)
  • All division with NDFrame - likes is now truedivision, regardless of the future import. You can use // and floordiv to do 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]: pd.Series(arr) / pd.Series(arr2) # no future import required
Out[6]:
0    0.200000
1    0.666667
2    1.500000
3    4.000000
dtype: float64
  • raise/warn SettingWithCopyError/Warning exception/warning when setting of a copy thru chained assignment is detected, settable via option mode.chained_assignment
  • test the list of NA values in the csv parser. add N/A, #NA as independent default na values (GH5521)
  • The refactoring involving``Series`` deriving from NDFrame breaks rpy2<=2.3.8. an Issue has been opened against rpy2 and a workaround is detailed in GH5698. Thanks @JanSchulz.
  • 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)

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) See Internal Refactoring

  • 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)
    • isnull/notnull now available on NDFrame objects
  • 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 same API as original DataFrame filter
  • fillna refactored to core/generic.py, while > 3ndim is NotImplemented
  • Series now inherits from NDFrame rather than directly from ndarray. There are several minor changes that affect the API.
  • numpy functions that do not support the array interface will now return ndarrays rather than series, e.g. np.diff, np.ones_like, np.where
  • Series(0.5) would previously return the scalar 0.5, this is no longer supported
  • 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/DataFame, similar to dtypes, but indicates if the underlying is sparse/dense (as well as the dtype)
  • All NDFrame objects now have a _prop_attributes, which can be used to indicate various values to propagate to a new object from an existing (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, GH5216) 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
  • Series (for index) / Panel (for items) now as attribute access to its elements (GH1903)
  • Refactor clip methods to core/generic.py (GH4798)
  • Refactor of _get_numeric_data/_get_bool_data to core/generic.py, allowing Series/Panel functionality
  • Refactor of Series arithmetic with time-like objects (datetime/timedelta/time etc.) into a separate, cleaned up wrapper class. (GH4613)
  • Complex compat for Series with ndarray. (GH4819)
  • Removed unnecessary rwproperty from codebase in favor of builtin property. (GH4843)
  • Refactor object level numeric methods (mean/sum/min/max...) from object level modules to core/generic.py (GH4435).
  • Refactor cum objects to core/generic.py (GH4435), note that these have a more numpy-like function signature.
  • read_html() now uses TextParser to parse HTML data from bs4/lxml (GH4770).
  • Removed the keep_internal keyword parameter in pandas/core/groupby.py because it wasn’t being used (GH5102).
  • Base DateOffsets are no longer all instantiated on importing pandas, instead they are generated and cached on the fly. The internal representation and handling of DateOffsets has also been clarified. (GH5189, related GH5004)
  • MultiIndex constructor now validates that passed levels and labels are compatible. (GH5213, GH5214)
  • Unity dropna for Series/DataFrame signature (GH5250), tests from GH5234, courtesy of @rockg
  • Rewrite assert_almost_equal() in cython for performance (GH4398)
  • Added an internal _update_inplace method to facilitate updating NDFrame wrappers on inplace ops (only is for convenience of caller, doesn’t actually prevent copies). (GH5247)

Bug Fixes

  • HDFStore
    • raising an invalid TypeError rather than ValueError when appending with a different block ordering (GH4096)
    • read_hdf was not respecting as passed mode (GH4504)
    • appending a 0-len table will work correctly (GH4273)
    • to_hdf was raising when passing both arguments append and table (GH4584)
    • reading from a store with duplicate columns across dtypes would raise (GH4767)
    • Fixed a bug where ValueError wasn’t correctly raised when column names weren’t strings (GH4956)
    • A zero length series written in Fixed format not deserializing properly. (GH4708)
    • Fixed decoding perf issue on pyt3 (GH5441)
    • Validate levels in a multi-index before storing (GH5527)
    • Correctly handle data_columns with a Panel (GH5717)
  • Fixed bug in tslib.tz_convert(vals, tz1, tz2): it could raise IndexError exception while trying to access trans[pos + 1] (GH4496)
  • The by argument now works correctly with the layout argument (GH4102, GH4014) in *.hist plotting methods
  • Fixed bug in PeriodIndex.map where using str would return the str representation of the index (GH4136)
  • Fixed test failure test_time_series_plot_color_with_empty_kwargs when using custom matplotlib default colors (GH4345)
  • Fix running of stata IO tests. Now uses temporary files to write (GH4353)
  • Fixed an issue where DataFrame.sum was slower than DataFrame.mean for integer valued frames (GH4365)
  • read_html tests now work with Python 2.6 (GH4351)
  • Fixed bug where network testing was throwing NameError because a local variable was undefined (GH4381)
  • In to_json, raise if a passed orient would cause loss of data because of a duplicate index (GH4359)
  • In to_json, fix date handling so milliseconds are the default timestamp as the docstring says (GH4362).
  • as_index is no longer ignored when doing groupby apply (GH4648, GH3417)
  • JSON NaT handling fixed, NaTs are now serialized to null (GH4498)
  • Fixed JSON handling of escapable characters in JSON object keys (GH4593)
  • Fixed passing keep_default_na=False when na_values=None (GH4318)
  • Fixed bug with values raising an error on a DataFrame with duplicate columns and mixed dtypes, surfaced in (GH4377)
  • Fixed bug with duplicate columns and type conversion in read_json when orient='split' (GH4377)
  • Fixed JSON bug where locales with decimal separators other than ‘.’ threw exceptions when encoding / decoding certain values. (GH4918)
  • Fix .iat indexing with a PeriodIndex (GH4390)
  • Fixed an issue where PeriodIndex joining with self was returning a new instance rather than the same instance (GH4379); also adds a test for this for the other index types
  • Fixed a bug with all the dtypes being converted to object when using the CSV cparser with the usecols parameter (GH3192)
  • Fix an issue in merging blocks where the resulting DataFrame had partially set _ref_locs (GH4403)
  • Fixed an issue where hist subplots were being overwritten when they were called using the top level matplotlib API (GH4408)
  • Fixed a bug where calling Series.astype(str) would truncate the string (GH4405, GH4437)
  • Fixed a py3 compat issue where bytes were being repr’d as tuples (GH4455)
  • Fixed Panel attribute naming conflict if item is named ‘a’ (GH3440)
  • Fixed an issue where duplicate indexes were raising when plotting (GH4486)
  • Fixed an issue where cumsum and cumprod didn’t work with bool dtypes (GH4170, GH4440)
  • Fixed Panel slicing issued in xs that was returning an incorrect dimmed object (GH4016)
  • Fix resampling bug where custom reduce function not used if only one group (GH3849, GH4494)
  • Fixed Panel assignment with a transposed frame (GH3830)
  • Raise on set indexing with a Panel and a Panel as a value which needs alignment (GH3777)
  • frozenset objects now raise in the Series constructor (GH4482, GH4480)
  • Fixed issue with sorting a duplicate multi-index that has multiple dtypes (GH4516)
  • Fixed bug in DataFrame.set_values which was causing name attributes to be lost when expanding the index. (GH3742, GH4039)
  • Fixed issue where individual names, levels and labels could be set on MultiIndex without validation (GH3714, GH4039)
  • Fixed (GH3334) in pivot_table. Margins did not compute if values is the index.
  • Fix bug in having a rhs of np.timedelta64 or np.offsets.DateOffset when operating with datetimes (GH4532)
  • Fix arithmetic with series/datetimeindex and np.timedelta64 not working the same (GH4134) and buggy timedelta in numpy 1.6 (GH4135)
  • Fix bug in pd.read_clipboard on windows with PY3 (GH4561); not decoding properly
  • tslib.get_period_field() and tslib.get_period_field_arr() now raise if code argument out of range (GH4519, GH4520)
  • Fix boolean indexing on an empty series loses index names (GH4235), infer_dtype works with empty arrays.
  • Fix reindexing with multiple axes; if an axes match was not replacing the current axes, leading to a possible lazay frequency inference issue (GH3317)
  • Fixed issue where DataFrame.apply was reraising exceptions incorrectly (causing the original stack trace to be truncated).
  • Fix selection with ix/loc and non_unique selectors (GH4619)
  • Fix assignment with iloc/loc involving a dtype change in an existing column (GH4312, GH5702) have internal setitem_with_indexer in core/indexing to use Block.setitem
  • Fixed bug where thousands operator was not handled correctly for floating point numbers in csv_import (GH4322)
  • Fix an issue with CacheableOffset not properly being used by many DateOffset; this prevented the DateOffset from being cached (GH4609)
  • Fix boolean comparison with a DataFrame on the lhs, and a list/tuple on the rhs (GH4576)
  • Fix error/dtype conversion with setitem of None on Series/DataFrame (GH4667)
  • Fix decoding based on a passed in non-default encoding in pd.read_stata (GH4626)
  • Fix DataFrame.from_records with a plain-vanilla ndarray. (GH4727)
  • Fix some inconsistencies with Index.rename and MultiIndex.rename, etc. (GH4718, GH4628)
  • Bug in using iloc/loc with a cross-sectional and duplicate indicies (GH4726)
  • Bug with using QUOTE_NONE with to_csv causing Exception. (GH4328)
  • Bug with Series indexing not raising an error when the right-hand-side has an incorrect length (GH2702)
  • Bug in multi-indexing with a partial string selection as one part of a MultIndex (GH4758)
  • Bug with reindexing on the index with a non-unique index will now raise ValueError (GH4746)
  • Bug in setting with loc/ix a single indexer with a multi-index axis and a numpy array, related to (GH3777)
  • Bug in concatenation with duplicate columns across dtypes not merging with axis=0 (GH4771, GH4975)
  • Bug in iloc with a slice index failing (GH4771)
  • Incorrect error message with no colspecs or width in read_fwf. (GH4774)
  • Fix bugs in indexing in a Series with a duplicate index (GH4548, GH4550)
  • Fixed bug with reading compressed files with read_fwf in Python 3. (GH3963)
  • Fixed an issue with a duplicate index and assignment with a dtype change (GH4686)
  • Fixed bug with reading compressed files in as bytes rather than str in Python 3. Simplifies bytes-producing file-handling in Python 3 (GH3963, GH4785).
  • Fixed an issue related to ticklocs/ticklabels with log scale bar plots across different versions of matplotlib (GH4789)
  • Suppressed DeprecationWarning associated with internal calls issued by repr() (GH4391)
  • Fixed an issue with a duplicate index and duplicate selector with .loc (GH4825)
  • Fixed an issue with DataFrame.sort_index where, when sorting by a single column and passing a list for ascending, the argument for ascending was being interpreted as True (GH4839, GH4846)
  • Fixed Panel.tshift not working. Added freq support to Panel.shift (GH4853)
  • Fix an issue in TextFileReader w/ Python engine (i.e. PythonParser) with thousands != ”,” (GH4596)
  • Bug in getitem with a duplicate index when using where (GH4879)
  • Fix Type inference code coerces float column into datetime (GH4601)
  • Fixed _ensure_numeric does not check for complex numbers (GH4902)
  • Fixed a bug in Series.hist where two figures were being created when the by argument was passed (GH4112, GH4113).
  • Fixed a bug in convert_objects for > 2 ndims (GH4937)
  • Fixed a bug in DataFrame/Panel cache insertion and subsequent indexing (GH4939, GH5424)
  • Fixed string methods for FrozenNDArray and FrozenList (GH4929)
  • Fixed a bug with setting invalid or out-of-range values in indexing enlargement scenarios (GH4940)
  • Tests for fillna on empty Series (GH4346), thanks @immerrr
  • Fixed copy() to shallow copy axes/indices as well and thereby keep separate metadata. (GH4202, GH4830)
  • Fixed skiprows option in Python parser for read_csv (GH4382)
  • Fixed bug preventing cut from working with np.inf levels without explicitly passing labels (GH3415)
  • Fixed wrong check for overlapping in DatetimeIndex.union (GH4564)
  • Fixed conflict between thousands separator and date parser in csv_parser (GH4678)
  • Fix appending when dtypes are not the same (error showing mixing float/np.datetime64) (GH4993)
  • Fix repr for DateOffset. No longer show duplicate entries in kwds. Removed unused offset fields. (GH4638)
  • Fixed wrong index name during read_csv if using usecols. Applies to c parser only. (GH4201)
  • Timestamp objects can now appear in the left hand side of a comparison operation with a Series or DataFrame object (GH4982).
  • Fix a bug when indexing with np.nan via iloc/loc (GH5016)
  • Fixed a bug where low memory c parser could create different types in different chunks of the same file. Now coerces to numerical type or raises warning. (GH3866)
  • Fix a bug where reshaping a Series to its own shape raised TypeError (GH4554) and other reshaping issues.
  • Bug in setting with ix/loc and a mixed int/string index (GH4544)
  • Make sure series-series boolean comparisons are label based (GH4947)
  • Bug in multi-level indexing with a Timestamp partial indexer (GH4294)
  • Tests/fix for multi-index construction of an all-nan frame (GH4078)
  • Fixed a bug where read_html() wasn’t correctly inferring values of tables with commas (GH5029)
  • Fixed a bug where read_html() wasn’t providing a stable ordering of returned tables (GH4770, GH5029).
  • Fixed a bug where read_html() was incorrectly parsing when passed index_col=0 (GH5066).
  • Fixed a bug where read_html() was incorrectly inferring the type of headers (GH5048).
  • Fixed a bug where DatetimeIndex joins with PeriodIndex caused a stack overflow (GH3899).
  • Fixed a bug where groupby objects didn’t allow plots (GH5102).
  • Fixed a bug where groupby objects weren’t tab-completing column names (GH5102).
  • Fixed a bug where groupby.plot() and friends were duplicating figures multiple times (GH5102).
  • Provide automatic conversion of object dtypes on fillna, related (GH5103)
  • Fixed a bug where default options were being overwritten in the option parser cleaning (GH5121).
  • Treat a list/ndarray identically for iloc indexing with list-like (GH5006)
  • Fix MultiIndex.get_level_values() with missing values (GH5074)
  • Fix bound checking for Timestamp() with datetime64 input (GH4065)
  • Fix a bug where TestReadHtml wasn’t calling the correct read_html() function (GH5150).
  • Fix a bug with NDFrame.replace() which made replacement appear as though it was (incorrectly) using regular expressions (GH5143).
  • Fix better error message for to_datetime (GH4928)
  • Made sure different locales are tested on travis-ci (GH4918). Also adds a couple of utilities for getting locales and setting locales with a context manager.
  • Fixed segfault on isnull(MultiIndex) (now raises an error instead) (GH5123, GH5125)
  • Allow duplicate indices when performing operations that align (GH5185, GH5639)
  • Compound dtypes in a constructor raise NotImplementedError (GH5191)
  • Bug in comparing duplicate frames (GH4421) related
  • Bug in describe on duplicate frames
  • Bug in to_datetime with a format and coerce=True not raising (GH5195)
  • Bug in loc setting with multiple indexers and a rhs of a Series that needs broadcasting (GH5206)
  • Fixed bug where inplace setting of levels or labels on MultiIndex would not clear cached values property and therefore return wrong values. (GH5215)
  • Fixed bug where filtering a grouped DataFrame or Series did not maintain the original ordering (GH4621).
  • Fixed Period with a business date freq to always roll-forward if on a non-business date. (GH5203)
  • Fixed bug in Excel writers where frames with duplicate column names weren’t written correctly. (GH5235)
  • Fixed issue with drop and a non-unique index on Series (GH5248)
  • Fixed seg fault in C parser caused by passing more names than columns in the file. (GH5156)
  • Fix Series.isin with date/time-like dtypes (GH5021)
  • C and Python Parser can now handle the more common multi-index column format which doesn’t have a row for index names (GH4702)
  • Bug when trying to use an out-of-bounds date as an object dtype (GH5312)
  • Bug when trying to display an embedded PandasObject (GH5324)
  • Allows operating of Timestamps to return a datetime if the result is out-of-bounds related (GH5312)
  • Fix return value/type signature of initObjToJSON() to be compatible with numpy’s import_array() (GH5334, GH5326)
  • Bug when renaming then set_index on a DataFrame (GH5344)
  • Test suite no longer leaves around temporary files when testing graphics. (GH5347) (thanks for catching this @yarikoptic!)
  • Fixed html tests on win32. (GH4580)
  • Make sure that head/tail are iloc based, (GH5370)
  • Fixed bug for PeriodIndex string representation if there are 1 or 2 elements. (GH5372)
  • The GroupBy methods transform and filter can be used on Series and DataFrames that have repeated (non-unique) indices. (GH4620)
  • Fix empty series not printing name in repr (GH4651)
  • Make tests create temp files in temp directory by default. (GH5419)
  • pd.to_timedelta of a scalar returns a scalar (GH5410)
  • pd.to_timedelta accepts NaN and NaT, returning NaT instead of raising (GH5437)
  • performance improvements in isnull on larger size pandas objects
  • Fixed various setitem with 1d ndarray that does not have a matching length to the indexer (GH5508)
  • Bug in getitem with a multi-index and iloc (GH5528)
  • Bug in delitem on a Series (GH5542)
  • Bug fix in apply when using custom function and objects are not mutated (GH5545)
  • Bug in selecting from a non-unique index with loc (GH5553)
  • Bug in groupby returning non-consistent types when user function returns a None, (GH5592)
  • Work around regression in numpy 1.7.0 which erroneously raises IndexError from ndarray.item (GH5666)
  • Bug in repeated indexing of object with resultant non-unique index (GH5678)
  • Bug in fillna with Series and a passed series/dict (GH5703)
  • Bug in groupby transform with a datetime-like grouper (GH5712)
  • Bug in multi-index selection in PY3 when using certain keys (GH5725)
  • Row-wise concat of differing dtypes failing in certain cases (GH5754)

pandas 0.12.0

Release date: 2013-07-24

New Features

  • pd.read_html() can now parse HTML strings, files or urls and returns a list of DataFrame s courtesy of @cpcloud. (GH3477, GH3605, GH3606)
  • Support for reading Amazon S3 files. (GH3504)
  • Added module for reading and writing JSON strings/files: pandas.io.json includes to_json DataFrame/Series method, and a read_json top-level reader various issues (GH1226, GH3804, GH3876, GH3867, GH1305)
  • Added module for reading and writing Stata files: pandas.io.stata (GH1512) includes to_stata DataFrame method, and a read_stata top-level reader
  • Added support for writing in to_csv and reading in read_csv, multi-index columns. The header option in read_csv now accepts a list of the rows from which to read the index. Added the option, tupleize_cols to provide compatibility for the pre 0.12 behavior of writing and reading multi-index 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 multi-index column. Note: The default value will change in 0.12 to make the default to write and read multi-index columns in the new format. (GH3571, GH1651, GH3141)
  • Add iterator to Series.str (GH3638)
  • pd.set_option() now allows N option, value pairs (GH3667).
  • Added keyword parameters for different types of scatter_matrix subplots
  • A filter method on grouped Series or DataFrames returns a subset of the original (GH3680, GH919)
  • Access to historical Google Finance data in pandas.io.data (GH3814)
  • DataFrame plotting methods can sample column colors from a Matplotlib colormap via the colormap keyword. (GH3860)

