pandas contains extensive capabilities and features for working with time series data for all domains. Using the NumPy datetime64 and timedelta64 dtypes, pandas has consolidated a large number of features from other Python libraries like scikits.timeseries as well as created a tremendous amount of new functionality for manipulating time series data.
datetime64
timedelta64
scikits.timeseries
For example, pandas supports:
Parsing time series information from various sources and formats
In [1]: import datetime In [2]: dti = pd.to_datetime( ...: ["1/1/2018", np.datetime64("2018-01-01"), datetime.datetime(2018, 1, 1)] ...: ) ...: In [3]: dti Out[3]: DatetimeIndex(['2018-01-01', '2018-01-01', '2018-01-01'], dtype='datetime64[ns]', freq=None)
Generate sequences of fixed-frequency dates and time spans
In [4]: dti = pd.date_range("2018-01-01", periods=3, freq="H") In [5]: dti Out[5]: DatetimeIndex(['2018-01-01 00:00:00', '2018-01-01 01:00:00', '2018-01-01 02:00:00'], dtype='datetime64[ns]', freq='H')
Manipulating and converting date times with timezone information
In [6]: dti = dti.tz_localize("UTC") In [7]: dti Out[7]: DatetimeIndex(['2018-01-01 00:00:00+00:00', '2018-01-01 01:00:00+00:00', '2018-01-01 02:00:00+00:00'], dtype='datetime64[ns, UTC]', freq='H') In [8]: dti.tz_convert("US/Pacific") Out[8]: DatetimeIndex(['2017-12-31 16:00:00-08:00', '2017-12-31 17:00:00-08:00', '2017-12-31 18:00:00-08:00'], dtype='datetime64[ns, US/Pacific]', freq='H')
Resampling or converting a time series to a particular frequency
In [9]: idx = pd.date_range("2018-01-01", periods=5, freq="H") In [10]: ts = pd.Series(range(len(idx)), index=idx) In [11]: ts Out[11]: 2018-01-01 00:00:00 0 2018-01-01 01:00:00 1 2018-01-01 02:00:00 2 2018-01-01 03:00:00 3 2018-01-01 04:00:00 4 Freq: H, dtype: int64 In [12]: ts.resample("2H").mean() Out[12]: 2018-01-01 00:00:00 0.5 2018-01-01 02:00:00 2.5 2018-01-01 04:00:00 4.0 Freq: 2H, dtype: float64
Performing date and time arithmetic with absolute or relative time increments
In [13]: friday = pd.Timestamp("2018-01-05") In [14]: friday.day_name() Out[14]: 'Friday' # Add 1 day In [15]: saturday = friday + pd.Timedelta("1 day") In [16]: saturday.day_name() Out[16]: 'Saturday' # Add 1 business day (Friday --> Monday) In [17]: monday = friday + pd.offsets.BDay() In [18]: monday.day_name() Out[18]: 'Monday'
pandas provides a relatively compact and self-contained set of tools for performing the above tasks and more.
pandas captures 4 general time related concepts:
Date times: A specific date and time with timezone support. Similar to datetime.datetime from the standard library.
datetime.datetime
Time deltas: An absolute time duration. Similar to datetime.timedelta from the standard library.
datetime.timedelta
Time spans: A span of time defined by a point in time and its associated frequency.
Date offsets: A relative time duration that respects calendar arithmetic. Similar to dateutil.relativedelta.relativedelta from the dateutil package.
dateutil.relativedelta.relativedelta
dateutil
Concept
Scalar Class
Array Class
pandas Data Type
Primary Creation Method
Date times
Timestamp
DatetimeIndex
datetime64[ns] or datetime64[ns, tz]
datetime64[ns]
datetime64[ns, tz]
to_datetime or date_range
to_datetime
date_range
Time deltas
Timedelta
TimedeltaIndex
timedelta64[ns]
to_timedelta or timedelta_range
to_timedelta
timedelta_range
Time spans
Period
PeriodIndex
period[freq]
Period or period_range
period_range
Date offsets
DateOffset
None
For time series data, it’s conventional to represent the time component in the index of a Series or DataFrame so manipulations can be performed with respect to the time element.