Improvements to existing features

  • Fixed various issues with internal pprinting code, the repr() for various objects including TimeStamp and Index now produces valid python code strings and can be used to recreate the object, (GH3038, GH3379, GH3251, GH3460)
  • convert_objects now accepts a copy parameter (defaults to True)
  • HDFStore
    • will retain index attributes (freq,tz,name) on recreation (GH3499,:issue:4098)
    • 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)
    • table writing performance improvements.
    • support python3 (via PyTables 3.0.0) (GH3750)
  • Add modulo operator to Series, DataFrame
  • Add date method to DatetimeIndex
  • Add dropna argument to pivot_table (:issue: 3820)
  • Simplified the API and added a describe method to Categorical
  • melt now accepts the optional parameters var_name and value_name to specify custom column names of the returned DataFrame (GH3649), thanks @hoechenberger. If var_name is not specified and dataframe.columns.name is not None, then this will be used as the var_name (GH4144). Also support for MultiIndex columns.
  • clipboard functions use pyperclip (no dependencies on Windows, alternative dependencies offered for Linux) (GH3837).
  • 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 eliminates any spurious plots from showing up.
  • Added Faq section on repr display options, to help users customize their setup.
  • where operations that result in block splitting are much faster (GH3733)
  • Series and DataFrame hist methods now take a figsize argument (GH3834)
  • DatetimeIndexes no longer try to convert mixed-integer indexes during join operations (GH3877)
  • Add unit keyword to Timestamp and to_datetime to enable passing of integers or floats that are in an epoch unit of D, s, ms, us, ns, thanks @mtkini (GH3969) (e.g. unix timestamps or epoch s, with fractional seconds allowed) (GH3540)
  • DataFrame corr method (spearman) is now cythonized.
  • 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)
  • read_csv will now throw a more informative error message when a file contains no columns, e.g., all newline characters
  • Added layout keyword to DataFrame.hist() for more customizable layout (GH4050)
  • 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)

API Changes

  • HDFStore
    • When removing an object, remove(key) raises KeyError if the key is not a valid store object.
    • raise a TypeError on passing where or columns to select with a Storer; these are invalid parameters at this time (GH4189)
    • can now specify an encoding option to append/put to enable alternate encodings (GH3750)
    • enable support for iterator/chunksize with read_hdf
  • The repr() for (Multi)Index now obeys display.max_seq_items rather then numpy threshold print options. (GH3426, GH3466)
  • Added mangle_dupe_cols option to read_table/csv, allowing users to control legacy behaviour re dupe cols (A, A.1, A.2 vs A, A ) (GH3468) Note: The default value will change in 0.12 to the “no mangle” behaviour, If your code relies on this behaviour, explicitly specify mangle_dupe_cols=True in your calls.
  • 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
  • Do not allow datetimelike/timedeltalike creation except with valid types (e.g. cannot pass datetime64[ms]) (GH3423)
  • Add squeeze keyword to groupby to allow reduction from DataFrame -> Series if groups are unique. Regression from 0.10.1, partial revert on (GH2893) with (GH3596)
  • 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)
  • The raise_on_error option to plotting methods is obviated by GH3572, so it is removed. Plots now always raise when data cannot be plotted or the object being plotted has a dtype of object.
  • 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)
  • Deprecated display.height, display.width is now only a formatting option does not control triggering of summary, similar to < 0.11.0.
  • 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)
  • io API changes
    • added pandas.io.api for i/o imports
    • removed Excel support to pandas.io.excel
    • added top-level pd.read_sql and to_sql DataFrame methods
    • removed clipboard support to pandas.io.clipboard
    • replace top-level and instance methods save and load with top-level read_pickle and to_pickle instance method, save and load will give deprecation warning.
  • the method and axis arguments of DataFrame.replace() are deprecated
  • set FutureWarning to require data_source, and to replace year/month with expiry date in pandas.io options. This is in preparation to add options data from Google (GH3822)
  • the method and axis arguments of DataFrame.replace() are deprecated
  • Implement __nonzero__ for NDFrame objects (GH3691, GH3696)
  • as_matrix with mixed signed and unsigned dtypes will result in 2 x the lcd of the unsigned as an int, maxing with int64, to avoid precision issues (GH3733)
  • na_values in a list provided to read_csv/read_excel will match string and numeric versions e.g. na_values=['99'] will match 99 whether the column ends up being int, float, or string (GH3611)
  • 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
  • more consistency in the to_datetime return types (give string/array of string inputs) (GH3888)
  • 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)

Experimental Features

  • Added experimental CustomBusinessDay class to support DateOffsets with custom holiday calendars and custom weekmasks. (GH2301)

Bug Fixes

  • Fixed an esoteric excel reading bug, xlrd>= 0.9.0 now required for excel support. Should provide python3 support (for reading) which has been lacking. (GH3164)
  • Disallow Series constructor called with MultiIndex which caused segfault (GH4187)
  • Allow unioning of date ranges sharing a timezone (GH3491)
  • Fix to_csv issue when having a large number of rows and NaT in some columns (GH3437)
  • .loc was not raising when passed an integer list (GH3449)
  • Unordered time series selection was misbehaving when using label slicing (GH3448)
  • Fix sorting in a frame with a list of columns which contains datetime64[ns] dtypes (GH3461)
  • DataFrames fetched via FRED now handle ‘.’ as a NaN. (GH3469)
  • Fix regression in a DataFrame apply with axis=1, objects were not being converted back to base dtypes correctly (GH3480)
  • Fix issue when storing uint dtypes in an HDFStore. (GH3493)
  • Non-unique index support clarified (GH3468)
    • Addressed handling of dupe columns in df.to_csv new and old (GH3454, GH3457)
    • 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)
    • 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)
  • Fixed bug in groupby with empty series referencing a variable before assignment. (GH3510)
  • Allow index name to be used in groupby for non MultiIndex (GH4014)
  • Fixed bug in mixed-frame assignment with aligned series (GH3492)
  • Fixed bug in selecting month/quarter/year from a series would not select the time element on the last day (GH3546)
  • Fixed a couple of MultiIndex rendering bugs in df.to_html() (GH3547, GH3553)
  • Properly convert np.datetime64 objects in a Series (GH3416)
  • Raise a TypeError on invalid datetime/timedelta operations e.g. add datetimes, multiple timedelta x datetime
  • Fix .diff on datelike and timedelta operations (GH3100)
  • combine_first not returning the same dtype in cases where it can (GH3552)
  • Fixed bug with Panel.transpose argument aliases (GH3556)
  • Fixed platform bug in PeriodIndex.take (GH3579)
  • Fixed bud in incorrect conversion of datetime64[ns] in combine_first (GH3593)
  • Fixed bug in reset_index with NaN in a multi-index (GH3586)
  • fillna methods now raise a TypeError when the value parameter is a list or tuple.
  • Fixed bug where a time-series was being selected in preference to an actual column name in a frame (GH3594)
  • Make secondary_y work properly for bar plots (GH3598)
  • Fix modulo and integer division on Series,DataFrames to act similary to float dtypes to return np.nan or np.inf as appropriate (GH3590)
  • Fix incorrect dtype on groupby with as_index=False (GH3610)
  • Fix read_csv/read_excel to correctly encode identical na_values, e.g. na_values=[-999.0,-999] was failing (GH3611)
  • Disable HTML output in qtconsole again. (GH3657)
  • Reworked the new repr display logic, which users found confusing. (GH3663)
  • Fix indexing issue in ndim >= 3 with iloc (GH3617)
  • Correctly parse date columns with embedded (nan/NaT) into datetime64[ns] dtype in read_csv when parse_dates is specified (GH3062)
  • Fix not consolidating before to_csv (GH3624)
  • Fix alignment issue when setitem in a DataFrame with a piece of a DataFrame (GH3626) or a mixed DataFrame and a Series (GH3668)
  • Fix plotting of unordered DatetimeIndex (GH3601)
  • sql.write_frame failing when writing a single column to sqlite (GH3628), thanks to @stonebig
  • Fix pivoting with nan in the index (GH3558)
  • Fix running of bs4 tests when it is not installed (GH3605)
  • Fix parsing of html table (GH3606)
  • read_html() now only allows a single backend: html5lib (GH3616)
  • convert_objects with convert_dates='coerce' was parsing some single-letter strings into today’s date
  • DataFrame.from_records did not accept empty recarrays (GH3682)
  • DataFrame.to_csv will succeed with the deprecated option nanRep, @tdsmith
  • DataFrame.to_html and DataFrame.to_latex now accept a path for their first argument (GH3702)
  • Fix file tokenization error with r delimiter and quoted fields (GH3453)
  • Groupby transform with item-by-item not upcasting correctly (GH3740)
  • Incorrectly read a HDFStore multi-index Frame with a column specification (GH3748)
  • read_html now correctly skips tests (GH3741)
  • PandasObjects raise TypeError when trying to hash (GH3882)
  • Fix incorrect arguments passed to concat that are not list-like (e.g. concat(df1,df2)) (GH3481)
  • Correctly parse when passed the dtype=str (or other variable-len string dtypes) in read_csv (GH3795)
  • Fix index name not propagating when using loc/ix (GH3880)
  • Fix groupby when applying a custom function resulting in a returned DataFrame was not converting dtypes (GH3911)
  • Fixed a bug where DataFrame.replace with a compiled regular expression in the to_replace argument wasn’t working (GH3907)
  • Fixed __truediv__ in Python 2.7 with numexpr installed to actually do true division when dividing two integer arrays with at least 10000 cells total (GH3764)
  • Indexing with a string with seconds resolution not selecting from a time index (GH3925)
  • csv parsers would loop infinitely if iterator=True but no chunksize was specified (GH3967), python parser failing with chunksize=1
  • Fix index name not propagating when using shift
  • Fixed dropna=False being ignored with multi-index stack (GH3997)
  • Fixed flattening of columns when renaming MultiIndex columns DataFrame (GH4004)
  • Fix Series.clip for datetime series. NA/NaN threshold values will now throw ValueError (GH3996)
  • Fixed insertion issue into DataFrame, after rename (GH4032)
  • 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)
  • Fix bug where HDFStore will fail to append because of a different block ordering on-disk (GH4096)
  • Better error messages on inserting incompatible columns to a frame (GH4107)
  • Fixed bug in DataFrame.replace where a nested dict wasn’t being iterated over when regex=False (GH4115)
  • Fixed bug in convert_objects(convert_numeric=True) where a mixed numeric and object Series/Frame was not converting properly (GH4119)
  • Fixed bugs in multi-index selection with column multi-index and duplicates (GH4145, GH4146)
  • 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 Series.where where broadcasting a single element input vector to the length of the series resulted in multiplying the value inside the input (GH4192)
  • 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)

pandas 0.11.0

Release date: 2013-04-22

New Features

  • New documentation section, 10 Minutes to Pandas
  • New documentation section, Cookbook
  • Allow mixed dtypes (e.g float32/float64/int32/int16/int8) to coexist in DataFrames and propagate in operations
  • Add function to pandas.io.data for retrieving stock index components from Yahoo! finance (GH2795)
  • Support slicing with time objects (GH2681)
  • Added .iloc attribute, to support strict integer based indexing, analogous to .ix (GH2922)
  • Added .loc attribute, to support strict label based indexing, analogous to .ix (GH3053)
  • Added .iat attribute, to support fast scalar access via integers (replaces iget_value/iset_value)
  • Added .at attribute, to support fast scalar access via labels (replaces get_value/set_value)
  • Moved functionality from irow,icol,iget_value/iset_value to .iloc indexer (via _ixs methods in each object)
  • Added support for expression evaluation using the numexpr library
  • Added convert=boolean to take routines to translate negative indices to positive, defaults to True
  • Added to_series() method to indices, to facilitate the creation of indexers (GH3275)

Improvements to existing features

  • Improved performance of df.to_csv() by up to 10x in some cases. (GH3059)

  • added blocks attribute to DataFrames, to return a dict of dtypes to homogeneously dtyped DataFrames

  • added keyword convert_numeric to convert_objects() to try to convert object dtypes to numeric types (default is False)

  • convert_dates in convert_objects can now be coerce which will return a datetime64[ns] dtype with non-convertibles set as NaT; will preserve an all-nan object (e.g. strings), default is True (to perform soft-conversion

  • Series print output now includes the dtype by default

  • Optimize internal reindexing routines (GH2819, GH2867)

  • describe_option() now reports the default and current value of options.

  • Add format option to pandas.to_datetime with faster conversion of strings that can be parsed with datetime.strptime

  • Add axes property to Series for compatibility

  • Add xs function to Series for compatibility

  • Allow setitem in a frame where only mixed numerics are present (e.g. int and float), (GH3037)

  • HDFStore

    • 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)
    • support read_hdf/to_hdf API similar to read_csv/to_csv (GH3222)
  • Add squeeze method to possibly remove length 1 dimensions from an object.

    In [1]: p = Panel(randn(3,4,4),items=['ItemA','ItemB','ItemC'],
       ...:                    major_axis=date_range('20010102',periods=4),
       ...:                    minor_axis=['A','B','C','D'])
       ...: 
    
    In [2]: p
    Out[2]: 
    <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 [3]: p.reindex(items=['ItemA']).squeeze()
    Out[3]: 
                       A         B         C         D
    2001-01-02  0.469112 -0.282863 -1.509059 -1.135632
    2001-01-03  1.212112 -0.173215  0.119209 -1.044236
    2001-01-04 -0.861849 -2.104569 -0.494929  1.071804
    2001-01-05  0.721555 -0.706771 -1.039575  0.271860
    
    In [4]: p.reindex(items=['ItemA'],minor=['B']).squeeze()
    Out[4]: 
    2001-01-02   -0.282863
    2001-01-03   -0.173215
    2001-01-04   -2.104569
    2001-01-05   -0.706771
    Freq: D, Name: B, dtype: float64
    
  • Improvement to Yahoo API access in pd.io.data.Options (GH2758)

  • 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.

  • Allow selection semantics via a string with a datelike index to work in both Series and DataFrames (GH3070)

    In [5]: idx = date_range("2001-10-1", periods=5, freq='M')
    
    In [6]: ts = Series(np.random.rand(len(idx)),index=idx)
    
    In [7]: ts['2001']
    Out[7]: 
    2001-10-31    0.838796
    2001-11-30    0.897333
    2001-12-31    0.732592
    Freq: M, dtype: float64
    
    In [8]: df = DataFrame(dict(A = ts))
    
    In [9]: df['2001']
    Out[9]: 
                       A
    2001-10-31  0.838796
    2001-11-30  0.897333
    2001-12-31  0.732592
    
  • added option display.mpl_style providing a sleeker visual style for plots. Based on https://gist.github.com/huyng/816622 (GH3075).

  • Improved performance across several core functions by taking memory ordering of arrays into account. Courtesy of @stephenwlin (GH3130)

  • Improved performance of groupby transform method (GH2121)

  • Handle “ragged” CSV files missing trailing delimiters in rows with missing fields when also providing explicit list of column names (so the parser knows how many columns to expect in the result) (GH2981)

  • On a mixed DataFrame, allow setting with indexers with ndarray/DataFrame on rhs (GH3216)

  • Treat boolean values as integers (values 1 and 0) for numeric operations. (GH2641)

  • Add time method to DatetimeIndex (GH3180)

  • Return NA when using Series.str[...] for values that are not long enough (GH3223)

  • Display cursor coordinate information in time-series plots (GH1670)

  • to_html() now accepts an optional “escape” argument to control reserved HTML character escaping (enabled by default) and escapes &, in addition to < and >. (GH2919)

API Changes

  • Do not automatically upcast numeric specified dtypes to int64 or float64 (GH622 and GH797)

  • DataFrame construction of lists and scalars, with no dtype present, will result in casting to int64 or float64, regardless of platform. This is not an apparent change in the API, but noting it.

  • Guarantee that convert_objects() for Series/DataFrame always returns a copy

  • groupby operations will respect dtypes for numeric float operations (float32/float64); other types will be operated on, and will try to cast back to the input dtype (e.g. if an int is passed, as long as the output doesn’t have nans, then an int will be returned)

  • backfill/pad/take/diff/ohlc will now support float32/int16/int8 operations

  • Block types will upcast as needed in where/masking operations (GH2793)

  • Series now automatically will try to set the correct dtype based on passed datetimelike objects (datetime/Timestamp)

    • timedelta64 are returned in appropriate cases (e.g. Series - Series, when both are datetime64)
    • mixed datetimes and objects (GH2751) in a constructor will be cast correctly
    • astype on datetimes to object are now handled (as well as NaT conversions to np.nan)
    • all timedelta like objects will be correctly assigned to timedelta64 with mixed NaN and/or NaT allowed
  • arguments to DataFrame.clip were inconsistent to numpy and Series clipping (GH2747)

  • util.testing.assert_frame_equal now checks the column and index names (GH2964)

  • Constructors will now return a more informative ValueError on failures when invalid shapes are passed

  • Don’t suppress TypeError in GroupBy.agg (GH3238)

  • Methods return None when inplace=True (GH1893)

  • 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 will now automatically create data_columns for passed keys
  • Downcast on pivot if possible (GH3283), adds argument downcast to fillna

  • Introduced options display.height/width for explicitly specifying terminal height/width in characters. Deprecated display.line_width, now replaced by display.width. These defaults are in effect for scripts as well, so unless disabled, previously very wide output will now be output as “expand_repr” style wrapped output.

  • Various defaults for options (including display.max_rows) have been revised, after a brief survey concluded they were wrong for everyone. Now at w=80,h=60.

  • HTML repr output in IPython qtconsole is once again controlled by the option display.notebook_repr_html, and on by default.

Bug Fixes

  • Fix seg fault on empty data frame when fillna with pad or backfill (GH2778)
  • Single element ndarrays of datetimelike objects are handled (e.g. np.array(datetime(2001,1,1,0,0))), w/o dtype being passed
  • 0-dim ndarrays with a passed dtype are handled correctly (e.g. np.array(0.,dtype=’float32’))
  • Fix some boolean indexing inconsistencies in Series.__getitem__/__setitem__ (GH2776)
  • Fix issues with DataFrame and Series constructor with integers that overflow int64 and some mixed typed type lists (GH2845)
  • HDFStore
    • Fix weird PyTables error when using too many selectors in a where also correctly filter on any number of values in a Term expression (so not using numexpr filtering, but isin filtering)
    • Internally, change all variables to be private-like (now have leading underscore)
    • Fixes for query parsing to correctly interpret boolean and != (GH2849, GH2973)
    • Fixes for pathological case on SparseSeries with 0-len array and compression (GH2931)
    • Fixes bug with writing rows if part of a block was all-nan (GH3012)
    • Exceptions are now ValueError or TypeError as needed
    • A table will now raise if min_itemsize contains fields which are not queryables
  • Bug showing up in applymap where some object type columns are converted (GH2909) had an incorrect default in convert_objects
  • TimeDeltas
    • Series ops with a Timestamp on the rhs was throwing an exception (GH2898) added tests for Series ops with datetimes,timedeltas,Timestamps, and datelike Series on both lhs and rhs
    • Fixed subtle timedelta64 inference issue on py3 & numpy 1.7.0 (GH3094)
    • Fixed some formatting issues on timedelta when negative
    • Support null checking on timedelta64, representing (and formatting) with NaT
    • Support setitem with np.nan value, converts to NaT
    • Support min/max ops in a Dataframe (abs not working, nor do we error on non-supported ops)
    • Support idxmin/idxmax/abs/max/min in a Series (GH2989, GH2982)
  • Bug on in-place putmasking on an integer series that needs to be converted to float (GH2746)
  • Bug in argsort of datetime64[ns] Series with NaT (GH2967)
  • Bug in value_counts of datetime64[ns] Series (GH3002)
  • Fixed printing of NaT in an index
  • Bug in idxmin/idxmax of datetime64[ns] Series with NaT (GH2982)
  • Bug in icol, take with negative indicies was producing incorrect return values (see GH2922, GH2892), also check for out-of-bounds indices (GH3029)
  • Bug in DataFrame column insertion when the column creation fails, existing frame is left in an irrecoverable state (GH3010)
  • Bug in DataFrame update, combine_first where non-specified values could cause dtype changes (GH3016, GH3041)
  • Bug in groupby with first/last where dtypes could change (GH3041, GH2763)
  • Formatting of an index that has nan was inconsistent or wrong (would fill from other values), (GH2850)
  • Unstack of a frame with no nans would always cause dtype upcasting (GH2929)
  • Fix scalar datetime.datetime parsing bug in read_csv (GH3071)
  • Fixed slow printing of large Dataframes, due to inefficient dtype reporting (GH2807)
  • Fixed a segfault when using a function as grouper in groupby (GH3035)
  • Fix pretty-printing of infinite data structures (closes GH2978)
  • Fixed exception when plotting timeseries bearing a timezone (closes GH2877)
  • str.contains ignored na argument (GH2806)
  • Substitute warning for segfault when grouping with categorical grouper of mismatched length (GH3011)
  • Fix exception in SparseSeries.density (GH2083)
  • Fix upsampling bug with closed=’left’ and daily to daily data (GH3020)
  • Fixed missing tick bars on scatter_matrix plot (GH3063)
  • Fixed bug in Timestamp(d,tz=foo) when d is date() rather then datetime() (GH2993)
  • series.plot(kind=’bar’) now respects pylab color schem (GH3115)
  • Fixed bug in reshape if not passed correct input, now raises TypeError (GH2719)
  • Fixed a bug where Series ctor did not respect ordering if OrderedDict passed in (GH3282)
  • Fix NameError issue on RESO_US (GH2787)
  • Allow selection in an unordered timeseries to work similary to an ordered timeseries (GH2437).
  • Fix implemented .xs when called with axes=1 and a level parameter (GH2903)
  • Timestamp now supports the class method fromordinal similar to datetimes (GH3042)
  • Fix issue with indexing a series with a boolean key and specifiying a 1-len list on the rhs (GH2745) or a list on the rhs (GH3235)
  • Fixed bug in groupby apply when kernel generate list of arrays having unequal len (GH1738)
  • fixed handling of rolling_corr with center=True which could produce corr>1 (GH3155)
  • Fixed issues where indices can be passed as ‘index/column’ in addition to 0/1 for the axis parameter
  • PeriodIndex.tolist now boxes to Period (GH3178)
  • PeriodIndex.get_loc KeyError now reports Period instead of ordinal (GH3179)
  • df.to_records bug when handling MultiIndex (GH3189)
  • Fix Series.__getitem__ segfault when index less than -length (GH3168)
  • Fix bug when using Timestamp as a date parser (GH2932)
  • Fix bug creating date range from Timestamp with time zone and passing same time zone (GH2926)
  • Add comparison operators to Period object (GH2781)
  • Fix bug when concatenating two Series into a DataFrame when they have the same name (GH2797)
  • Fix automatic color cycling when plotting consecutive timeseries without color arguments (GH2816)
  • fixed bug in the pickling of PeriodIndex (GH2891)
  • Upcast/split blocks when needed in a mixed DataFrame when setitem with an indexer (GH3216)
  • Invoking df.applymap on a dataframe with dupe cols now raises a ValueError (GH2786)
  • Apply with invalid returned indices raise correct Exception (GH2808)
  • Fixed a bug in plotting log-scale bar plots (GH3247)
  • df.plot() grid on/off now obeys the mpl default style, just like series.plot(). (GH3233)
  • Fixed a bug in the legend of plotting.andrews_curves() (GH3278)
  • Produce a series on apply if we only generate a singular series and have a simple index (GH2893)
  • Fix Python ASCII file parsing when integer falls outside of floating point spacing (GH3258)
  • fixed pretty priniting of sets (GH3294)
  • Panel() and Panel.from_dict() now respects ordering when give OrderedDict (GH3303)
  • DataFrame where with a datetimelike incorrectly selecting (GH3311)
  • Ensure index casts work even in Int64Index
  • Fix set_index segfault when passing MultiIndex (GH3308)
  • Ensure pickles created in py2 can be read in py3
  • Insert ellipsis in MultiIndex summary repr (GH3348)
  • Groupby will handle mutation among an input groups columns (and fallback to non-fast apply) (GH3380)
  • Eliminated unicode errors on FreeBSD when using MPL GTK backend (GH3360)
  • Period.strftime should return unicode strings always (GH3363)
  • Respect passed read_* chunksize in get_chunk function (GH3406)

pandas 0.10.1

Release date: 2013-01-22

New Features

  • Add data interface to World Bank WDI pandas.io.wb (GH2592)

API Changes

  • Restored inplace=True behavior returning self (same object) with deprecation warning until 0.11 (GH1893)
  • HDFStore
    • refactored HFDStore to deal with non-table stores as objects, will allow future enhancements
    • removed keyword compression from put (replaced by keyword complib to be consistent across library)
    • warn PerformanceWarning if you are attempting to store types that will be pickled by PyTables

Improvements to existing features

  • HDFStore
    • enables storing of multi-index dataframes (closes GH1277)
    • support data column indexing and selection, via data_columns keyword in append
    • support write chunking to reduce memory footprint, via chunksize keyword to append
    • support automagic indexing via index keyword to append
    • support expectedrows keyword in append to inform PyTables about the expected tablesize
    • support start and stop keywords in select to limit the row selection space
    • added get_store context manager to automatically import with pandas
    • added column filtering via columns keyword in select
    • added methods append_to_multiple/select_as_multiple/select_as_coordinates to do multiple-table append/selection
    • added support for datetime64 in columns
    • added method unique to select the unique values in an indexable or data column
    • added method copy to copy an existing store (and possibly upgrade)
    • show the shape of the data on disk for non-table stores when printing the store
    • added ability to read PyTables flavor tables (allows compatibility to other HDF5 systems)
  • Add logx option to DataFrame/Series.plot (GH2327, GH2565)
  • Support reading gzipped data from file-like object
  • pivot_table aggfunc can be anything used in GroupBy.aggregate (GH2643)
  • Implement DataFrame merges in case where set cardinalities might overflow 64-bit integer (GH2690)
  • Raise exception in C file parser if integer dtype specified and have NA values. (GH2631)
  • Attempt to parse ISO8601 format dates when parse_dates=True in read_csv for major performance boost in such cases (GH2698)
  • Add methods neg and inv to Series
  • Implement kind option in ExcelFile to indicate whether it’s an XLS or XLSX file (GH2613)
  • Documented a fast-path in pd.read_csv when parsing iso8601 datetime strings yielding as much as a 20x speedup. (GH5993)

Bug Fixes

  • Fix read_csv/read_table multithreading issues (GH2608)
  • HDFStore
    • correctly handle nan elements in string columns; serialize via the nan_rep keyword to append
    • raise correctly on non-implemented column types (unicode/date)
    • handle correctly Term passed types (e.g. index<1000, when index is Int64), (closes GH512)
    • handle Timestamp correctly in data_columns (closes GH2637)
    • contains correctly matches on non-natural names
    • correctly store float32 dtypes in tables (if not other float types in the same table)
  • Fix DataFrame.info bug with UTF8-encoded columns. (GH2576)
  • Fix DatetimeIndex handling of FixedOffset tz (GH2604)
  • More robust detection of being in IPython session for wide DataFrame console formatting (GH2585)
  • Fix platform issues with file:/// in unit test (GH2564)
  • Fix bug and possible segfault when grouping by hierarchical level that contains NA values (GH2616)
  • Ensure that MultiIndex tuples can be constructed with NAs (GH2616)
  • Fix int64 overflow issue when unstacking MultiIndex with many levels (GH2616)
  • Exclude non-numeric data from DataFrame.quantile by default (GH2625)
  • Fix a Cython C int64 boxing issue causing read_csv to return incorrect results (GH2599)
  • Fix groupby summing performance issue on boolean data (GH2692)
  • Don’t bork Series containing datetime64 values with to_datetime (GH2699)
  • Fix DataFrame.from_records corner case when passed columns, index column, but empty record list (GH2633)
  • Fix C parser-tokenizer bug with trailing fields. (GH2668)
  • Don’t exclude non-numeric data from GroupBy.max/min (GH2700)
  • Don’t lose time zone when calling DatetimeIndex.drop (GH2621)
  • Fix setitem on a Series with a boolean key and a non-scalar as value (GH2686)
  • Box datetime64 values in Series.apply/map (GH2627, GH2689)
  • Upconvert datetime + datetime64 values when concatenating frames (GH2624)
  • Raise a more helpful error message in merge operations when one DataFrame has duplicate columns (GH2649)
  • Fix partial date parsing issue occuring only when code is run at EOM (GH2618)
  • Prevent MemoryError when using counting sort in sortlevel with high-cardinality MultiIndex objects (GH2684)
  • Fix Period resampling bug when all values fall into a single bin (GH2070)
  • Fix buggy interaction with usecols argument in read_csv when there is an implicit first index column (GH2654)
  • Fix bug in Index.summary() where string format methods were being called incorrectly. (GH3869)

pandas 0.10.0

Release date: 2012-12-17

New Features

  • Brand new high-performance delimited file parsing engine written in C and Cython. 50% or better performance in many standard use cases with a fraction as much memory usage. (GH407, GH821)
  • Many new file parser (read_csv, read_table) features:
    • Support for on-the-fly gzip or bz2 decompression (compression option)
    • Ability to get back numpy.recarray instead of DataFrame (as_recarray=True)
    • dtype option: explicit column dtypes
    • usecols option: specify list of columns to be read from a file. Good for reading very wide files with many irrelevant columns (GH1216 GH926, GH2465)
    • Enhanced unicode decoding support via encoding option
    • skipinitialspace dialect option
    • Can specify strings to be recognized as True (true_values) or False (false_values)
    • High-performance delim_whitespace option for whitespace-delimited files; a preferred alternative to the ‘s+’ regular expression delimiter
    • Option to skip “bad” lines (wrong number of fields) that would otherwise have caused an error in the past (error_bad_lines and warn_bad_lines options)
    • Substantially improved performance in the parsing of integers with thousands markers and lines with comments
    • Easy of European (and other) decimal formats (decimal option) (GH584, GH2466)
    • Custom line terminators (e.g. lineterminator=’~’) (GH2457)
    • Handling of no trailing commas in CSV files (GH2333)
    • Ability to handle fractional seconds in date_converters (GH2209)
    • read_csv allow scalar arg to na_values (GH1944)
    • Explicit column dtype specification in read_* functions (GH1858)
    • Easier CSV dialect specification (GH1743)
    • Improve parser performance when handling special characters (GH1204)
  • Google Analytics API integration with easy oauth2 workflow (GH2283)
  • Add error handling to Series.str.encode/decode (GH2276)
  • Add where and mask to Series (GH2337)
  • Grouped histogram via by keyword in Series/DataFrame.hist (GH2186)
  • Support optional min_periods keyword in corr and cov for both Series and DataFrame (GH2002)
  • Add duplicated and drop_duplicates functions to Series (GH1923)
  • Add docs for HDFStore table format
  • ‘density’ property in SparseSeries (GH2384)
  • Add ffill and bfill convenience functions for forward- and backfilling time series data (GH2284)
  • New option configuration system and functions set_option, get_option, describe_option, and reset_option. Deprecate set_printoptions and reset_printoptions (GH2393). You can also access options as attributes via pandas.options.X
  • Wide DataFrames can be viewed more easily in the console with new expand_frame_repr and line_width configuration options. This is on by default now (GH2436)
  • Scikits.timeseries-like moving window functions via rolling_window (GH1270)

Experimental Features

  • Add support for Panel4D, a named 4 Dimensional structure
  • Add support for ndpanel factory functions, to create custom, domain-specific N-Dimensional containers

API Changes

  • The default binning/labeling behavior for resample has been changed to closed=’left’, label=’left’ for daily and lower frequencies. This had been a large source of confusion for users. See “what’s new” page for more on this. (GH2410)
  • Methods with inplace option now return None instead of the calling (modified) object (GH1893)
  • The special case DataFrame - TimeSeries doing column-by-column broadcasting has been deprecated. Users should explicitly do e.g. df.sub(ts, axis=0) instead. This is a legacy hack and can lead to subtle bugs.
  • inf/-inf are no longer considered as NA by isnull/notnull. To be clear, this is legacy cruft from early pandas. This behavior can be globally re-enabled using the new option mode.use_inf_as_null (GH2050, GH1919)
  • pandas.merge will now default to sort=False. For many use cases sorting the join keys is not necessary, and doing it by default is wasteful
  • Specify header=0 explicitly to replace existing column names in file in read_* functions.
  • Default column names for header-less parsed files (yielded by read_csv, etc.) are now the integers 0, 1, .... A new argument prefix has been added; to get the v0.9.x behavior specify prefix='X' (GH2034). This API change was made to make the default column names more consistent with the DataFrame constructor’s default column names when none are specified.
  • DataFrame selection using a boolean frame now preserves input shape
  • If function passed to Series.apply yields a Series, result will be a DataFrame (GH2316)
  • Values like YES/NO/yes/no will not be considered as boolean by default any longer in the file parsers. This can be customized using the new true_values and false_values options (GH2360)
  • obj.fillna() is no longer valid; make method=’pad’ no longer the default option, to be more explicit about what kind of filling to perform. Add ffill/bfill convenience functions per above (GH2284)
  • HDFStore.keys() now returns an absolute path-name for each key
  • to_string() now always returns a unicode string. (GH2224)
  • File parsers will not handle NA sentinel values arising from passed converter functions

Improvements to existing features

  • Add nrows option to DataFrame.from_records for iterators (GH1794)
  • Unstack/reshape algorithm rewrite to avoid high memory use in cases where the number of observed key-tuples is much smaller than the total possible number that could occur (GH2278). Also improves performance in most cases.
  • Support duplicate columns in DataFrame.from_records (GH2179)
  • Add normalize option to Series/DataFrame.asfreq (GH2137)
  • SparseSeries and SparseDataFrame construction from empty and scalar values now no longer create dense ndarrays unnecessarily (GH2322)
  • HDFStore now supports hierarchical keys (GH2397)
  • Support multiple query selection formats for HDFStore tables (GH1996)
  • Support del store['df'] syntax to delete HDFStores
  • Add multi-dtype support for HDFStore tables
  • min_itemsize parameter can be specified in HDFStore table creation
  • Indexing support in HDFStore tables (GH698)
  • Add line_terminator option to DataFrame.to_csv (GH2383)
  • added implementation of str(x)/unicode(x)/bytes(x) to major pandas data structures, which should do the right thing on both py2.x and py3.x. (GH2224)
  • Reduce groupby.apply overhead substantially by low-level manipulation of internal NumPy arrays in DataFrames (GH535)
  • Implement value_vars in melt and add melt to pandas namespace (GH2412)
  • Added boolean comparison operators to Panel
  • Enable Series.str.strip/lstrip/rstrip methods to take an argument (GH2411)
  • The DataFrame ctor now respects column ordering when given an OrderedDict (GH2455)
  • Assigning DatetimeIndex to Series changes the class to TimeSeries (GH2139)
  • Improve performance of .value_counts method on non-integer data (GH2480)
  • get_level_values method for MultiIndex return Index instead of ndarray (GH2449)
  • convert_to_r_dataframe conversion for datetime values (GH2351)
  • Allow DataFrame.to_csv to represent inf and nan differently (GH2026)
  • Add min_i argument to nancorr to specify minimum required observations (GH2002)
  • Add inplace option to sortlevel / sort functions on DataFrame (GH1873)
  • Enable DataFrame to accept scalar constructor values like Series (GH1856)
  • DataFrame.from_records now takes optional size parameter (GH1794)
  • include iris dataset (GH1709)
  • No datetime64 DataFrame column conversion of datetime.datetime with tzinfo (GH1581)
  • Micro-optimizations in DataFrame for tracking state of internal consolidation (GH217)
  • Format parameter in DataFrame.to_csv (GH1525)
  • Partial string slicing for DatetimeIndex for daily and higher frequencies (GH2306)
  • Implement col_space parameter in to_html and to_string in DataFrame (GH1000)
  • Override Series.tolist and box datetime64 types (GH2447)
  • Optimize unstack memory usage by compressing indices (GH2278)
  • Fix HTML repr in IPython qtconsole if opening window is small (GH2275)
  • Escape more special characters in console output (GH2492)
  • df.select now invokes bool on the result of crit(x) (GH2487)