Series
DataFrame
In [19]: pd.Series(range(3), index=pd.date_range("2000", freq="D", periods=3)) Out[19]: 2000-01-01 0 2000-01-02 1 2000-01-03 2 Freq: D, dtype: int64
However, Series and DataFrame can directly also support the time component as data itself.
In [20]: pd.Series(pd.date_range("2000", freq="D", periods=3)) Out[20]: 0 2000-01-01 1 2000-01-02 2 2000-01-03 dtype: datetime64[ns]
Series and DataFrame have extended data type support and functionality for datetime, timedelta and Period data when passed into those constructors. DateOffset data however will be stored as object data.
datetime
timedelta
object
In [21]: pd.Series(pd.period_range("1/1/2011", freq="M", periods=3)) Out[21]: 0 2011-01 1 2011-02 2 2011-03 dtype: period[M] In [22]: pd.Series([pd.DateOffset(1), pd.DateOffset(2)]) Out[22]: 0 <DateOffset> 1 <2 * DateOffsets> dtype: object In [23]: pd.Series(pd.date_range("1/1/2011", freq="M", periods=3)) Out[23]: 0 2011-01-31 1 2011-02-28 2 2011-03-31 dtype: datetime64[ns]
Lastly, pandas represents null date times, time deltas, and time spans as NaT which is useful for representing missing or null date like values and behaves similar as np.nan does for float data.
NaT
np.nan
In [24]: pd.Timestamp(pd.NaT) Out[24]: NaT In [25]: pd.Timedelta(pd.NaT) Out[25]: NaT In [26]: pd.Period(pd.NaT) Out[26]: NaT # Equality acts as np.nan would In [27]: pd.NaT == pd.NaT Out[27]: False
Timestamped data is the most basic type of time series data that associates values with points in time. For pandas objects it means using the points in time.
In [28]: pd.Timestamp(datetime.datetime(2012, 5, 1)) Out[28]: Timestamp('2012-05-01 00:00:00') In [29]: pd.Timestamp("2012-05-01") Out[29]: Timestamp('2012-05-01 00:00:00') In [30]: pd.Timestamp(2012, 5, 1) Out[30]: Timestamp('2012-05-01 00:00:00')
However, in many cases it is more natural to associate things like change variables with a time span instead. The span represented by Period can be specified explicitly, or inferred from datetime string format.
For example:
In [31]: pd.Period("2011-01") Out[31]: Period('2011-01', 'M') In [32]: pd.Period("2012-05", freq="D") Out[32]: Period('2012-05-01', 'D')
Timestamp and Period can serve as an index. Lists of Timestamp and Period are automatically coerced to DatetimeIndex and PeriodIndex respectively.
In [33]: dates = [ ....: pd.Timestamp("2012-05-01"), ....: pd.Timestamp("2012-05-02"), ....: pd.Timestamp("2012-05-03"), ....: ] ....: In [34]: ts = pd.Series(np.random.randn(3), dates) In [35]: type(ts.index) Out[35]: pandas.core.indexes.datetimes.DatetimeIndex In [36]: ts.index Out[36]: DatetimeIndex(['2012-05-01', '2012-05-02', '2012-05-03'], dtype='datetime64[ns]', freq=None) In [37]: ts Out[37]: 2012-05-01 0.469112 2012-05-02 -0.282863 2012-05-03 -1.509059 dtype: float64 In [38]: periods = [pd.Period("2012-01"), pd.Period("2012-02"), pd.Period("2012-03")] In [39]: ts = pd.Series(np.random.randn(3), periods) In [40]: type(ts.index) Out[40]: pandas.core.indexes.period.PeriodIndex In [41]: ts.index Out[41]: PeriodIndex(['2012-01', '2012-02', '2012-03'], dtype='period[M]', freq='M') In [42]: ts Out[42]: 2012-01 -1.135632 2012-02 1.212112 2012-03 -0.173215 Freq: M, dtype: float64
pandas allows you to capture both representations and convert between them. Under the hood, pandas represents timestamps using instances of Timestamp and sequences of timestamps using instances of DatetimeIndex. For regular time spans, pandas uses Period objects for scalar values and PeriodIndex for sequences of spans. Better support for irregular intervals with arbitrary start and end points are forth-coming in future releases.
To convert a Series or list-like object of date-like objects e.g. strings, epochs, or a mixture, you can use the to_datetime function. When passed a Series, this returns a Series (with the same index), while a list-like is converted to a DatetimeIndex:
In [43]: pd.to_datetime(pd.Series(["Jul 31, 2009", "2010-01-10", None])) Out[43]: 0 2009-07-31 1 2010-01-10 2 NaT dtype: datetime64[ns] In [44]: pd.to_datetime(["2005/11/23", "2010.12.31"]) Out[44]: DatetimeIndex(['2005-11-23', '2010-12-31'], dtype='datetime64[ns]', freq=None)
If you use dates which start with the day first (i.e. European style), you can pass the dayfirst flag:
dayfirst
In [45]: pd.to_datetime(["04-01-2012 10:00"], dayfirst=True) Out[45]: DatetimeIndex(['2012-01-04 10:00:00'], dtype='datetime64[ns]', freq=None) In [46]: pd.to_datetime(["14-01-2012", "01-14-2012"], dayfirst=True) Out[46]: DatetimeIndex(['2012-01-14', '2012-01-14'], dtype='datetime64[ns]', freq=None)
Warning
You see in the above example that dayfirst isn’t strict, so if a date can’t be parsed with the day being first it will be parsed as if dayfirst were False.
If you pass a single string to to_datetime, it returns a single Timestamp. Timestamp can also accept string input, but it doesn’t accept string parsing options like dayfirst or format, so use to_datetime if these are required.
format
In [47]: pd.to_datetime("2010/11/12") Out[47]: Timestamp('2010-11-12 00:00:00') In [48]: pd.Timestamp("2010/11/12") Out[48]: Timestamp('2010-11-12 00:00:00')
You can also use the DatetimeIndex constructor directly:
In [49]: pd.DatetimeIndex(["2018-01-01", "2018-01-03", "2018-01-05"]) Out[49]: DatetimeIndex(['2018-01-01', '2018-01-03', '2018-01-05'], dtype='datetime64[ns]', freq=None)
The string ‘infer’ can be passed in order to set the frequency of the index as the inferred frequency upon creation:
In [50]: pd.DatetimeIndex(["2018-01-01", "2018-01-03", "2018-01-05"], freq="infer") Out[50]: DatetimeIndex(['2018-01-01', '2018-01-03', '2018-01-05'], dtype='datetime64[ns]', freq='2D')
In addition to the required datetime string, a format argument can be passed to ensure specific parsing. This could also potentially speed up the conversion considerably.
In [51]: pd.to_datetime("2010/11/12", format="%Y/%m/%d") Out[51]: Timestamp('2010-11-12 00:00:00') In [52]: pd.to_datetime("12-11-2010 00:00", format="%d-%m-%Y %H:%M") Out[52]: Timestamp('2010-11-12 00:00:00')
For more information on the choices available when specifying the format option, see the Python datetime documentation.
You can also pass a DataFrame of integer or string columns to assemble into a Series of Timestamps.
Timestamps
In [53]: df = pd.DataFrame( ....: {"year": [2015, 2016], "month": [2, 3], "day": [4, 5], "hour": [2, 3]} ....: ) ....: In [54]: pd.to_datetime(df) Out[54]: 0 2015-02-04 02:00:00 1 2016-03-05 03:00:00 dtype: datetime64[ns]
You can pass only the columns that you need to assemble.
In [55]: pd.to_datetime(df[["year", "month", "day"]]) Out[55]: 0 2015-02-04 1 2016-03-05 dtype: datetime64[ns]
pd.to_datetime looks for standard designations of the datetime component in the column names, including:
pd.to_datetime
required: year, month, day
year
month
day
optional: hour, minute, second, millisecond, microsecond, nanosecond
hour
minute
second
millisecond
microsecond
nanosecond
The default behavior, errors='raise', is to raise when unparsable:
errors='raise'
In [2]: pd.to_datetime(['2009/07/31', 'asd'], errors='raise') ValueError: Unknown string format
Pass errors='ignore' to return the original input when unparsable:
errors='ignore'
In [56]: pd.to_datetime(["2009/07/31", "asd"], errors="ignore") Out[56]: Index(['2009/07/31', 'asd'], dtype='object')
Pass errors='coerce' to convert unparsable data to NaT (not a time):
errors='coerce'
In [57]: pd.to_datetime(["2009/07/31", "asd"], errors="coerce") Out[57]: DatetimeIndex(['2009-07-31', 'NaT'], dtype='datetime64[ns]', freq=None)
pandas supports converting integer or float epoch times to Timestamp and DatetimeIndex. The default unit is nanoseconds, since that is how Timestamp objects are stored internally. However, epochs are often stored in another unit which can be specified. These are computed from the starting point specified by the origin parameter.