Bug Fixes

  • Fix major performance regression in DataFrame.iteritems (GH2273)
  • Fixes bug when negative period passed to Series/DataFrame.diff (GH2266)
  • Escape tabs in console output to avoid alignment issues (GH2038)
  • Properly box datetime64 values when retrieving cross-section from mixed-dtype DataFrame (GH2272)
  • Fix concatenation bug leading to GH2057, GH2257
  • Fix regression in Index console formatting (GH2319)
  • Box Period data when assigning PeriodIndex to frame column (GH2243, GH2281)
  • Raise exception on calling reset_index on Series with inplace=True (GH2277)
  • Enable setting multiple columns in DataFrame with hierarchical columns (GH2295)
  • Respect dtype=object in DataFrame constructor (GH2291)
  • Fix DatetimeIndex.join bug with tz-aware indexes and how=’outer’ (GH2317)
  • pop(...) and del works with DataFrame with duplicate columns (GH2349)
  • Treat empty strings as NA in date parsing (rather than let dateutil do something weird) (GH2263)
  • Prevent uint64 -> int64 overflows (GH2355)
  • Enable joins between MultiIndex and regular Index (GH2024)
  • Fix time zone metadata issue when unioning non-overlapping DatetimeIndex objects (GH2367)
  • Raise/handle int64 overflows in parsers (GH2247)
  • Deleting of consecutive rows in HDFStore tables` is much faster than before
  • Appending on a HDFStore would fail if the table was not first created via put
  • Use col_space argument as minimum column width in DataFrame.to_html (GH2328)
  • Fix tz-aware DatetimeIndex.to_period (GH2232)
  • Fix DataFrame row indexing case with MultiIndex (GH2314)
  • Fix to_excel exporting issues with Timestamp objects in index (GH2294)
  • Fixes assigning scalars and array to hierarchical column chunk (GH1803)
  • Fixed a UnicodeDecodeError with series tidy_repr (GH2225)
  • Fixed issued with duplicate keys in an index (GH2347, GH2380)
  • Fixed issues re: Hash randomization, default on starting w/ py3.3 (GH2331)
  • Fixed issue with missing attributes after loading a pickled dataframe (GH2431)
  • Fix Timestamp formatting with tzoffset time zone in dateutil 2.1 (GH2443)
  • Fix GroupBy.apply issue when using BinGrouper to do ts binning (GH2300)
  • Fix issues resulting from datetime.datetime columns being converted to datetime64 when calling DataFrame.apply. (GH2374)
  • Raise exception when calling to_panel on non uniquely-indexed frame (GH2441)
  • Improved detection of console encoding on IPython zmq frontends (GH2458)
  • Preserve time zone when .append-ing two time series (GH2260)
  • Box timestamps when calling reset_index on time-zone-aware index rather than creating a tz-less datetime64 column (GH2262)
  • Enable searching non-string columns in DataFrame.filter(like=...) (GH2467)
  • Fixed issue with losing nanosecond precision upon conversion to DatetimeIndex(GH2252)
  • Handle timezones in Datetime.normalize (GH2338)
  • Fix test case where dtype specification with endianness causes failures on big endian machines (GH2318)
  • Fix plotting bug where upsampling causes data to appear shifted in time (GH2448)
  • Fix read_csv failure for UTF-16 with BOM and skiprows(GH2298)
  • read_csv with names arg not implicitly setting header=None(GH2459)
  • Unrecognized compression mode causes segfault in read_csv(GH2474)
  • In read_csv, header=0 and passed names should discard first row(GH2269)
  • Correctly route to stdout/stderr in read_table (GH2071)
  • Fix exception when Timestamp.to_datetime is called on a Timestamp with tzoffset (GH2471)
  • Fixed unintentional conversion of datetime64 to long in groupby.first() (GH2133)
  • Union of empty DataFrames now return empty with concatenated index (GH2307)
  • DataFrame.sort_index raises more helpful exception if sorting by column with duplicates (GH2488)
  • DataFrame.to_string formatters can be list, too (GH2520)
  • DataFrame.combine_first will always result in the union of the index and columns, even if one DataFrame is length-zero (GH2525)
  • Fix several DataFrame.icol/irow with duplicate indices issues (GH2228, GH2259)
  • Use Series names for column names when using concat with axis=1 (GH2489)
  • Raise Exception if start, end, periods all passed to date_range (GH2538)
  • Fix Panel resampling issue (GH2537)

pandas 0.9.1

Release date: 2012-11-14

New Features

  • Can specify multiple sort orders in DataFrame/Series.sort/sort_index (GH928)
  • New top and bottom options for handling NAs in rank (GH1508, GH2159)
  • Add where and mask functions to DataFrame (GH2109, GH2151)
  • Add at_time and between_time functions to DataFrame (GH2149)
  • Add flexible pow and rpow methods to DataFrame (GH2190)

API Changes

  • Upsampling period index “spans” intervals. Example: annual periods upsampled to monthly will span all months in each year
  • Period.end_time will yield timestamp at last nanosecond in the interval (GH2124, GH2125, GH1764)
  • File parsers no longer coerce to float or bool for columns that have custom converters specified (GH2184)

Improvements to existing features

  • Time rule inference for week-of-month (e.g. WOM-2FRI) rules (GH2140)
  • Improve performance of datetime + business day offset with large number of offset periods
  • Improve HTML display of DataFrame objects with hierarchical columns
  • Enable referencing of Excel columns by their column names (GH1936)
  • DataFrame.dot can accept ndarrays (GH2042)
  • Support negative periods in Panel.shift (GH2164)
  • Make .drop(...) work with non-unique indexes (GH2101)
  • Improve performance of Series/DataFrame.diff (re: GH2087)
  • Support unary ~ (__invert__) in DataFrame (GH2110)
  • Turn off pandas-style tick locators and formatters (GH2205)
  • DataFrame[DataFrame] uses DataFrame.where to compute masked frame (GH2230)

Bug Fixes

  • Fix some duplicate-column DataFrame constructor issues (GH2079)
  • Fix bar plot color cycle issues (GH2082)
  • Fix off-center grid for stacked bar plots (GH2157)
  • Fix plotting bug if inferred frequency is offset with N > 1 (GH2126)
  • Implement comparisons on date offsets with fixed delta (GH2078)
  • Handle inf/-inf correctly in read_* parser functions (GH2041)
  • Fix matplotlib unicode interaction bug
  • Make WLS r-squared match statsmodels 0.5.0 fixed value
  • Fix zero-trimming DataFrame formatting bug
  • Correctly compute/box datetime64 min/max values from Series.min/max (GH2083)
  • Fix unstacking edge case with unrepresented groups (GH2100)
  • Fix Series.str failures when using pipe pattern ‘|’ (GH2119)
  • Fix pretty-printing of dict entries in Series, DataFrame (GH2144)
  • Cast other datetime64 values to nanoseconds in DataFrame ctor (GH2095)
  • Alias Timestamp.astimezone to tz_convert, so will yield Timestamp (GH2060)
  • Fix timedelta64 formatting from Series (GH2165, GH2146)
  • Handle None values gracefully in dict passed to Panel constructor (GH2075)
  • Box datetime64 values as Timestamp objects in Series/DataFrame.iget (GH2148)
  • Fix Timestamp indexing bug in DatetimeIndex.insert (GH2155)
  • Use index name(s) (if any) in DataFrame.to_records (GH2161)
  • Don’t lose index names in Panel.to_frame/DataFrame.to_panel (GH2163)
  • Work around length-0 boolean indexing NumPy bug (GH2096)
  • Fix partial integer indexing bug in DataFrame.xs (GH2107)
  • Fix variety of cut/qcut string-bin formatting bugs (GH1978, GH1979)
  • Raise Exception when xs view not possible of MultiIndex’d DataFrame (GH2117)
  • Fix groupby(...).first() issue with datetime64 (GH2133)
  • Better floating point error robustness in some rolling_* functions (GH2114, GH2527)
  • Fix ewma NA handling in the middle of Series (GH2128)
  • Fix numerical precision issues in diff with integer data (GH2087)
  • Fix bug in MultiIndex.__getitem__ with NA values (GH2008)
  • Fix DataFrame.from_records dict-arg bug when passing columns (GH2179)
  • Fix Series and DataFrame.diff for integer dtypes (GH2087, GH2174)
  • Fix bug when taking intersection of DatetimeIndex with empty index (GH2129)
  • Pass through timezone information when calling DataFrame.align (GH2127)
  • Properly sort when joining on datetime64 values (GH2196)
  • Fix indexing bug in which False/True were being coerced to 0/1 (GH2199)
  • Many unicode formatting fixes (GH2201)
  • Fix improper MultiIndex conversion issue when assigning e.g. DataFrame.index (GH2200)
  • Fix conversion of mixed-type DataFrame to ndarray with dup columns (GH2236)
  • Fix duplicate columns issue (GH2218, GH2219)
  • Fix SparseSeries.__pow__ issue with NA input (GH2220)
  • Fix icol with integer sequence failure (GH2228)
  • Fixed resampling tz-aware time series issue (GH2245)
  • SparseDataFrame.icol was not returning SparseSeries (GH2227, GH2229)
  • Enable ExcelWriter to handle PeriodIndex (GH2240)
  • Fix issue constructing DataFrame from empty Series with name (GH2234)
  • Use console-width detection in interactive sessions only (GH1610)
  • Fix parallel_coordinates legend bug with mpl 1.2.0 (GH2237)
  • Make tz_localize work in corner case of empty Series (GH2248)

pandas 0.9.0

Release date: 10/7/2012

New Features

  • Add str.encode and str.decode to Series (GH1706)
  • Add to_latex method to DataFrame (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)
  • Recognize and convert more boolean values in file parsing (Yes, No, TRUE, FALSE, variants thereof) (GH1691, GH1295)
  • Add Panel.update method, analogous to DataFrame.update (GH1999, GH1988)

Improvements to existing features

  • Proper handling of NA values in merge operations (GH1990)
  • Add flags option for re.compile in some Series.str methods (GH1659)
  • Parsing of UTC date strings in read_* functions (GH1693)
  • Handle generator input to Series (GH1679)
  • Add na_action=’ignore’ to Series.map to quietly propagate NAs (GH1661)
  • Add args/kwds options to Series.apply (GH1829)
  • Add inplace option to Series/DataFrame.reset_index (GH1797)
  • Add level parameter to Series.reset_index
  • Add quoting option for DataFrame.to_csv (GH1902)
  • Indicate long column value truncation in DataFrame output with ... (GH1854)
  • DataFrame.dot will not do data alignment, and also work with Series (GH1915)
  • Add na option for missing data handling in some vectorized string methods (GH1689)
  • If index_label=False in DataFrame.to_csv, do not print fields/commas in the text output. Results in easier importing into R (GH1583)
  • Can pass tuple/list of axes to DataFrame.dropna to simplify repeated calls (dropping both columns and rows) (GH924)
  • Improve DataFrame.to_html output for hierarchically-indexed rows (do not repeat levels) (GH1929)
  • TimeSeries.between_time can now select times across midnight (GH1871)
  • Enable skip_footer parameter in ExcelFile.parse (GH1843)

API Changes

  • Change default header names in read_* functions to more Pythonic X0, X1, etc. instead of X.1, X.2. (GH2000)
  • Deprecated day_of_year API removed from PeriodIndex, use dayofyear (GH1723)
  • Don’t modify NumPy suppress printoption 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)
  • Enable skipfooter parameter in text parsers as an alias for skip_footer

Bug Fixes

  • Perform arithmetic column-by-column in mixed-type DataFrame to avoid type upcasting issues. Caused downstream DataFrame.diff bug (GH1896)
  • Fix matplotlib auto-color assignment when no custom spectrum passed. Also respect passed color keyword argument (GH1711)
  • Fix resampling logical error with closed=’left’ (GH1726)
  • Fix critical DatetimeIndex.union bugs (GH1730, GH1719, GH1745, GH1702, GH1753)
  • Fix critical DatetimeIndex.intersection bug with unanchored offsets (GH1708)
  • Fix MM-YYYY time series indexing case (GH1672)
  • Fix case where Categorical group key was not being passed into index in GroupBy result (GH1701)
  • Handle Ellipsis in Series.__getitem__/__setitem__ (GH1721)
  • Fix some bugs with handling datetime64 scalars of other units in NumPy 1.6 and 1.7 (GH1717)
  • Fix performance issue in MultiIndex.format (GH1746)
  • Fixed GroupBy bugs interacting with DatetimeIndex asof / map methods (GH1677)
  • Handle factors with NAs in pandas.rpy (GH1615)
  • Fix statsmodels import in pandas.stats.var (GH1734)
  • Fix DataFrame repr/info summary with non-unique columns (GH1700)
  • Fix Series.iget_value for non-unique indexes (GH1694)
  • Don’t lose tzinfo when passing DatetimeIndex as DataFrame column (GH1682)
  • Fix tz conversion with time zones that haven’t had any DST transitions since first date in the array (GH1673)
  • Fix field access with UTC->local conversion on unsorted arrays (GH1756)
  • Fix isnull handling of array-like (list) inputs (GH1755)
  • Fix regression in handling of Series in Series constructor (GH1671)
  • Fix comparison of Int64Index with DatetimeIndex (GH1681)
  • Fix min_periods handling in new rolling_max/min at array start (GH1695)
  • Fix errors with how=’median’ and generic NumPy resampling in some cases caused by SeriesBinGrouper (GH1648, GH1688)
  • When grouping by level, exclude unobserved levels (GH1697)
  • Don’t lose tzinfo in DatetimeIndex when shifting by different offset (GH1683)
  • Hack to support storing data with a zero-length axis in HDFStore (GH1707)
  • Fix DatetimeIndex tz-aware range generation issue (GH1674)
  • Fix method=’time’ interpolation with intraday data (GH1698)
  • Don’t plot all-NA DataFrame columns as zeros (GH1696)
  • Fix bug in scatter_plot with by option (GH1716)
  • Fix performance problem in infer_freq with lots of non-unique stamps (GH1686)
  • Fix handling of PeriodIndex as argument to create MultiIndex (GH1705)
  • Fix re: unicode MultiIndex level names in Series/DataFrame repr (GH1736)
  • Handle PeriodIndex in to_datetime instance method (GH1703)
  • Support StaticTzInfo in DatetimeIndex infrastructure (GH1692)
  • Allow MultiIndex setops with length-0 other type indexes (GH1727)
  • Fix handling of DatetimeIndex in DataFrame.to_records (GH1720)
  • Fix handling of general objects in isnull on which bool(...) fails (GH1749)
  • Fix .ix indexing with MultiIndex ambiguity (GH1678)
  • Fix .ix setting logic error with non-unique MultiIndex (GH1750)
  • Basic indexing now works on MultiIndex with > 1000000 elements, regression from earlier version of pandas (GH1757)
  • Handle non-float64 dtypes in fast DataFrame.corr/cov code paths (GH1761)
  • Fix DatetimeIndex.isin to function properly (GH1763)
  • Fix conversion of array of tz-aware datetime.datetime to DatetimeIndex with right time zone (GH1777)
  • Fix DST issues with generating ancxhored date ranges (GH1778)
  • Fix issue calling sort on result of Series.unique (GH1807)
  • Fix numerical issue leading to square root of negative number in rolling_std (GH1840)
  • Let Series.str.split accept no arguments (like str.split) (GH1859)
  • Allow user to have dateutil 2.1 installed on a Python 2 system (GH1851)
  • Catch ImportError less aggressively in pandas/__init__.py (GH1845)
  • Fix pip source installation bug when installing from GitHub (GH1805)
  • Fix error when window size > array size in rolling_apply (GH1850)
  • Fix pip source installation issues via SSH from GitHub
  • Fix OLS.summary when column is a tuple (GH1837)
  • Fix bug in __doc__ patching when -OO passed to interpreter (GH1792 GH1741 GH1774)
  • Fix unicode console encoding issue in IPython notebook (GH1782, GH1768)
  • Fix unicode formatting issue with Series.name (GH1782)
  • Fix bug in DataFrame.duplicated with datetime64 columns (GH1833)
  • Fix bug in Panel internals resulting in error when doing fillna after truncate not changing size of panel (GH1823)
  • Prevent segfault due to MultiIndex not being supported in HDFStore table format (GH1848)
  • Fix UnboundLocalError in Panel.__setitem__ and add better error (GH1826)
  • Fix to_csv issues with list of string entries. Isnull works on list of strings now too (GH1791)
  • Fix Timestamp comparisons with datetime values outside the nanosecond range (1677-2262)
  • Revert to prior behavior of normalize_date with datetime.date objects (return datetime)
  • Fix broken interaction between np.nansum and Series.any/all
  • Fix bug with multiple column date parsers (GH1866)
  • DatetimeIndex.union(Int64Index) was broken
  • Make plot x vs y interface consistent with integer indexing (GH1842)
  • set_index inplace modified data even if unique check fails (GH1831)
  • Only use Q-OCT/NOV/DEC in quarterly frequency inference (GH1789)
  • Upcast to dtype=object when unstacking boolean DataFrame (GH1820)
  • Fix float64/float32 merging bug (GH1849)
  • Fixes to Period.start_time for non-daily frequencies (GH1857)
  • Fix failure when converter used on index_col in read_csv (GH1835)
  • Implement PeriodIndex.append so that pandas.concat works correctly (GH1815)
  • Avoid Cython out-of-bounds access causing segfault sometimes in pad_2d, backfill_2d
  • Fix resampling error with intraday times and anchored target time (like AS-DEC) (GH1772)
  • Fix .ix indexing bugs with mixed-integer indexes (GH1799)
  • Respect passed color keyword argument in Series.plot (GH1890)
  • Fix rolling_min/max when the window is larger than the size of the input array. Check other malformed inputs (GH1899, GH1897)
  • Rolling variance / standard deviation with only a single observation in window (GH1884)
  • Fix unicode sheet name failure in to_excel (GH1828)
  • Override DatetimeIndex.min/max to return Timestamp objects (GH1895)
  • Fix column name formatting issue in length-truncated column (GH1906)
  • Fix broken handling of copying Index metadata to new instances created by view(...) calls inside the NumPy infrastructure
  • Support datetime.date again in DateOffset.rollback/rollforward
  • Raise Exception if set passed to Series constructor (GH1913)
  • Add TypeError when appending HDFStore table w/ wrong index type (GH1881)
  • Don’t raise exception on empty inputs in EW functions (e.g. ewma) (GH1900)
  • Make asof work correctly with PeriodIndex (GH1883)
  • Fix extlinks in doc build
  • Fill boolean DataFrame with NaN when calling shift (GH1814)
  • Fix setuptools bug causing pip not to Cythonize .pyx files sometimes
  • Fix negative integer indexing regression in .ix from 0.7.x (GH1888)
  • Fix error while retrieving timezone and utc offset from subclasses of datetime.tzinfo without .zone and ._utcoffset attributes (GH1922)
  • Fix DataFrame formatting of small, non-zero FP numbers (GH1911)
  • Various fixes by upcasting of date -> datetime (GH1395)
  • Raise better exception when passing multiple functions with the same name, such as lambdas, to GroupBy.aggregate
  • Fix DataFrame.apply with axis=1 on a non-unique index (GH1878)
  • Proper handling of Index subclasses in pandas.unique (GH1759)
  • Set index names in DataFrame.from_records (GH1744)
  • Fix time series indexing error with duplicates, under and over hash table size cutoff (GH1821)
  • Handle list keys in addition to tuples in DataFrame.xs when partial-indexing a hierarchically-indexed DataFrame (GH1796)
  • Support multiple column selection in DataFrame.__getitem__ with duplicate columns (GH1943)
  • Fix time zone localization bug causing improper fields (e.g. hours) in time zones that have not had a UTC transition in a long time (GH1946)
  • Fix errors when parsing and working with with fixed offset timezones (GH1922, GH1928)
  • Fix text parser bug when handling UTC datetime objects generated by dateutil (GH1693)
  • Fix plotting bug when ‘B’ is the inferred frequency but index actually contains weekends (GH1668, GH1669)
  • Fix plot styling bugs (GH1666, GH1665, GH1658)
  • Fix plotting bug with index/columns with unicode (GH1685)
  • Fix DataFrame constructor bug when passed Series with datetime64 dtype in a dict (GH1680)
  • Fixed regression in generating DatetimeIndex using timezone aware datetime.datetime (GH1676)
  • Fix DataFrame bug when printing concatenated DataFrames with duplicated columns (GH1675)
  • Fixed bug when plotting time series with multiple intraday frequencies (GH1732)
  • Fix bug in DataFrame.duplicated to enable iterables other than list-types as input argument (GH1773)
  • Fix resample bug when passed list of lambdas as how argument (GH1808)
  • Repr fix for MultiIndex level with all NAs (GH1971)
  • Fix PeriodIndex slicing bug when slice start/end are out-of-bounds (GH1977)
  • Fix read_table bug when parsing unicode (GH1975)
  • Fix BlockManager.iget bug when dealing with non-unique MultiIndex as columns (GH1970)
  • Fix reset_index bug if both drop and level are specified (GH1957)
  • Work around unsafe NumPy object->int casting with Cython function (GH1987)
  • Fix datetime64 formatting bug in DataFrame.to_csv (GH1993)
  • Default start date in pandas.io.data to 1/1/2000 as the docs say (GH2011)

pandas 0.8.1

Release date: July 22, 2012

New Features

  • Add vectorized, NA-friendly string methods to Series (GH1621, GH620)
  • Can pass dict of per-column line styles to DataFrame.plot (GH1559)
  • Selective plotting to secondary y-axis on same subplot (GH1640)
  • Add new bootstrap_plot plot function
  • Add new parallel_coordinates plot function (GH1488)
  • Add radviz plot function (GH1566)
  • Add multi_sparse option to set_printoptions to modify display of hierarchical indexes (GH1538)
  • Add dropna method to Panel (GH171)