unit
origin
In [58]: pd.to_datetime( ....: [1349720105, 1349806505, 1349892905, 1349979305, 1350065705], unit="s" ....: ) ....: Out[58]: DatetimeIndex(['2012-10-08 18:15:05', '2012-10-09 18:15:05', '2012-10-10 18:15:05', '2012-10-11 18:15:05', '2012-10-12 18:15:05'], dtype='datetime64[ns]', freq=None) In [59]: pd.to_datetime( ....: [1349720105100, 1349720105200, 1349720105300, 1349720105400, 1349720105500], ....: unit="ms", ....: ) ....: Out[59]: DatetimeIndex(['2012-10-08 18:15:05.100000', '2012-10-08 18:15:05.200000', '2012-10-08 18:15:05.300000', '2012-10-08 18:15:05.400000', '2012-10-08 18:15:05.500000'], dtype='datetime64[ns]', freq=None)
Note
The unit parameter does not use the same strings as the format parameter that was discussed above). The available units are listed on the documentation for pandas.to_datetime().
pandas.to_datetime()
Changed in version 1.0.0.
Constructing a Timestamp or DatetimeIndex with an epoch timestamp with the tz argument specified will raise a ValueError. If you have epochs in wall time in another timezone, you can read the epochs as timezone-naive timestamps and then localize to the appropriate timezone:
tz
In [60]: pd.Timestamp(1262347200000000000).tz_localize("US/Pacific") Out[60]: Timestamp('2010-01-01 12:00:00-0800', tz='US/Pacific') In [61]: pd.DatetimeIndex([1262347200000000000]).tz_localize("US/Pacific") Out[61]: DatetimeIndex(['2010-01-01 12:00:00-08:00'], dtype='datetime64[ns, US/Pacific]', freq=None)
Epoch times will be rounded to the nearest nanosecond.
Conversion of float epoch times can lead to inaccurate and unexpected results. Python floats have about 15 digits precision in decimal. Rounding during conversion from float to high precision Timestamp is unavoidable. The only way to achieve exact precision is to use a fixed-width types (e.g. an int64).
In [62]: pd.to_datetime([1490195805.433, 1490195805.433502912], unit="s") Out[62]: DatetimeIndex(['2017-03-22 15:16:45.433000088', '2017-03-22 15:16:45.433502913'], dtype='datetime64[ns]', freq=None) In [63]: pd.to_datetime(1490195805433502912, unit="ns") Out[63]: Timestamp('2017-03-22 15:16:45.433502912')
See also
Using the origin Parameter
To invert the operation from above, namely, to convert from a Timestamp to a ‘unix’ epoch:
In [64]: stamps = pd.date_range("2012-10-08 18:15:05", periods=4, freq="D") In [65]: stamps Out[65]: DatetimeIndex(['2012-10-08 18:15:05', '2012-10-09 18:15:05', '2012-10-10 18:15:05', '2012-10-11 18:15:05'], dtype='datetime64[ns]', freq='D')
We subtract the epoch (midnight at January 1, 1970 UTC) and then floor divide by the “unit” (1 second).
In [66]: (stamps - pd.Timestamp("1970-01-01")) // pd.Timedelta("1s") Out[66]: Int64Index([1349720105, 1349806505, 1349892905, 1349979305], dtype='int64')
Using the origin parameter, one can specify an alternative starting point for creation of a DatetimeIndex. For example, to use 1960-01-01 as the starting date:
In [67]: pd.to_datetime([1, 2, 3], unit="D", origin=pd.Timestamp("1960-01-01")) Out[67]: DatetimeIndex(['1960-01-02', '1960-01-03', '1960-01-04'], dtype='datetime64[ns]', freq=None)
The default is set at origin='unix', which defaults to 1970-01-01 00:00:00. Commonly called ‘unix epoch’ or POSIX time.