Improvements to existing features

  • Use moving min/max algorithms from Bottleneck in rolling_min/rolling_max for > 100x speedup. (GH1504, GH50)
  • Add Cython group median method for >15x speedup (GH1358)
  • Drastically improve to_datetime performance on ISO8601 datetime strings (with no time zones) (GH1571)
  • Improve single-key groupby performance on large data sets, accelerate use of groupby with a Categorical variable
  • Add ability to append hierarchical index levels with set_index and to drop single levels with reset_index (GH1569, GH1577)
  • Always apply passed functions in resample, even if upsampling (GH1596)
  • Avoid unnecessary copies in DataFrame constructor with explicit dtype (GH1572)
  • Cleaner DatetimeIndex string representation with 1 or 2 elements (GH1611)
  • Improve performance of array-of-Period to PeriodIndex, convert such arrays to PeriodIndex inside Index (GH1215)
  • More informative string representation for weekly Period objects (GH1503)
  • Accelerate 3-axis multi data selection from homogeneous Panel (GH979)
  • Add adjust option to ewma to disable adjustment factor (GH1584)
  • Add new matplotlib converters for high frequency time series plotting (GH1599)
  • Handling of tz-aware datetime.datetime objects in to_datetime; raise Exception unless utc=True given (GH1581)

Bug Fixes

  • Fix NA handling in DataFrame.to_panel (GH1582)
  • Handle TypeError issues inside PyObject_RichCompareBool calls in khash (GH1318)
  • Fix resampling bug to lower case daily frequency (GH1588)
  • Fix kendall/spearman DataFrame.corr bug with no overlap (GH1595)
  • Fix bug in DataFrame.set_index (GH1592)
  • Don’t ignore axes in boxplot if by specified (GH1565)
  • Fix Panel .ix indexing with integers bug (GH1603)
  • Fix Partial indexing bugs (years, months, ...) with PeriodIndex (GH1601)
  • Fix MultiIndex console formatting issue (GH1606)
  • Unordered index with duplicates doesn’t yield scalar location for single entry (GH1586)
  • Fix resampling of tz-aware time series with “anchored” freq (GH1591)
  • Fix DataFrame.rank error on integer data (GH1589)
  • Selection of multiple SparseDataFrame columns by list in __getitem__ (GH1585)
  • Override Index.tolist for compatibility with MultiIndex (GH1576)
  • Fix hierarchical summing bug with MultiIndex of length 1 (GH1568)
  • Work around numpy.concatenate use/bug in Series.set_value (GH1561)
  • Ensure Series/DataFrame are sorted before resampling (GH1580)
  • Fix unhandled IndexError when indexing very large time series (GH1562)
  • Fix DatetimeIndex intersection logic error with irregular indexes (GH1551)
  • Fix unit test errors on Python 3 (GH1550)
  • Fix .ix indexing bugs in duplicate DataFrame index (GH1201)
  • Better handle errors with non-existing objects in HDFStore (GH1254)
  • Don’t copy int64 array data in DatetimeIndex when copy=False (GH1624)
  • Fix resampling of conforming periods quarterly to annual (GH1622)
  • Don’t lose index name on resampling (GH1631)
  • Support python-dateutil version 2.1 (GH1637)
  • Fix broken scatter_matrix axis labeling, esp. with time series (GH1625)
  • Fix cases where extra keywords weren’t being passed on to matplotlib from Series.plot (GH1636)
  • Fix BusinessMonthBegin logic for dates before 1st bday of month (GH1645)
  • Ensure string alias converted (valid in DatetimeIndex.get_loc) in DataFrame.xs / __getitem__ (GH1644)
  • Fix use of string alias timestamps with tz-aware time series (GH1647)
  • Fix Series.max/min and Series.describe on len-0 series (GH1650)
  • Handle None values in dict passed to concat (GH1649)
  • Fix Series.interpolate with method=’values’ and DatetimeIndex (GH1646)
  • Fix IndexError in left merges on a DataFrame with 0-length (GH1628)
  • Fix DataFrame column width display with UTF-8 encoded characters (GH1620)
  • Handle case in pandas.io.data.get_data_yahoo where Yahoo! returns duplicate dates for most recent business day
  • Avoid downsampling when plotting mixed frequencies on the same subplot (GH1619)
  • Fix read_csv bug when reading a single line (GH1553)
  • Fix bug in C code causing monthly periods prior to December 1969 to be off (GH1570)

pandas 0.8.0

Release date: 6/29/2012

New Features

  • New unified DatetimeIndex class for nanosecond-level timestamp data
  • New Timestamp datetime.datetime subclass with easy time zone conversions, and support for nanoseconds
  • New PeriodIndex class for timespans, calendar logic, and Period scalar object
  • High performance resampling of timestamp and period data. New resample method of all pandas data structures
  • New frequency names plus shortcut string aliases like ‘15h’, ‘1h30min’
  • Time series string indexing shorthand (GH222)
  • Add week, dayofyear array and other timestamp array-valued field accessor functions to DatetimeIndex
  • Add GroupBy.prod optimized aggregation function and ‘prod’ fast time series conversion method (GH1018)
  • Implement robust frequency inference function and inferred_freq attribute on DatetimeIndex (GH391)
  • New tz_convert and tz_localize methods in Series / DataFrame
  • Convert DatetimeIndexes to UTC if time zones are different in join/setops (GH864)
  • Add limit argument for forward/backward filling to reindex, fillna, etc. (GH825 and others)
  • Add support for indexes (dates or otherwise) with duplicates and common sense indexing/selection functionality
  • Series/DataFrame.update methods, in-place variant of combine_first (GH961)
  • Add match function to API (GH502)
  • Add Cython-optimized first, last, min, max, prod functions to GroupBy (GH994, GH1043)
  • Dates can be split across multiple columns (GH1227, GH1186)
  • Add experimental support for converting pandas DataFrame to R data.frame via rpy2 (GH350, GH1212)
  • Can pass list of (name, function) to GroupBy.aggregate to get aggregates in a particular order (GH610)
  • Can pass dicts with lists of functions or dicts to GroupBy aggregate to do much more flexible multiple function aggregation (GH642, GH610)
  • New ordered_merge functions for merging DataFrames with ordered data. Also supports group-wise merging for panel data (GH813)
  • Add keys() method to DataFrame
  • Add flexible replace method for replacing potentially values to Series and DataFrame (GH929, GH1241)
  • Add ‘kde’ plot kind for Series/DataFrame.plot (GH1059)
  • More flexible multiple function aggregation with GroupBy
  • Add pct_change function to Series/DataFrame
  • Add option to interpolate by Index values in Series.interpolate (GH1206)
  • Add max_colwidth option for DataFrame, defaulting to 50
  • Conversion of DataFrame through rpy2 to R data.frame (GH1282, )
  • Add keys() method on DataFrame (GH1240)
  • Add new match function to API (similar to R) (GH502)
  • Add dayfirst option to parsers (GH854)
  • Add method argument to align method for forward/backward fillin (GH216)
  • Add Panel.transpose method for rearranging axes (GH695)
  • Add new cut function (patterned after R) for discretizing data into equal range-length bins or arbitrary breaks of your choosing (GH415)
  • Add new qcut for cutting with quantiles (GH1378)
  • Add value_counts top level array method (GH1392)
  • Added Andrews curves plot tupe (GH1325)
  • Add lag plot (GH1440)
  • Add autocorrelation_plot (GH1425)
  • Add support for tox and Travis CI (GH1382)
  • Add support for Categorical use in GroupBy (GH292)
  • Add any and all methods to DataFrame (GH1416)
  • Add secondary_y option to Series.plot
  • Add experimental lreshape function for reshaping wide to long

Improvements to existing features

  • Switch to klib/khash-based hash tables in Index classes for better performance in many cases and lower memory footprint
  • Shipping some functions from scipy.stats to reduce dependency, e.g. Series.describe and DataFrame.describe (GH1092)
  • Can create MultiIndex by passing list of lists or list of arrays to Series, DataFrame constructor, etc. (GH831)
  • Can pass arrays in addition to column names to DataFrame.set_index (GH402)
  • Improve the speed of “square” reindexing of homogeneous DataFrame objects by significant margin (GH836)
  • Handle more dtypes when passed MaskedArrays in DataFrame constructor (GH406)
  • Improved performance of join operations on integer keys (GH682)
  • Can pass multiple columns to GroupBy object, e.g. grouped[[col1, col2]] to only aggregate a subset of the value columns (GH383)
  • Add histogram / kde plot options for scatter_matrix diagonals (GH1237)
  • Add inplace option to Series/DataFrame.rename and sort_index, DataFrame.drop_duplicates (GH805, GH207)
  • More helpful error message when nothing passed to Series.reindex (GH1267)
  • Can mix array and scalars as dict-value inputs to DataFrame ctor (GH1329)
  • Use DataFrame columns’ name for legend title in plots
  • Preserve frequency in DatetimeIndex when possible in boolean indexing operations
  • Promote datetime.date values in data alignment operations (GH867)
  • Add order method to Index classes (GH1028)
  • Avoid hash table creation in large monotonic hash table indexes (GH1160)
  • Store time zones in HDFStore (GH1232)
  • Enable storage of sparse data structures in HDFStore (GH85)
  • Enable Series.asof to work with arrays of timestamp inputs
  • Cython implementation of DataFrame.corr speeds up by > 100x (GH1349, GH1354)
  • Exclude “nuisance” columns automatically in GroupBy.transform (GH1364)
  • Support functions-as-strings in GroupBy.transform (GH1362)
  • Use index name as xlabel/ylabel in plots (GH1415)
  • Add convert_dtype option to Series.apply to be able to leave data as dtype=object (GH1414)
  • Can specify all index level names in concat (GH1419)
  • Add dialect keyword to parsers for quoting conventions (GH1363)
  • Enable DataFrame[bool_DataFrame] += value (GH1366)
  • Add retries argument to get_data_yahoo to try to prevent Yahoo! API 404s (GH826)
  • Improve performance of reshaping by using O(N) categorical sorting
  • Series names will be used for index of DataFrame if no index passed (GH1494)
  • Header argument in DataFrame.to_csv can accept a list of column names to use instead of the object’s columns (GH921)
  • Add raise_conflict argument to DataFrame.update (GH1526)
  • Support file-like objects in ExcelFile (GH1529)

API Changes

  • Rename pandas._tseries to pandas.lib
  • Rename Factor to Categorical and add improvements. Numerous Categorical bug fixes
  • Frequency name overhaul, WEEKDAY/EOM and rules with @ deprecated. get_legacy_offset_name backwards compatibility function added
  • Raise ValueError in DataFrame.__nonzero__, so “if df” no longer works (GH1073)
  • Change BDay (business day) to not normalize dates by default (GH506)
  • Remove deprecated DataMatrix name
  • Default merge suffixes for overlap now have underscores instead of periods to facilitate tab completion, etc. (GH1239)
  • Deprecation of offset, time_rule timeRule parameters throughout codebase
  • Series.append and DataFrame.append no longer check for duplicate indexes by default, add verify_integrity parameter (GH1394)
  • Refactor Factor class, old constructor moved to Factor.from_array
  • Modified internals of MultiIndex to use less memory (no longer represented as array of tuples) internally, speed up construction time and many methods which construct intermediate hierarchical indexes (GH1467)

Bug Fixes

  • Fix OverflowError from storing pre-1970 dates in HDFStore by switching to datetime64 (GH179)
  • Fix logical error with February leap year end in YearEnd offset
  • Series([False, nan]) was getting casted to float64 (GH1074)
  • Fix binary operations between boolean Series and object Series with booleans and NAs (GH1074, GH1079)
  • Couldn’t assign whole array to column in mixed-type DataFrame via .ix (GH1142)
  • Fix label slicing issues with float index values (GH1167)
  • Fix segfault caused by empty groups passed to groupby (GH1048)
  • Fix occasionally misbehaved reindexing in the presence of NaN labels (GH522)
  • Fix imprecise logic causing weird Series results from .apply (GH1183)
  • Unstack multiple levels in one shot, avoiding empty columns in some cases. Fix pivot table bug (GH1181)
  • Fix formatting of MultiIndex on Series/DataFrame when index name coincides with label (GH1217)
  • Handle Excel 2003 #N/A as NaN from xlrd (GH1213, GH1225)
  • Fix timestamp locale-related deserialization issues with HDFStore by moving to datetime64 representation (GH1081, GH809)
  • Fix DataFrame.duplicated/drop_duplicates NA value handling (GH557)
  • Actually raise exceptions in fast reducer (GH1243)
  • Fix various timezone-handling bugs from 0.7.3 (GH969)
  • GroupBy on level=0 discarded index name (GH1313)
  • Better error message with unmergeable DataFrames (GH1307)
  • Series.__repr__ alignment fix with unicode index values (GH1279)
  • Better error message if nothing passed to reindex (GH1267)
  • More robust NA handling in DataFrame.drop_duplicates (GH557)
  • Resolve locale-based and pre-epoch HDF5 timestamp deserialization issues (GH973, GH1081, GH179)
  • Implement Series.repeat (GH1229)
  • Fix indexing with namedtuple and other tuple subclasses (GH1026)
  • Fix float64 slicing bug (GH1167)
  • Parsing integers with commas (GH796)
  • Fix groupby improper data type when group consists of one value (GH1065)
  • Fix negative variance possibility in nanvar resulting from floating point error (GH1090)
  • Consistently set name on groupby pieces (GH184)
  • Treat dict return values as Series in GroupBy.apply (GH823)
  • Respect column selection for DataFrame in in GroupBy.transform (GH1365)
  • Fix MultiIndex partial indexing bug (GH1352)
  • Enable assignment of rows in mixed-type DataFrame via .ix (GH1432)
  • Reset index mapping when grouping Series in Cython (GH1423)
  • Fix outer/inner DataFrame.join with non-unique indexes (GH1421)
  • Fix MultiIndex groupby bugs with empty lower levels (GH1401)
  • Calling fillna with a Series will have same behavior as with dict (GH1486)
  • SparseSeries reduction bug (GH1375)
  • Fix unicode serialization issue in HDFStore (GH1361)
  • Pass keywords to pyplot.boxplot in DataFrame.boxplot (GH1493)
  • Bug fixes in MonthBegin (GH1483)
  • Preserve MultiIndex names in drop (GH1513)
  • Fix Panel DataFrame slice-assignment bug (GH1533)
  • Don’t use locals() in read_* functions (GH1547)

pandas 0.7.3

Release date: April 12, 2012

New Features

  • Support for non-unique indexes: indexing and selection, many-to-one and many-to-many joins (GH1306)
  • Added fixed-width file reader, read_fwf (GH952)
  • Add group_keys argument to groupby to not add group names to MultiIndex in result of apply (GH938)
  • DataFrame can now accept non-integer label slicing (GH946). Previously only DataFrame.ix was able to do so.
  • DataFrame.apply now retains name attributes on Series objects (GH983)
  • Numeric DataFrame comparisons with non-numeric values now raises proper TypeError (GH943). Previously raise “PandasError: DataFrame constructor not properly called!”
  • Add kurt methods to Series and DataFrame (GH964)
  • Can pass dict of column -> list/set NA values for text parsers (GH754)
  • Allows users specified NA values in text parsers (GH754)
  • Parsers checks for openpyxl dependency and raises ImportError if not found (GH1007)
  • New factory function to create HDFStore objects that can be used in a with statement so users do not have to explicitly call HDFStore.close (GH1005)
  • pivot_table is now more flexible with same parameters as groupby (GH941)
  • Added stacked bar plots (GH987)
  • scatter_matrix method in pandas/tools/plotting.py (GH935)
  • DataFrame.boxplot returns plot results for ex-post styling (GH985)
  • Short version number accessible as pandas.version.short_version (GH930)
  • Additional documentation in panel.to_frame (GH942)
  • More informative Series.apply docstring regarding element-wise apply (GH977)
  • Notes on rpy2 installation (GH1006)
  • Add rotation and font size options to hist method (GH1012)
  • Use exogenous / X variable index in result of OLS.y_predict. Add OLS.predict method (GH1027, GH1008)

API Changes

  • Calling apply on grouped Series, e.g. describe(), will no longer yield DataFrame by default. Will have to call unstack() to get prior behavior
  • NA handling in non-numeric comparisons has been tightened up (GH933, GH953)
  • No longer assign dummy names key_0, key_1, etc. to groupby index (GH1291)