origin='unix'
1970-01-01 00:00:00
In [68]: pd.to_datetime([1, 2, 3], unit="D") Out[68]: DatetimeIndex(['1970-01-02', '1970-01-03', '1970-01-04'], dtype='datetime64[ns]', freq=None)
To generate an index with timestamps, you can use either the DatetimeIndex or Index constructor and pass in a list of datetime objects:
Index
In [69]: dates = [ ....: datetime.datetime(2012, 5, 1), ....: datetime.datetime(2012, 5, 2), ....: datetime.datetime(2012, 5, 3), ....: ] ....: # Note the frequency information In [70]: index = pd.DatetimeIndex(dates) In [71]: index Out[71]: DatetimeIndex(['2012-05-01', '2012-05-02', '2012-05-03'], dtype='datetime64[ns]', freq=None) # Automatically converted to DatetimeIndex In [72]: index = pd.Index(dates) In [73]: index Out[73]: DatetimeIndex(['2012-05-01', '2012-05-02', '2012-05-03'], dtype='datetime64[ns]', freq=None)
In practice this becomes very cumbersome because we often need a very long index with a large number of timestamps. If we need timestamps on a regular frequency, we can use the date_range() and bdate_range() functions to create a DatetimeIndex. The default frequency for date_range is a calendar day while the default for bdate_range is a business day:
date_range()
bdate_range()
bdate_range
In [74]: start = datetime.datetime(2011, 1, 1) In [75]: end = datetime.datetime(2012, 1, 1) In [76]: index = pd.date_range(start, end) In [77]: index Out[77]: DatetimeIndex(['2011-01-01', '2011-01-02', '2011-01-03', '2011-01-04', '2011-01-05', '2011-01-06', '2011-01-07', '2011-01-08', '2011-01-09', '2011-01-10', ... '2011-12-23', '2011-12-24', '2011-12-25', '2011-12-26', '2011-12-27', '2011-12-28', '2011-12-29', '2011-12-30', '2011-12-31', '2012-01-01'], dtype='datetime64[ns]', length=366, freq='D') In [78]: index = pd.bdate_range(start, end) In [79]: index Out[79]: DatetimeIndex(['2011-01-03', '2011-01-04', '2011-01-05', '2011-01-06', '2011-01-07', '2011-01-10', '2011-01-11', '2011-01-12', '2011-01-13', '2011-01-14', ... '2011-12-19', '2011-12-20', '2011-12-21', '2011-12-22', '2011-12-23', '2011-12-26', '2011-12-27', '2011-12-28', '2011-12-29', '2011-12-30'], dtype='datetime64[ns]', length=260, freq='B')
Convenience functions like date_range and bdate_range can utilize a variety of frequency aliases:
In [80]: pd.date_range(start, periods=1000, freq="M") Out[80]: DatetimeIndex(['2011-01-31', '2011-02-28', '2011-03-31', '2011-04-30', '2011-05-31', '2011-06-30', '2011-07-31', '2011-08-31', '2011-09-30', '2011-10-31', ... '2093-07-31', '2093-08-31', '2093-09-30', '2093-10-31', '2093-11-30', '2093-12-31', '2094-01-31', '2094-02-28', '2094-03-31', '2094-04-30'], dtype='datetime64[ns]', length=1000, freq='M') In [81]: pd.bdate_range(start, periods=250, freq="BQS") Out[81]: DatetimeIndex(['2011-01-03', '2011-04-01', '2011-07-01', '2011-10-03', '2012-01-02', '2012-04-02', '2012-07-02', '2012-10-01', '2013-01-01', '2013-04-01', ... '2071-01-01', '2071-04-01', '2071-07-01', '2071-10-01', '2072-01-01', '2072-04-01', '2072-07-01', '2072-10-03', '2073-01-02', '2073-04-03'], dtype='datetime64[ns]', length=250, freq='BQS-JAN')