Bug Fixes

  • Fix logic error when selecting part of a row in a DataFrame with a MultiIndex index (GH1013)
  • Series comparison with Series of differing length causes crash (GH1016).
  • Fix bug in indexing when selecting section of hierarchically-indexed row (GH1013)
  • DataFrame.plot(logy=True) has no effect (GH1011).
  • Broken arithmetic operations between SparsePanel-Panel (GH1015)
  • Unicode repr issues in MultiIndex with non-ASCII characters (GH1010)
  • DataFrame.lookup() returns inconsistent results if exact match not present (GH1001)
  • DataFrame arithmetic operations not treating None as NA (GH992)
  • DataFrameGroupBy.apply returns incorrect result (GH991)
  • Series.reshape returns incorrect result for multiple dimensions (GH989)
  • Series.std and Series.var ignores ddof parameter (GH934)
  • DataFrame.append loses index names (GH980)
  • DataFrame.plot(kind=’bar’) ignores color argument (GH958)
  • Inconsistent Index comparison results (GH948)
  • Improper int dtype DataFrame construction from data with NaN (GH846)
  • Removes default ‘result’ name in groupby results (GH995)
  • DataFrame.from_records no longer mutate input columns (GH975)
  • Use Index name when grouping by it (GH1313)

pandas 0.7.2

Release date: March 16, 2012

New Features

  • Add additional tie-breaking methods in DataFrame.rank (GH874)
  • Add ascending parameter to rank in Series, DataFrame (GH875)
  • Add sort_columns parameter to allow unsorted plots (GH918)
  • IPython tab completion on GroupBy objects

API Changes

  • Series.sum returns 0 instead of NA when called on an empty series. Analogously for a DataFrame whose rows or columns are length 0 (GH844)

Improvements to existing features

  • Don’t use groups dict in Grouper.size (GH860)
  • Use khash for Series.value_counts, add raw function to algorithms.py (GH861)
  • Enable column access via attributes on GroupBy (GH882)
  • Enable setting existing columns (only) via attributes on DataFrame, Panel (GH883)
  • Intercept __builtin__.sum in groupby (GH885)
  • Can pass dict to DataFrame.fillna to use different values per column (GH661)
  • Can select multiple hierarchical groups by passing list of values in .ix (GH134)
  • Add level keyword to drop for dropping values from a level (GH159)
  • Add coerce_float option on DataFrame.from_records (GH893)
  • Raise exception if passed date_parser fails in read_csv
  • Add axis option to DataFrame.fillna (GH174)
  • Fixes to Panel to make it easier to subclass (GH888)

Bug Fixes

  • Fix overflow-related bugs in groupby (GH850, GH851)
  • Fix unhelpful error message in parsers (GH856)
  • Better err msg for failed boolean slicing of dataframe (GH859)
  • Series.count cannot accept a string (level name) in the level argument (GH869)
  • Group index platform int check (GH870)
  • concat on axis=1 and ignore_index=True raises TypeError (GH871)
  • Further unicode handling issues resolved (GH795)
  • Fix failure in multiindex-based access in Panel (GH880)
  • Fix DataFrame boolean slice assignment failure (GH881)
  • Fix combineAdd NotImplementedError for SparseDataFrame (GH887)
  • Fix DataFrame.to_html encoding and columns (GH890, GH891, GH909)
  • Fix na-filling handling in mixed-type DataFrame (GH910)
  • Fix to DataFrame.set_value with non-existant row/col (GH911)
  • Fix malformed block in groupby when excluding nuisance columns (GH916)
  • Fix inconsistant NA handling in dtype=object arrays (GH925)
  • Fix missing center-of-mass computation in ewmcov (GH862)
  • Don’t raise exception when opening read-only HDF5 file (GH847)
  • Fix possible out-of-bounds memory access in 0-length Series (GH917)

pandas 0.7.1

Release date: February 29, 2012

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

Improvements to existing features

  • Improve performance and memory usage of fillna on DataFrame
  • Can concatenate a list of Series along axis=1 to obtain a DataFrame (GH787)

Bug Fixes

  • Fix memory leak when inserting large number of columns into a single DataFrame (GH790)
  • Appending length-0 DataFrame with new columns would not result in those new columns being part of the resulting concatenated DataFrame (GH782)
  • Fixed groupby corner case when passing dictionary grouper and as_index is False (GH819)
  • Fixed bug whereby bool array sometimes had object dtype (GH820)
  • Fix exception thrown on np.diff (GH816)
  • Fix to_records where columns are non-strings (GH822)
  • Fix Index.intersection where indices have incomparable types (GH811)
  • Fix ExcelFile throwing an exception for two-line file (GH837)
  • Add clearer error message in csv parser (GH835)
  • Fix loss of fractional seconds in HDFStore (GH513)
  • Fix DataFrame join where columns have datetimes (GH787)
  • Work around numpy performance issue in take (GH817)
  • Improve comparison operations for NA-friendliness (GH801)
  • Fix indexing operation for floating point values (GH780, GH798)
  • Fix groupby case resulting in malformed dataframe (GH814)
  • Fix behavior of reindex of Series dropping name (GH812)
  • Improve on redudant groupby computation (GH775)
  • Catch possible NA assignment to int/bool series with exception (GH839)

pandas 0.7.0

Release date: 2/9/2012

New Features

  • New 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 concat function for concatenating DataFrame or Panel objects along an axis. Can form union or intersection of the other axes. Improves performance of DataFrame.append (GH468, GH479, GH273)
  • Handle differently-indexed output values in DataFrame.apply (GH498)
  • Can pass list of dicts (e.g., a list of shallow JSON objects) to DataFrame constructor (GH526)
  • Add reorder_levels method to Series and DataFrame (GH534)
  • Add dict-like get function to DataFrame and Panel (GH521)
  • DataFrame.iterrows method for efficiently iterating through the rows of a DataFrame
  • Added DataFrame.to_panel with code adapted from LongPanel.to_long
  • 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 (PR GH554)
  • Add logy option to Series.plot for log-scaling on the Y axis
  • Add index, header, and justify options to DataFrame.to_string. Add option to (GH570, GH571)
  • Can pass multiple DataFrames to DataFrame.join to join on index (GH115)
  • Can pass multiple Panels to Panel.join (GH115)
  • Can pass multiple DataFrames to DataFrame.append to concatenate (stack) and multiple Series to Series.append too
  • 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)
  • Add verbose option to read_csv and read_table to show number of NA values inserted in non-numeric columns (GH614)
  • Can pass a list of dicts or Series to DataFrame.append to concatenate multiple rows (GH464)
  • Add level argument to DataFrame.xs for selecting data from other MultiIndex levels. Can take one or more levels with potentially a tuple of keys for flexible retrieval of data (GH371, GH629)
  • New crosstab function for easily computing frequency tables (GH170)
  • Can pass a list of functions to aggregate with groupby on a DataFrame, yielding an aggregated result with hierarchical columns (GH166)
  • Add integer-indexing functions iget in Series and irow / iget in DataFrame (GH628)
  • Add new Series.unique function, significantly faster than numpy.unique (GH658)
  • Add new cummin and cummax instance methods to Series and DataFrame (GH647)
  • Add new value_range function to return min/max of a dataframe (GH288)
  • Add drop parameter to reset_index method of DataFrame and added method to Series as well (GH699)
  • Add isin method to Index objects, works just like Series.isin (GH GH657)
  • Implement array interface on Panel so that ufuncs work (re: GH740)
  • Add sort option to DataFrame.join (GH731)
  • Improved handling of NAs (propagation) in binary operations with dtype=object arrays (GH737)
  • Add abs method to Pandas objects
  • Added algorithms module to start collecting central algos

API Changes

  • Label-indexing with integer indexes now raises KeyError if a label is not found instead of falling back on location-based indexing (GH700)
  • Label-based slicing via ix or [] on Series will now only work if exact matches for the labels are found or if the index is monotonic (for range selections)
  • Label-based slicing and sequences of labels can be passed to [] on a Series for both getting and setting (GH86)
  • [] operator (__getitem__ and __setitem__) will raise KeyError with integer indexes when an index is not contained in the index. The prior behavior would fall back on position-based indexing if a key was not found in the index which would lead to subtle bugs. This is now consistent with the behavior of .ix on DataFrame and friends (GH328)
  • Rename DataFrame.delevel to DataFrame.reset_index and add deprecation warning
  • Series.sort (an in-place operation) called on a Series which is a view on a larger array (e.g. a column in a DataFrame) will generate an Exception to prevent accidentally modifying the data source (GH316)
  • Refactor to remove deprecated LongPanel class (GH552)
  • Deprecated Panel.to_long, renamed to to_frame
  • Deprecated colSpace argument in DataFrame.to_string, renamed to col_space
  • Rename precision to accuracy in engineering float formatter (GH GH395)
  • The default delimiter for read_csv is comma rather than letting csv.Sniffer infer it
  • Rename col_or_columns argument in DataFrame.drop_duplicates (GH GH734)

Improvements to existing features

  • 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 (GH 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)
  • Some performance enhancements in constructing a Panel from a dict of DataFrame objects
  • Made Index._get_duplicates a public method by removing the underscore
  • Prettier printing of floats, and column spacing fix (GH395, GH571)
  • Add bold_rows option to DataFrame.to_html (GH586)
  • Improve the performance of DataFrame.sort_index by up to 5x or more when sorting by multiple columns
  • Substantially improve performance of DataFrame and Series constructors when passed a nested dict or dict, respectively (GH540, GH621)
  • Modified setup.py so that pip / setuptools will install dependencies (GH GH507, various pull requests)
  • Unstack called on DataFrame with non-MultiIndex will return Series (GH GH477)
  • Improve DataFrame.to_string and console formatting to be more consistent in the number of displayed digits (GH395)
  • Use bottleneck if available for performing NaN-friendly statistical operations that it implemented (GH91)
  • Monkey-patch context to traceback in DataFrame.apply to indicate which row/column the function application failed on (GH614)
  • Improved ability of read_table and read_clipboard to parse console-formatted DataFrames (can read the row of index names, etc.)
  • Can pass list of group labels (without having to convert to an ndarray yourself) to groupby in some cases (GH659)
  • Use kind argument to Series.order for selecting different sort kinds (GH668)
  • Add option to Series.to_csv to omit the index (GH684)
  • Add delimiter as an alternative to sep in read_csv and other parsing functions
  • Substantially improved performance of groupby on DataFrames with many columns by aggregating blocks of columns all at once (GH745)
  • Can pass a file handle or StringIO to Series/DataFrame.to_csv (GH765)
  • Can pass sequence of integers to DataFrame.irow(icol) and Series.iget, (GH GH654)
  • Prototypes for some vectorized string functions
  • Add float64 hash table to solve the Series.unique problem with NAs (GH714)
  • Memoize objects when reading from file to reduce memory footprint
  • Can get and set a column of a DataFrame with hierarchical columns containing “empty” (‘’) lower levels without passing the empty levels (PR GH768)

Bug Fixes

  • Raise exception in out-of-bounds indexing of Series instead of seg-faulting, regression from earlier releases (GH495)
  • Fix error when joining DataFrames of different dtypes within the same typeclass (e.g. float32 and float64) (GH486)
  • Fix bug in Series.min/Series.max on objects like datetime.datetime (GH GH487)
  • Preserve index names in Index.union (GH501)
  • Fix bug in Index joining causing subclass information (like DateRange type) to be lost in some cases (GH500)
  • Accept empty list as input to DataFrame constructor, regression from 0.6.0 (GH491)
  • Can output DataFrame and Series with ndarray objects in a dtype=object array (GH490)
  • Return empty string from Series.to_string when called on empty Series (GH GH488)
  • Fix exception passing empty list to DataFrame.from_records
  • Fix Index.format bug (excluding name field) with datetimes with time info
  • Fix scalar value access in Series to always return NumPy scalars, regression from prior versions (GH510)
  • Handle rows skipped at beginning of file in read_* functions (GH505)
  • Handle improper dtype casting in set_value methods
  • Unary ‘-‘ / __neg__ operator on DataFrame was returning integer values
  • Unbox 0-dim ndarrays from certain operators like all, any in Series
  • Fix handling of missing columns (was combine_first-specific) in DataFrame.combine for general case (GH529)
  • Fix type inference logic with boolean lists and arrays in DataFrame indexing
  • Use centered sum of squares in R-square computation if entity_effects=True in panel regression
  • Handle all NA case in Series.{corr, cov}, was raising exception (GH548)
  • Aggregating by multiple levels with level argument to DataFrame, Series stat method, was broken (GH545)
  • Fix Cython buf when converter passed to read_csv produced a numeric array (buffer dtype mismatch when passed to Cython type inference function) (GH GH546)
  • Fix exception when setting scalar value using .ix on a DataFrame with a MultiIndex (GH551)
  • Fix outer join between two DateRanges with different offsets that returned an invalid DateRange
  • Cleanup DataFrame.from_records failure where index argument is an integer
  • Fix Data.from_records failure when passed a dictionary
  • Fix NA handling in {Series, DataFrame}.rank with non-floating point dtypes
  • Fix bug related to integer type-checking in .ix-based indexing
  • Handle non-string index name passed to DataFrame.from_records
  • DataFrame.insert caused the columns name(s) field to be discarded (GH527)
  • Fix erroneous in monotonic many-to-one left joins
  • Fix DataFrame.to_string to remove extra column white space (GH571)
  • Format floats to default to same number of digits (GH395)
  • Added decorator to copy docstring from one function to another (GH449)
  • Fix error in monotonic many-to-one left joins
  • Fix __eq__ comparison between DateOffsets with different relativedelta keywords passed
  • Fix exception caused by parser converter returning strings (GH583)
  • Fix MultiIndex formatting bug with integer names (GH601)
  • Fix bug in handling of non-numeric aggregates in Series.groupby (GH612)
  • Fix TypeError with tuple subclasses (e.g. namedtuple) in DataFrame.from_records (GH611)
  • Catch misreported console size when running IPython within Emacs
  • Fix minor bug in pivot table margins, loss of index names and length-1 ‘All’ tuple in row labels
  • Add support for legacy WidePanel objects to be read from HDFStore
  • Fix out-of-bounds segfault in pad_object and backfill_object methods when either source or target array are empty
  • Could not create a new column in a DataFrame from a list of tuples
  • Fix bugs preventing SparseDataFrame and SparseSeries working with groupby (GH666)
  • Use sort kind in Series.sort / argsort (GH668)
  • Fix DataFrame operations on non-scalar, non-pandas objects (GH672)
  • Don’t convert DataFrame column to integer type when passing integer to __setitem__ (GH669)
  • Fix downstream bug in pivot_table caused by integer level names in MultiIndex (GH678)
  • Fix SparseSeries.combine_first when passed a dense Series (GH687)
  • Fix performance regression in HDFStore loading when DataFrame or Panel stored in table format with datetimes
  • Raise Exception in DateRange when offset with n=0 is passed (GH683)
  • Fix get/set inconsistency with .ix property and integer location but non-integer index (GH707)
  • Use right dropna function for SparseSeries. Return dense Series for NA fill value (GH730)
  • Fix Index.format bug causing incorrectly string-formatted Series with datetime indexes (GH726, GH758)
  • Fix errors caused by object dtype arrays passed to ols (GH759)
  • Fix error where column names lost when passing list of labels to DataFrame.__getitem__, (GH662)
  • Fix error whereby top-level week iterator overwrote week instance
  • Fix circular reference causing memory leak in sparse array / series / frame, (GH663)
  • Fix integer-slicing from integers-as-floats (GH670)
  • Fix zero division errors in nanops from object dtype arrays in all NA case (GH676)
  • Fix csv encoding when using unicode (GH705, GH717, GH738)
  • Fix assumption that each object contains every unique block type in concat, (GH708)
  • Fix sortedness check of multiindex in to_panel (GH719, 720)
  • Fix that None was not treated as NA in PyObjectHashtable
  • Fix hashing dtype because of endianness confusion (GH747, GH748)
  • Fix SparseSeries.dropna to return dense Series in case of NA fill value (GH GH730)
  • Use map_infer instead of np.vectorize. handle NA sentinels if converter yields numeric array, (GH753)
  • Fixes and improvements to DataFrame.rank (GH742)
  • Fix catching AttributeError instead of NameError for bottleneck
  • Try to cast non-MultiIndex to better dtype when calling reset_index (GH726 GH440)
  • Fix #1.QNAN0’ float bug on 2.6/win64
  • Allow subclasses of dicts in DataFrame constructor, with tests
  • Fix problem whereby set_index destroys column multiindex (GH764)
  • Hack around bug in generating DateRange from naive DateOffset (GH770)
  • Fix bug in DateRange.intersection causing incorrect results with some overlapping ranges (GH771)

Thanks

  • Craig Austin
  • Chris Billington
  • Marius Cobzarenco
  • Mario Gamboa-Cavazos
  • Hans-Martin Gaudecker
  • Arthur Gerigk
  • Yaroslav Halchenko
  • Jeff Hammerbacher
  • Matt Harrison
  • Andreas Hilboll
  • Luc Kesters
  • Adam Klein
  • Gregg Lind
  • Solomon Negusse
  • Wouter Overmeire
  • Christian Prinoth
  • Jeff Reback
  • Sam Reckoner
  • Craig Reeson
  • Jan Schulz
  • Skipper Seabold
  • Ted Square
  • Graham Taylor
  • Aman Thakral
  • Chris Uga
  • Dieter Vandenbussche
  • Texas P.
  • Pinxing Ye
  • ... and everyone I forgot

pandas 0.6.1

Release date: 12/13/2011

API Changes

  • Rename names argument in DataFrame.from_records to columns. Add deprecation warning
  • Boolean get/set operations on Series with boolean Series will reindex instead of requiring that the indexes be exactly equal (GH429)

New Features

  • Can pass Series to DataFrame.append with ignore_index=True for appending a single row (GH430)
  • Add Spearman and Kendall correlation options to Series.corr and DataFrame.corr (GH428)
  • Add new get_value and set_value methods to Series, DataFrame, and Panel to very low-overhead access to scalar elements. df.get_value(row, column) is about 3x faster than df[column][row] by handling fewer cases (GH437, GH438). Add similar methods to sparse data structures for compatibility
  • Add Qt table widget to sandbox (GH435)
  • DataFrame.align can accept Series arguments, add axis keyword (GH461)
  • Implement new SparseList and SparseArray data structures. SparseSeries now derives from SparseArray (GH463)
  • max_columns / max_rows options in set_printoptions (GH453)
  • Implement Series.rank and DataFrame.rank, fast versions of scipy.stats.rankdata (GH428)
  • Implement DataFrame.from_items alternate constructor (GH444)
  • DataFrame.convert_objects method for inferring better dtypes for object columns (GH302)
  • Add rolling_corr_pairwise function for computing Panel of correlation matrices (GH189)
  • Add margins option to pivot_table for computing subgroup aggregates (GH GH114)
  • Add Series.from_csv function (GH482)