date_range and bdate_range make it easy to generate a range of dates using various combinations of parameters like start, end, periods, and freq. The start and end dates are strictly inclusive, so dates outside of those specified will not be generated:
start
end
periods
freq
In [82]: pd.date_range(start, end, freq="BM") Out[82]: DatetimeIndex(['2011-01-31', '2011-02-28', '2011-03-31', '2011-04-29', '2011-05-31', '2011-06-30', '2011-07-29', '2011-08-31', '2011-09-30', '2011-10-31', '2011-11-30', '2011-12-30'], dtype='datetime64[ns]', freq='BM') In [83]: pd.date_range(start, end, freq="W") Out[83]: DatetimeIndex(['2011-01-02', '2011-01-09', '2011-01-16', '2011-01-23', '2011-01-30', '2011-02-06', '2011-02-13', '2011-02-20', '2011-02-27', '2011-03-06', '2011-03-13', '2011-03-20', '2011-03-27', '2011-04-03', '2011-04-10', '2011-04-17', '2011-04-24', '2011-05-01', '2011-05-08', '2011-05-15', '2011-05-22', '2011-05-29', '2011-06-05', '2011-06-12', '2011-06-19', '2011-06-26', '2011-07-03', '2011-07-10', '2011-07-17', '2011-07-24', '2011-07-31', '2011-08-07', '2011-08-14', '2011-08-21', '2011-08-28', '2011-09-04', '2011-09-11', '2011-09-18', '2011-09-25', '2011-10-02', '2011-10-09', '2011-10-16', '2011-10-23', '2011-10-30', '2011-11-06', '2011-11-13', '2011-11-20', '2011-11-27', '2011-12-04', '2011-12-11', '2011-12-18', '2011-12-25', '2012-01-01'], dtype='datetime64[ns]', freq='W-SUN') In [84]: pd.bdate_range(end=end, periods=20) Out[84]: DatetimeIndex(['2011-12-05', '2011-12-06', '2011-12-07', '2011-12-08', '2011-12-09', '2011-12-12', '2011-12-13', '2011-12-14', '2011-12-15', '2011-12-16', '2011-12-19', '2011-12-20', '2011-12-21', '2011-12-22', '2011-12-23', '2011-12-26', '2011-12-27', '2011-12-28', '2011-12-29', '2011-12-30'], dtype='datetime64[ns]', freq='B') In [85]: pd.bdate_range(start=start, periods=20) Out[85]: DatetimeIndex(['2011-01-03', '2011-01-04', '2011-01-05', '2011-01-06', '2011-01-07', '2011-01-10', '2011-01-11', '2011-01-12', '2011-01-13', '2011-01-14', '2011-01-17', '2011-01-18', '2011-01-19', '2011-01-20', '2011-01-21', '2011-01-24', '2011-01-25', '2011-01-26', '2011-01-27', '2011-01-28'], dtype='datetime64[ns]', freq='B')
Specifying start, end, and periods will generate a range of evenly spaced dates from start to end inclusively, with periods number of elements in the resulting DatetimeIndex:
In [86]: pd.date_range("2018-01-01", "2018-01-05", periods=5) Out[86]: DatetimeIndex(['2018-01-01', '2018-01-02', '2018-01-03', '2018-01-04', '2018-01-05'], dtype='datetime64[ns]', freq=None) In [87]: pd.date_range("2018-01-01", "2018-01-05", periods=10) Out[87]: DatetimeIndex(['2018-01-01 00:00:00', '2018-01-01 10:40:00', '2018-01-01 21:20:00', '2018-01-02 08:00:00', '2018-01-02 18:40:00', '2018-01-03 05:20:00', '2018-01-03 16:00:00', '2018-01-04 02:40:00', '2018-01-04 13:20:00', '2018-01-05 00:00:00'], dtype='datetime64[ns]', freq=None)
bdate_range can also generate a range of custom frequency dates by using the weekmask and holidays parameters. These parameters will only be used if a custom frequency string is passed.