Improvements to existing features

  • Improve memory usage of DataFrame.describe (do not copy data unnecessarily) (GH425)
  • Use same formatting function for outputting floating point Series to console as in DataFrame (GH420)
  • DataFrame.delevel will try to infer better dtype for new columns (GH440)
  • Exclude non-numeric types in DataFrame.{corr, cov}
  • Override Index.astype to enable dtype casting (GH412)
  • Use same float formatting function for Series.__repr__ (GH420)
  • Use available console width to output DataFrame columns (GH453)
  • Accept ndarrays when setting items in Panel (GH452)
  • Infer console width when printing __repr__ of DataFrame to console (PR GH453)
  • Optimize scalar value lookups in the general case by 25% or more in Series and DataFrame
  • Can pass DataFrame/DataFrame and DataFrame/Series to rolling_corr/rolling_cov (GH462)
  • 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 GH158)
  • MultiIndex.get_level_values can take the level name
  • More helpful error message when DataFrame.plot fails on one of the columns (GH478)
  • Improve performance of DataFrame.{index, columns} attribute lookup

Bug Fixes

  • Fix O(K^2) memory leak caused by inserting many columns without consolidating, had been present since 0.4.0 (GH467)
  • DataFrame.count should return Series with zero instead of NA with length-0 axis (GH423)
  • Fix Yahoo! Finance API usage in pandas.io.data (GH419, GH427)
  • Fix upstream bug causing failure in Series.align with empty Series (GH434)
  • Function passed to DataFrame.apply can return a list, as long as it’s the right length. Regression from 0.4 (GH432)
  • Don’t “accidentally” upcast scalar values when indexing using .ix (GH431)
  • Fix groupby exception raised with as_index=False and single column selected (GH421)
  • Implement DateOffset.__ne__ causing downstream bug (GH456)
  • Fix __doc__-related issue when converting py -> pyo with py2exe
  • Bug fix in left join Cython code with duplicate monotonic labels
  • Fix bug when unstacking multiple levels described in GH451
  • Exclude NA values in dtype=object arrays, regression from 0.5.0 (GH469)
  • Use Cython map_infer function in DataFrame.applymap to properly infer output type, handle tuple return values and other things that were breaking (GH465)
  • Handle floating point index values in HDFStore (GH454)
  • Fixed stale column reference bug (cached Series object) caused by type change / item deletion in DataFrame (GH473)
  • Index.get_loc should always raise Exception when there are duplicates
  • Handle differently-indexed Series input to DataFrame constructor (GH475)
  • Omit nuisance columns in multi-groupby with Python function
  • Buglet in handling of single grouping in general apply
  • Handle type inference properly when passing list of lists or tuples to DataFrame constructor (GH484)
  • Preserve Index / MultiIndex names in GroupBy.apply concatenation step (GH GH481)

Thanks

  • Ralph Bean
  • Luca Beltrame
  • Marius Cobzarenco
  • Andreas Hilboll
  • Jev Kuznetsov
  • Adam Lichtenstein
  • Wouter Overmeire
  • Fernando Perez
  • Nathan Pinger
  • Christian Prinoth
  • Alex Reyfman
  • Joon Ro
  • Chang She
  • Ted Square
  • Chris Uga
  • Dieter Vandenbussche

pandas 0.6.0

Release date: 11/25/2011

API Changes

  • Arithmetic methods like sum will attempt to sum dtype=object values by default instead of excluding them (GH382)

New Features

  • Add melt function to pandas.core.reshape
  • Add level parameter to group by level in Series and DataFrame descriptive statistics (GH313)
  • Add head and tail methods to Series, analogous to to DataFrame (PR GH296)
  • Add Series.isin function which checks if each value is contained in a passed sequence (GH289)
  • Add float_format option to Series.to_string
  • Add skip_footer (GH291) and converters (GH343) options to read_csv and read_table
  • Add proper, tested weighted least squares to standard and panel OLS (GH GH303)
  • Add drop_duplicates and duplicated functions for removing duplicate DataFrame rows and checking for duplicate rows, respectively (GH319)
  • Implement logical (boolean) operators &, |, ^ on DataFrame (GH347)
  • Add Series.mad, mean absolute deviation, matching DataFrame
  • Add QuarterEnd DateOffset (GH321)
  • Add matrix multiplication function dot to DataFrame (GH65)
  • Add orient option to Panel.from_dict to ease creation of mixed-type Panels (GH359, GH301)
  • Add DataFrame.from_dict with similar orient option
  • Can now pass list of tuples or list of lists to DataFrame.from_records for fast conversion to DataFrame (GH357)
  • Can pass multiple levels to groupby, e.g. df.groupby(level=[0, 1]) (GH GH103)
  • Can sort by multiple columns in DataFrame.sort_index (GH92, GH362)
  • Add fast get_value and put_value methods to DataFrame and micro-performance tweaks (GH360)
  • Add cov instance methods to Series and DataFrame (GH194, GH362)
  • Add bar plot option to DataFrame.plot (GH348)
  • Add idxmin and idxmax functions to Series and DataFrame for computing index labels achieving maximum and minimum values (GH286)
  • Add read_clipboard function for parsing DataFrame from OS clipboard, should work across platforms (GH300)
  • Add nunique function to Series for counting unique elements (GH297)
  • DataFrame constructor will use Series name if no columns passed (GH373)
  • Support regular expressions and longer delimiters in read_table/read_csv, but does not handle quoted strings yet (GH364)
  • Add DataFrame.to_html for formatting DataFrame to HTML (GH387)
  • MaskedArray can be passed to DataFrame constructor and masked values will be converted to NaN (GH396)
  • Add DataFrame.boxplot function (GH368, others)
  • Can pass extra args, kwds to DataFrame.apply (GH376)

Improvements to existing features

  • Raise more helpful exception if date parsing fails in DateRange (GH298)
  • Vastly improved performance of GroupBy on axes with a MultiIndex (GH299)
  • Print level names in hierarchical index in Series repr (GH305)
  • Return DataFrame when performing GroupBy on selected column and as_index=False (GH308)
  • Can pass vector to on argument in DataFrame.join (GH312)
  • Don’t show Series name if it’s None in the repr, also omit length for short Series (GH317)
  • Show legend by default in DataFrame.plot, add legend boolean flag (GH GH324)
  • Significantly improved performance of Series.order, which also makes np.unique called on a Series faster (GH327)
  • Faster cythonized count by level in Series and DataFrame (GH341)
  • Raise exception if dateutil 2.0 installed on Python 2.x runtime (GH346)
  • Significant GroupBy performance enhancement with multiple keys with many “empty” combinations
  • New Cython vectorized function map_infer speeds up Series.apply and Series.map significantly when passed elementwise Python function, motivated by GH355
  • Cythonized cache_readonly, resulting in substantial micro-performance enhancements throughout the codebase (GH361)
  • Special Cython matrix iterator for applying arbitrary reduction operations with 3-5x better performance than np.apply_along_axis (GH309)
  • Add raw option to DataFrame.apply for getting better performance when the passed function only requires an ndarray (GH309)
  • Improve performance of MultiIndex.from_tuples
  • Can pass multiple levels to stack and unstack (GH370)
  • Can pass multiple values columns to pivot_table (GH381)
  • Can call DataFrame.delevel with standard Index with name set (GH393)
  • Use Series name in GroupBy for result index (GH363)
  • Refactor Series/DataFrame stat methods to use common set of NaN-friendly function
  • Handle NumPy scalar integers at C level in Cython conversion routines

Bug Fixes

  • Fix bug in DataFrame.to_csv when writing a DataFrame with an index name (GH290)
  • DataFrame should clear its Series caches on consolidation, was causing “stale” Series to be returned in some corner cases (GH304)
  • DataFrame constructor failed if a column had a list of tuples (GH293)
  • Ensure that Series.apply always returns a Series and implement Series.round (GH314)
  • Support boolean columns in Cythonized groupby functions (GH315)
  • DataFrame.describe should not fail if there are no numeric columns, instead return categorical describe (GH323)
  • Fixed bug which could cause columns to be printed in wrong order in DataFrame.to_string if specific list of columns passed (GH325)
  • Fix legend plotting failure if DataFrame columns are integers (GH326)
  • Shift start date back by one month for Yahoo! Finance API in pandas.io.data (GH329)
  • Fix DataFrame.join failure on unconsolidated inputs (GH331)
  • DataFrame.min/max will no longer fail on mixed-type DataFrame (GH337)
  • Fix read_csv / read_table failure when passing list to index_col that is not in ascending order (GH349)
  • Fix failure passing Int64Index to Index.union when both are monotonic
  • Fix error when passing SparseSeries to (dense) DataFrame constructor
  • Added missing bang at top of setup.py (GH352)
  • Change is_monotonic on MultiIndex so it properly compares the tuples
  • Fix MultiIndex outer join logic (GH351)
  • Set index name attribute with single-key groupby (GH358)
  • Bug fix in reflexive binary addition in Series and DataFrame for non-commutative operations (like string concatenation) (GH353)
  • setupegg.py will invoke Cython (GH192)
  • Fix block consolidation bug after inserting column into MultiIndex (GH366)
  • Fix bug in join operations between Index and Int64Index (GH367)
  • Handle min_periods=0 case in moving window functions (GH365)
  • Fixed corner cases in DataFrame.apply/pivot with empty DataFrame (GH378)
  • Fixed repr exception when Series name is a tuple
  • Always return DateRange from asfreq (GH390)
  • Pass level names to swaplavel (GH379)
  • Don’t lose index names in MultiIndex.droplevel (GH394)
  • Infer more proper return type in DataFrame.apply when no columns or rows depending on whether the passed function is a reduction (GH389)
  • Always return NA/NaN from Series.min/max and DataFrame.min/max when all of a row/column/values are NA (GH384)
  • Enable partial setting with .ix / advanced indexing (GH397)
  • Handle mixed-type DataFrames correctly in unstack, do not lose type information (GH403)
  • Fix integer name formatting bug in Index.format and in Series.__repr__
  • Handle label types other than string passed to groupby (GH405)
  • Fix bug in .ix-based indexing with partial retrieval when a label is not contained in a level
  • Index name was not being pickled (GH408)
  • Level name should be passed to result index in GroupBy.apply (GH416)

Thanks

  • Craig Austin
  • Marius Cobzarenco
  • Joel Cross
  • Jeff Hammerbacher
  • Adam Klein
  • Thomas Kluyver
  • Jev Kuznetsov
  • Kieran O’Mahony
  • Wouter Overmeire
  • Nathan Pinger
  • Christian Prinoth
  • Skipper Seabold
  • Chang She
  • Ted Square
  • Aman Thakral
  • Chris Uga
  • Dieter Vandenbussche
  • carljv
  • rsamson

pandas 0.5.0

Release date: 10/24/2011

This release of pandas includes a number of API changes (see below) and cleanup of deprecated APIs from pre-0.4.0 releases. There are also bug fixes, new features, numerous significant performance enhancements, and includes a new ipython completer hook to enable tab completion of DataFrame columns accesses and attributes (a new feature).

In addition to the changes listed here from 0.4.3 to 0.5.0, the minor releases 4.1, 0.4.2, and 0.4.3 brought some significant new functionality and performance improvements that are worth taking a look at.

Thanks to all for bug reports, contributed patches and generally providing feedback on the library.

API Changes

  • read_table, read_csv, and ExcelFile.parse default arguments for index_col is now None. To use one or more of the columns as the resulting DataFrame’s index, these must be explicitly specified now
  • Parsing functions like read_csv no longer parse dates by default (GH GH225)
  • Removed weights option in panel regression which was not doing anything principled (GH155)
  • Changed buffer argument name in Series.to_string to buf
  • Series.to_string and DataFrame.to_string now return strings by default instead of printing to sys.stdout
  • Deprecated nanRep argument in various to_string and to_csv functions in favor of na_rep. Will be removed in 0.6 (GH275)
  • Renamed delimiter to sep in DataFrame.from_csv for consistency
  • Changed order of Series.clip arguments to match those of numpy.clip and added (unimplemented) out argument so numpy.clip can be called on a Series (GH272)
  • Series functions renamed (and thus deprecated) in 0.4 series have been removed:
    • asOf, use asof
    • toDict, use to_dict
    • toString, use to_string
    • toCSV, use to_csv
    • merge, use map
    • applymap, use apply
    • combineFirst, use combine_first
    • _firstTimeWithValue use first_valid_index
    • _lastTimeWithValue use last_valid_index
  • DataFrame functions renamed / deprecated in 0.4 series have been removed:
    • asMatrix method, use as_matrix or values attribute
    • combineFirst, use combine_first
    • getXS, use xs
    • merge, use join
    • fromRecords, use from_records
    • fromcsv, use from_csv
    • toRecords, use to_records
    • toDict, use to_dict
    • toString, use to_string
    • toCSV, use to_csv
    • _firstTimeWithValue use first_valid_index
    • _lastTimeWithValue use last_valid_index
    • toDataMatrix is no longer needed
    • rows() method, use index attribute
    • cols() method, use columns attribute
    • dropEmptyRows(), use dropna(how=’all’)
    • dropIncompleteRows(), use dropna()
    • tapply(f), use apply(f, axis=1)
    • tgroupby(keyfunc, aggfunc), use groupby with axis=1

Deprecations Removed

  • indexField argument in DataFrame.from_records
  • missingAtEnd argument in Series.order. Use na_last instead
  • Series.fromValue classmethod, use regular Series constructor instead
  • Functions parseCSV, parseText, and parseExcel methods in pandas.io.parsers have been removed
  • Index.asOfDate function
  • Panel.getMinorXS (use minor_xs) and Panel.getMajorXS (use major_xs)
  • Panel.toWide, use Panel.to_wide instead

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
  • Add 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
  • Added column attribute access to DataFrame, e.g. df.A equivalent to df[‘A’] if ‘A’ is a column in the DataFrame (GH213)
  • Added IPython tab completion hook for DataFrame columns. (GH233, GH230)
  • Implement Series.describe for Series containing objects (GH241)
  • Add inner join option to DataFrame.join when joining on key(s) (GH248)
  • Can select set of DataFrame columns by passing a list to __getitem__ (GH GH253)
  • Can use & and | to intersection / union Index objects, respectively (GH 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
  • Implemented Panel.take
  • Add set_eng_float_format function for setting alternate DataFrame floating point string formatting
  • Add convenience set_index function for creating a DataFrame index from its existing columns

Improvements to existing features

  • Major performance improvements in file parsing functions read_csv and read_table
  • Added Cython function for converting tuples to ndarray very fast. Speeds up many MultiIndex-related operations
  • File parsing functions like read_csv and read_table will explicitly check if a parsed index has duplicates and raise a more helpful exception rather than deferring the check until later
  • Refactored merging / joining code into a tidy class and disabled unnecessary computations in the float/object case, thus getting about 10% better performance (GH211)
  • Improved speed of DataFrame.xs on mixed-type DataFrame objects by about 5x, regression from 0.3.0 (GH215)
  • With new DataFrame.align method, speeding up binary operations between differently-indexed DataFrame objects by 10-25%.
  • Significantly sped up conversion of nested dict into DataFrame (GH212)
  • Can pass hierarchical index level name to groupby instead of the level number if desired (GH223)
  • Add support for different delimiters in DataFrame.to_csv (GH244)
  • Add more helpful error message when importing pandas post-installation from the source directory (GH250)
  • Significantly speed up DataFrame __repr__ and count on large mixed-type DataFrame objects
  • Better handling of pyx file dependencies in Cython module build (GH271)

Bug Fixes

  • read_csv / read_table fixes
    • Be less aggressive about converting float->int in cases of floating point representations of integers like 1.0, 2.0, etc.
    • “True”/”False” will not get correctly converted to boolean
    • Index name attribute will get set when specifying an index column
    • Passing column names should force header=None (GH257)
    • Don’t modify passed column names when index_col is not None (GH258)
    • Can sniff CSV separator in zip file (since seek is not supported, was failing before)
  • Worked around matplotlib “bug” in which series[:, np.newaxis] fails. Should be reported upstream to matplotlib (GH224)
  • DataFrame.iteritems was not returning Series with the name attribute set. Also neither was DataFrame._series
  • Can store datetime.date objects in HDFStore (GH231)
  • Index and Series names are now stored in HDFStore
  • Fixed problem in which data would get upcasted to object dtype in GroupBy.apply operations (GH237)
  • Fixed outer join bug with empty DataFrame (GH238)
  • Can create empty Panel (GH239)
  • Fix join on single key when passing list with 1 entry (GH246)
  • Don’t raise Exception on plotting DataFrame with an all-NA column (GH251, GH254)
  • Bug min/max errors when called on integer DataFrames (GH241)
  • DataFrame.iteritems and DataFrame._series not assigning name attribute
  • Panel.__repr__ raised exception on length-0 major/minor axes
  • DataFrame.join on key with empty DataFrame produced incorrect columns
  • Implemented MultiIndex.diff (GH260)
  • Int64Index.take and MultiIndex.take lost name field, fix downstream issue GH262
  • Can pass list of tuples to Series (GH270)
  • Can pass level name to DataFrame.stack
  • Support set operations between MultiIndex and Index
  • Fix many corner cases in MultiIndex set operations - Fix MultiIndex-handling bug with GroupBy.apply when returned groups are not indexed the same
  • Fix corner case bugs in DataFrame.apply
  • Setting DataFrame index did not cause Series cache to get cleared
  • Various int32 -> int64 platform-specific issues
  • Don’t be too aggressive converting to integer when parsing file with MultiIndex (GH285)
  • Fix bug when slicing Series with negative indices before beginning

Thanks

  • Thomas Kluyver
  • Daniel Fortunov
  • Aman Thakral
  • Luca Beltrame
  • Wouter Overmeire

pandas 0.4.3

Release date: 10/9/2011

is is largely a bugfix release from 0.4.2 but also includes a handful of new d enhanced features. Also, pandas can now be installed and used on Python 3 hanks Thomas Kluyver!).