weekmask
holidays
In [88]: weekmask = "Mon Wed Fri" In [89]: holidays = [datetime.datetime(2011, 1, 5), datetime.datetime(2011, 3, 14)] In [90]: pd.bdate_range(start, end, freq="C", weekmask=weekmask, holidays=holidays) Out[90]: DatetimeIndex(['2011-01-03', '2011-01-07', '2011-01-10', '2011-01-12', '2011-01-14', '2011-01-17', '2011-01-19', '2011-01-21', '2011-01-24', '2011-01-26', ... '2011-12-09', '2011-12-12', '2011-12-14', '2011-12-16', '2011-12-19', '2011-12-21', '2011-12-23', '2011-12-26', '2011-12-28', '2011-12-30'], dtype='datetime64[ns]', length=154, freq='C') In [91]: pd.bdate_range(start, end, freq="CBMS", weekmask=weekmask) Out[91]: DatetimeIndex(['2011-01-03', '2011-02-02', '2011-03-02', '2011-04-01', '2011-05-02', '2011-06-01', '2011-07-01', '2011-08-01', '2011-09-02', '2011-10-03', '2011-11-02', '2011-12-02'], dtype='datetime64[ns]', freq='CBMS')
Custom business days
Since pandas represents timestamps in nanosecond resolution, the time span that can be represented using a 64-bit integer is limited to approximately 584 years:
In [92]: pd.Timestamp.min Out[92]: Timestamp('1677-09-21 00:12:43.145225') In [93]: pd.Timestamp.max Out[93]: Timestamp('2262-04-11 23:47:16.854775807')
Representing out-of-bounds spans
One of the main uses for DatetimeIndex is as an index for pandas objects. The DatetimeIndex class contains many time series related optimizations:
A large range of dates for various offsets are pre-computed and cached under the hood in order to make generating subsequent date ranges very fast (just have to grab a slice).
Fast shifting using the shift method on pandas objects.
shift
Unioning of overlapping DatetimeIndex objects with the same frequency is very fast (important for fast data alignment).
Quick access to date fields via properties such as year, month, etc.
Regularization functions like snap and very fast asof logic.
snap
asof
DatetimeIndex objects have all the basic functionality of regular Index objects, and a smorgasbord of advanced time series specific methods for easy frequency processing.
Reindexing methods
While pandas does not force you to have a sorted date index, some of these methods may have unexpected or incorrect behavior if the dates are unsorted.
DatetimeIndex can be used like a regular index and offers all of its intelligent functionality like selection, slicing, etc.
In [94]: rng = pd.date_range(start, end, freq="BM") In [95]: ts = pd.Series(np.random.randn(len(rng)), index=rng) In [96]: ts.index Out[96]: DatetimeIndex(['2011-01-31', '2011-02-28', '2011-03-31', '2011-04-29', '2011-05-31', '2011-06-30', '2011-07-29', '2011-08-31', '2011-09-30', '2011-10-31', '2011-11-30', '2011-12-30'], dtype='datetime64[ns]', freq='BM') In [97]: ts[:5].index Out[97]: DatetimeIndex(['2011-01-31', '2011-02-28', '2011-03-31', '2011-04-29', '2011-05-31'], dtype='datetime64[ns]', freq='BM') In [98]: ts[::2].index Out[98]: DatetimeIndex(['2011-01-31', '2011-03-31', '2011-05-31', '2011-07-29', '2011-09-30', '2011-11-30'], dtype='datetime64[ns]', freq='2BM')
Dates and strings that parse to timestamps can be passed as indexing parameters:
In [99]: ts["1/31/2011"] Out[99]: 0.11920871129693428 In [100]: ts[datetime.datetime(2011, 12, 25):] Out[100]: 2011-12-30 0.56702 Freq: BM, dtype: float64 In [101]: ts["10/31/2011":"12/31/2011"] Out[101]: 2011-10-31 0.271860 2011-11-30 -0.424972 2011-12-30 0.567020 Freq: BM, dtype: float64
To provide convenience for accessing longer time series, you can also pass in the year or year and month as strings:
In [102]: ts["2011"] Out[102]: 2011-01-31 0.119209 2011-02-28 -1.044236 2011-03-31 -0.861849 2011-04-29 -2.104569 2011-05-31 -0.494929 2011-06-30 1.071804 2011-07-29 0.721555 2011-08-31 -0.706771 2011-09-30 -1.039575 2011-10-31 0.271860 2011-11-30 -0.424972 2011-12-30 0.567020 Freq: BM, dtype: float64 In [103]: ts["2011-6"] Out[103]: 2011-06-30 1.071804 Freq: BM, dtype: float64
This type of slicing will work on a DataFrame with a DatetimeIndex as well. Since the partial string selection is a form of label slicing, the endpoints will be included. This would include matching times on an included date:
Indexing DataFrame rows with a single string with getitem (e.g. frame[dtstring]) is deprecated starting with pandas 1.2.0 (given the ambiguity whether it is indexing the rows or selecting a column) and will be removed in a future version. The equivalent with .loc (e.g. frame.loc[dtstring]) is still supported.