New Features

  • Python 3 support using 2to3 (GH200, Thomas Kluyver)
  • Add name attribute to Series and added relevant logic and tests. Name now prints as part of Series.__repr__
  • Add name attribute to standard Index so that stacking / unstacking does not discard names and so that indexed DataFrame objects can be reliably round-tripped to flat files, pickle, HDF5, etc.
  • Add isnull and notnull as instance methods on Series (GH209, GH203)

Improvements to existing features

  • Skip xlrd-related unit tests if not installed
  • Index.append and MultiIndex.append can accept a list of Index objects to concatenate together
  • 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)
  • Refactored Series.__repr__ to be a bit more clean and consistent

API Changes

  • Series.describe and DataFrame.describe now bring the 25% and 75% quartiles instead of the 10% and 90% deciles. The other outputs have not changed
  • Series.toString will print deprecation warning, has been de-camelCased to to_string

Bug Fixes

  • Fix broken interaction between Index and Int64Index when calling intersection. Implement Int64Index.intersection
  • MultiIndex.sortlevel discarded the level names (GH202)
  • Fix bugs in groupby, join, and append due to improper concatenation of MultiIndex objects (GH201)
  • Fix regression from 0.4.1, isnull and notnull ceased to work on other kinds of Python scalar objects like datetime.datetime
  • Raise more helpful exception when attempting to write empty DataFrame or LongPanel to HDFStore (GH204)
  • Use stdlib csv module to properly escape strings with commas in DataFrame.to_csv (GH206, Thomas Kluyver)
  • Fix Python ndarray access in Cython code for sparse blocked index integrity check
  • Fix bug writing Series to CSV in Python 3 (GH209)
  • Miscellaneous Python 3 bugfixes

Thanks

  • Thomas Kluyver
  • rsamson

pandas 0.4.2

Release date: 10/3/2011

is is a performance optimization release with several bug fixes. The new t64Index and new merging / joining Cython code and related Python frastructure are the main new additions

New Features

  • Added fast Int64Index type with specialized join, union, intersection. Will result in significant performance enhancements for int64-based time series (e.g. using NumPy’s datetime64 one day) and also faster operations on DataFrame objects storing record array-like data.
  • Refactored Index classes to have a join method and associated data alignment routines throughout the codebase to be able to leverage optimized joining / merging routines.
  • Added Series.align method for aligning two series with choice of join method
  • Wrote faster Cython data alignment / merging routines resulting in substantial speed increases
  • Added is_monotonic property to Index classes with associated Cython code to evaluate the monotonicity of the Index values
  • Add method get_level_values to MultiIndex
  • Implemented shallow copy of BlockManager object in DataFrame internals

Improvements to existing features

  • Improved performance of isnull and notnull, a regression from v0.3.0 (GH187)
  • Wrote templating / code generation script to auto-generate Cython code for various functions which need to be available for the 4 major data types used in pandas (float64, bool, object, int64)
  • 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
  • Improved performance of DateRange.union with overlapping ranges and non-cacheable offsets (like Minute). Implemented analogous fast DateRange.intersection for overlapping ranges.
  • 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
  • Added informative Exception when passing dict to DataFrame groupby aggregation with axis != 0

API Changes

Bug Fixes

  • Fixed minor unhandled exception in Cython code implementing fast groupby aggregation operations
  • Fixed bug in unstacking code manifesting with more than 3 hierarchical levels
  • Throw exception when step specified in label-based slice (GH185)
  • Fix isnull to correctly work with np.float32. Fix upstream bug described in GH182
  • Finish implementation of as_index=False in groupby for DataFrame aggregation (GH181)
  • Raise SkipTest for pre-epoch HDFStore failure. Real fix will be sorted out via datetime64 dtype

Thanks

  • Uri Laserson
  • Scott Sinclair

pandas 0.4.1

Release date: 9/25/2011

is is primarily a bug fix release but includes some new features and improvements

New Features

  • Added new DataFrame methods get_dtype_counts and property dtypes
  • Setting of values using .ix indexing attribute in mixed-type DataFrame objects has been implemented (fixes GH135)
  • read_csv can read multiple columns into a MultiIndex. DataFrame’s to_csv method will properly write out a MultiIndex which can be read back (GH151, thanks to Skipper Seabold)
  • Wrote fast time series merging / joining methods in Cython. Will be integrated later into DataFrame.join and related functions
  • Added ignore_index option to DataFrame.append for combining unindexed records stored in a DataFrame

Improvements to existing features

  • Some speed enhancements with internal Index type-checking function
  • DataFrame.rename has a new copy parameter which can rename a DataFrame in place
  • Enable unstacking by level name (GH142)
  • Enable sortlevel to work by level name (GH141)
  • read_csv can automatically “sniff” other kinds of delimiters using csv.Sniffer (GH146)
  • Improved speed of unit test suite by about 40%
  • Exception will not be raised calling HDFStore.remove on non-existent node with where clause
  • Optimized _ensure_index function resulting in performance savings in type-checking Index objects

API Changes

Bug Fixes

  • Fixed DataFrame constructor bug causing downstream problems (e.g. .copy() failing) when passing a Series as the values along with a column name and index
  • Fixed single-key groupby on DataFrame with as_index=False (GH160)
  • Series.shift was failing on integer Series (GH154)
  • unstack methods were producing incorrect output in the case of duplicate hierarchical labels. An exception will now be raised (GH147)
  • Calling count with level argument caused reduceat failure or segfault in earlier NumPy (GH169)
  • Fixed DataFrame.corrwith to automatically exclude non-numeric data (GH GH144)
  • Unicode handling bug fixes in DataFrame.to_string (GH138)
  • Excluding OLS degenerate unit test case that was causing platform specific failure (GH149)
  • Skip blosc-dependent unit tests for PyTables < 2.2 (GH137)
  • Calling copy on DateRange did not copy over attributes to the new object (GH168)
  • Fix bug in HDFStore in which Panel data could be appended to a Table with different item order, thus resulting in an incorrect result read back

Thanks

  • Yaroslav Halchenko
  • Jeff Reback
  • Skipper Seabold
  • Dan Lovell
  • Nick Pentreath

pandas 0.4.0

Release date: 9/12/2011

New Features

  • pandas.core.sparse module: “Sparse” (mostly-NA, or some other fill value) versions of Series, DataFrame, and Panel. For low-density data, this will result in significant performance boosts, and smaller memory footprint. Added to_sparse methods to Series, DataFrame, and Panel. See online documentation for more on these
  • Fancy indexing operator on Series / DataFrame, e.g. via .ix operator. Both getting and setting of values is supported; however, setting values will only currently work on homogeneously-typed DataFrame objects. Things like:
    • series.ix[[d1, d2, d3]]
    • frame.ix[5:10, [‘C’, ‘B’, ‘A’]], frame.ix[5:10, ‘A’:’C’]
    • frame.ix[date1:date2]
  • Significantly enhanced groupby functionality
    • Can groupby multiple keys, e.g. df.groupby([‘key1’, ‘key2’]). Iteration with multiple groupings products a flattened tuple
    • “Nuisance” columns (non-aggregatable) will automatically be excluded from DataFrame aggregation operations
    • Added automatic “dispatching to Series / DataFrame methods to more easily invoke methods on groups. e.g. s.groupby(crit).std() will work even though std is not implemented on the GroupBy class
  • Hierarchical / multi-level indexing
    • New the MultiIndex class. Integrated MultiIndex into Series and DataFrame fancy indexing, slicing, __getitem__ and __setitem, reindexing, etc. Added level keyword argument to groupby to enable grouping by a level of a MultiIndex
  • New data reshaping functions: stack and unstack on DataFrame and Series
    • Integrate with MultiIndex to enable sophisticated reshaping of data
  • Index objects (labels for axes) are now capable of holding tuples
  • Series.describe, DataFrame.describe: produces an R-like table of summary statistics about each data column
  • DataFrame.quantile, Series.quantile for computing sample quantiles of data across requested axis
  • Added general DataFrame.dropna method to replace dropIncompleteRows and dropEmptyRows, deprecated those.
  • Series arithmetic methods with optional fill_value for missing data, e.g. a.add(b, fill_value=0). If a location is missing for both it will still be missing in the result though.
  • fill_value option has been added to DataFrame.{add, mul, sub, div} methods similar to Series
  • Boolean indexing with DataFrame objects: data[data > 0.1] = 0.1 or data[data> other] = 1.
  • pytz / tzinfo support in DateRange
    • tz_localize, tz_normalize, and tz_validate methods added
  • Added ExcelFile class to pandas.io.parsers for parsing multiple sheets out of a single Excel 2003 document
  • GroupBy aggregations can now optionally broadcast, e.g. produce an object of the same size with the aggregated value propagated
  • Added select function in all data structures: reindex axis based on arbitrary criterion (function returning boolean value), e.g. frame.select(lambda x: ‘foo’ in x, axis=1)
  • DataFrame.consolidate method, API function relating to redesigned internals
  • DataFrame.insert method for inserting column at a specified location rather than the default __setitem__ behavior (which puts it at the end)
  • HDFStore class in pandas.io.pytables has been largely rewritten using patches from Jeff Reback from others. It now supports mixed-type DataFrame and Series data and can store Panel objects. It also has the option to query DataFrame and Panel data. Loading data from legacy HDFStore files is supported explicitly in the code
  • Added set_printoptions method to modify appearance of DataFrame tabular output
  • rolling_quantile functions; a moving version of Series.quantile / DataFrame.quantile
  • Generic rolling_apply moving window function
  • New drop method added to Series, DataFrame, etc. which can drop a set of labels from an axis, producing a new object
  • reindex methods now sport a copy option so that data is not forced to be copied then the resulting object is indexed the same
  • Added sort_index methods to Series and Panel. Renamed DataFrame.sort to sort_index. Leaving DataFrame.sort for now.
  • Added skipna option to statistical instance methods on all the data structures
  • pandas.io.data module providing a consistent interface for reading time series data from several different sources

Improvements to existing features

  • The 2-dimensional DataFrame and DataMatrix classes have been extensively redesigned internally into a single class DataFrame, preserving where possible their optimal performance characteristics. This should reduce confusion from users about which class to use.
    • Note that under the hood there is a new essentially “lazy evaluation” scheme within respect to adding columns to DataFrame. During some operations, like-typed blocks will be “consolidated” but not before.
  • DataFrame accessing columns repeatedly is now significantly faster than DataMatrix used to be in 0.3.0 due to an internal Series caching mechanism (which are all views on the underlying data)
  • Column ordering for mixed type data is now completely consistent in DataFrame. In prior releases, there was inconsistent column ordering in DataMatrix
  • Improved console / string formatting of DataMatrix with negative numbers
  • Improved tabular data parsing functions, read_table and read_csv:
    • Added skiprows and na_values arguments to pandas.io.parsers functions for more flexible IO
    • parseCSV / read_csv functions and others in pandas.io.parsers now can take a list of custom NA values, and also a list of rows to skip
  • Can slice DataFrame and get a view of the data (when homogeneously typed), e.g. frame.xs(idx, copy=False) or frame.ix[idx]
  • Many speed optimizations throughout Series and DataFrame
  • Eager evaluation of groups when calling groupby functions, so if there is an exception with the grouping function it will raised immediately versus sometime later on when the groups are needed
  • datetools.WeekOfMonth offset can be parameterized with n different than 1 or -1.
  • Statistical methods on DataFrame like mean, std, var, skew will now ignore non-numerical data. Before a not very useful error message was generated. A flag numeric_only has been added to DataFrame.sum and DataFrame.count to enable this behavior in those methods if so desired (disabled by default)
  • DataFrame.pivot generalized to enable pivoting multiple columns into a DataFrame with hierarchical columns
  • DataFrame constructor can accept structured / record arrays
  • Panel constructor can accept a dict of DataFrame-like objects. Do not need to use from_dict anymore (from_dict is there to stay, though).

API Changes

  • The DataMatrix variable now refers to DataFrame, will be removed within two releases
  • WidePanel is now known as Panel. The WidePanel variable in the pandas namespace now refers to the renamed Panel class
  • LongPanel and Panel / WidePanel now no longer have a common subclass. LongPanel is now a subclass of DataFrame having a number of additional methods and a hierarchical index instead of the old LongPanelIndex object, which has been removed. Legacy LongPanel pickles may not load properly
  • Cython is now required to build pandas from a development branch. This was done to avoid continuing to check in cythonized C files into source control. Builds from released source distributions will not require Cython
  • Cython code has been moved up to a top level pandas/src directory. Cython extension modules have been renamed and promoted from the lib subpackage to the top level, i.e.
    • pandas.lib.tseries -> pandas._tseries
    • pandas.lib.sparse -> pandas._sparse
  • DataFrame pickling format has changed. Backwards compatibility for legacy pickles is provided, but it’s recommended to consider PyTables-based HDFStore for storing data with a longer expected shelf life
  • A copy argument has been added to the DataFrame constructor to avoid unnecessary copying of data. Data is no longer copied by default when passed into the constructor
  • Handling of boolean dtype in DataFrame has been improved to support storage of boolean data with NA / NaN values. Before it was being converted to float64 so this should not (in theory) cause API breakage
  • To optimize performance, Index objects now only check that their labels are unique when uniqueness matters (i.e. when someone goes to perform a lookup). This is a potentially dangerous tradeoff, but will lead to much better performance in many places (like groupby).
  • Boolean indexing using Series must now have the same indices (labels)
  • Backwards compatibility support for begin/end/nPeriods keyword arguments in DateRange class has been removed
  • More intuitive / shorter filling aliases ffill (for pad) and bfill (for backfill) have been added to the functions that use them: reindex, asfreq, fillna.
  • pandas.core.mixins code moved to pandas.core.generic
  • buffer keyword arguments (e.g. DataFrame.toString) renamed to buf to avoid using Python built-in name
  • DataFrame.rows() removed (use DataFrame.index)
  • Added deprecation warning to DataFrame.cols(), to be removed in next release
  • DataFrame deprecations and de-camelCasing: merge, asMatrix, toDataMatrix, _firstTimeWithValue, _lastTimeWithValue, toRecords, fromRecords, tgroupby, toString
  • pandas.io.parsers method deprecations
    • parseCSV is now read_csv and keyword arguments have been de-camelCased
    • parseText is now read_table
    • parseExcel is replaced by the ExcelFile class and its parse method
  • fillMethod arguments (deprecated in prior release) removed, should be replaced with method
  • Series.fill, DataFrame.fill, and Panel.fill removed, use fillna instead
  • groupby functions now exclude NA / NaN values from the list of groups. This matches R behavior with NAs in factors e.g. with the tapply function
  • Removed parseText, parseCSV and parseExcel from pandas namespace
  • Series.combineFunc renamed to Series.combine and made a bit more general with a fill_value keyword argument defaulting to NaN
  • Removed pandas.core.pytools module. Code has been moved to pandas.core.common
  • Tacked on groupName attribute for groups in GroupBy renamed to name
  • Panel/LongPanel dims attribute renamed to shape to be more conformant
  • Slicing a Series returns a view now
  • More Series deprecations / renaming: toCSV to to_csv, asOf to asof, merge to map, applymap to apply, toDict to to_dict, combineFirst to combine_first. Will print FutureWarning.
  • DataFrame.to_csv does not write an “index” column label by default anymore since the output file can be read back without it. However, there is a new index_label argument. So you can do index_label='index' to emulate the old behavior
  • datetools.Week argument renamed from dayOfWeek to weekday
  • timeRule argument in shift has been deprecated in favor of using the offset argument for everything. So you can still pass a time rule string to offset
  • Added optional encoding argument to read_csv, read_table, to_csv, from_csv to handle unicode in python 2.x

Bug Fixes

  • Column ordering in pandas.io.parsers.parseCSV will match CSV in the presence of mixed-type data
  • Fixed handling of Excel 2003 dates in pandas.io.parsers
  • DateRange caching was happening with high resolution DateOffset objects, e.g. DateOffset(seconds=1). This has been fixed
  • Fixed __truediv__ issue in DataFrame
  • Fixed DataFrame.toCSV bug preventing IO round trips in some cases
  • Fixed bug in Series.plot causing matplotlib to barf in exceptional cases
  • Disabled Index objects from being hashable, like ndarrays
  • Added __ne__ implementation to Index so that operations like ts[ts != idx] will work
  • Added __ne__ implementation to DataFrame
  • Bug / unintuitive result when calling fillna on unordered labels
  • Bug calling sum on boolean DataFrame
  • Bug fix when creating a DataFrame from a dict with scalar values
  • Series.{sum, mean, std, ...} now return NA/NaN when the whole Series is NA
  • NumPy 1.4 through 1.6 compatibility fixes
  • Fixed bug in bias correction in rolling_cov, was affecting rolling_corr too
  • R-square value was incorrect in the presence of fixed and time effects in the PanelOLS classes
  • HDFStore can handle duplicates in table format, will take

Thanks

  • Joon Ro
  • Michael Pennington
  • Chris Uga
  • Chris Withers
  • Jeff Reback
  • Ted Square
  • Craig Austin
  • William Ferreira
  • Daniel Fortunov
  • Tony Roberts
  • Martin Felder
  • John Marino
  • Tim McNamara
  • Justin Berka
  • Dieter Vandenbussche
  • Shane Conway
  • Skipper Seabold
  • Chris Jordan-Squire

pandas 0.3.0

Release date: February 20, 2011

New features

  • corrwith function to compute column- or row-wise correlations between two DataFrame objects
  • Can boolean-index DataFrame objects, e.g. df[df > 2] = 2, px[px > last_px] = 0
  • Added comparison magic methods (__lt__, __gt__, etc.)
  • Flexible explicit arithmetic methods (add, mul, sub, div, etc.)
  • Added reindex_like method
  • Added reindex_like method to WidePanel
  • Convenience functions for accessing SQL-like databases in pandas.io.sql module
  • Added (still experimental) HDFStore class for storing pandas data structures using HDF5 / PyTables in pandas.io.pytables module
  • Added WeekOfMonth date offset
  • pandas.rpy (experimental) module created, provide some interfacing / conversion between rpy2 and pandas

Improvements to existing features

  • Unit test coverage: 100% line coverage of core data structures
  • Speed enhancement to rolling_{median, max, min}
  • Column ordering between DataFrame and DataMatrix is now consistent: before DataFrame would not respect column order
  • Improved {Series, DataFrame}.plot methods to be more flexible (can pass matplotlib Axis arguments, plot DataFrame columns in multiple subplots, etc.)

API Changes

  • Exponentially-weighted moment functions in pandas.stats.moments have a more consistent API and accept a min_periods argument like their regular moving counterparts.
  • fillMethod argument in Series, DataFrame changed to method, FutureWarning added.
  • fill method in Series, DataFrame/DataMatrix, WidePanel renamed to fillna, FutureWarning added to fill
  • Renamed DataFrame.getXS to xs, FutureWarning added
  • Removed cap and floor functions from DataFrame, renamed to clip_upper and clip_lower for consistency with NumPy

Bug Fixes

  • Fixed bug in IndexableSkiplist Cython code that was breaking rolling_max function
  • Numerous numpy.int64-related indexing fixes
  • Several NumPy 1.4.0 NaN-handling fixes
  • Bug fixes to pandas.io.parsers.parseCSV
  • Fixed DateRange caching issue with unusual date offsets
  • Fixed bug in DateRange.union
  • Fixed corner case in IndexableSkiplist implementation