frame[dtstring]
.loc
frame.loc[dtstring]
In [104]: dft = pd.DataFrame( .....: np.random.randn(100000, 1), .....: columns=["A"], .....: index=pd.date_range("20130101", periods=100000, freq="T"), .....: ) .....: In [105]: dft Out[105]: A 2013-01-01 00:00:00 0.276232 2013-01-01 00:01:00 -1.087401 2013-01-01 00:02:00 -0.673690 2013-01-01 00:03:00 0.113648 2013-01-01 00:04:00 -1.478427 ... ... 2013-03-11 10:35:00 -0.747967 2013-03-11 10:36:00 -0.034523 2013-03-11 10:37:00 -0.201754 2013-03-11 10:38:00 -1.509067 2013-03-11 10:39:00 -1.693043 [100000 rows x 1 columns] In [106]: dft.loc["2013"] Out[106]: A 2013-01-01 00:00:00 0.276232 2013-01-01 00:01:00 -1.087401 2013-01-01 00:02:00 -0.673690 2013-01-01 00:03:00 0.113648 2013-01-01 00:04:00 -1.478427 ... ... 2013-03-11 10:35:00 -0.747967 2013-03-11 10:36:00 -0.034523 2013-03-11 10:37:00 -0.201754 2013-03-11 10:38:00 -1.509067 2013-03-11 10:39:00 -1.693043 [100000 rows x 1 columns]
This starts on the very first time in the month, and includes the last date and time for the month:
In [107]: dft["2013-1":"2013-2"] Out[107]: A 2013-01-01 00:00:00 0.276232 2013-01-01 00:01:00 -1.087401 2013-01-01 00:02:00 -0.673690 2013-01-01 00:03:00 0.113648 2013-01-01 00:04:00 -1.478427 ... ... 2013-02-28 23:55:00 0.850929 2013-02-28 23:56:00 0.976712 2013-02-28 23:57:00 -2.693884 2013-02-28 23:58:00 -1.575535 2013-02-28 23:59:00 -1.573517 [84960 rows x 1 columns]
This specifies a stop time that includes all of the times on the last day:
In [108]: dft["2013-1":"2013-2-28"] Out[108]: A 2013-01-01 00:00:00 0.276232 2013-01-01 00:01:00 -1.087401 2013-01-01 00:02:00 -0.673690 2013-01-01 00:03:00 0.113648 2013-01-01 00:04:00 -1.478427 ... ... 2013-02-28 23:55:00 0.850929 2013-02-28 23:56:00 0.976712 2013-02-28 23:57:00 -2.693884 2013-02-28 23:58:00 -1.575535 2013-02-28 23:59:00 -1.573517 [84960 rows x 1 columns]
This specifies an exact stop time (and is not the same as the above):
In [109]: dft["2013-1":"2013-2-28 00:00:00"] Out[109]: A 2013-01-01 00:00:00 0.276232 2013-01-01 00:01:00 -1.087401 2013-01-01 00:02:00 -0.673690 2013-01-01 00:03:00 0.113648 2013-01-01 00:04:00 -1.478427 ... ... 2013-02-27 23:56:00 1.197749 2013-02-27 23:57:00 0.720521 2013-02-27 23:58:00 -0.072718 2013-02-27 23:59:00 -0.681192 2013-02-28 00:00:00 -0.557501 [83521 rows x 1 columns]
We are stopping on the included end-point as it is part of the index:
In [110]: dft["2013-1-15":"2013-1-15 12:30:00"] Out[110]: A 2013-01-15 00:00:00 -0.984810 2013-01-15 00:01:00 0.941451 2013-01-15 00:02:00 1.559365 2013-01-15 00:03:00 1.034374 2013-01-15 00:04:00 -1.480656 ... ... 2013-01-15 12:26:00 0.371454 2013-01-15 12:27:00 -0.930806 2013-01-15 12:28:00 -0.069177 2013-01-15 12:29:00 0.066510 2013-01-15 12:30:00 -0.003945 [751 rows x 1 columns]
DatetimeIndex partial string indexing also works on a DataFrame with a MultiIndex:
MultiIndex
In [111]: dft2 = pd.DataFrame( .....: np.random.randn(20, 1), .....: columns=["A"], .....: index=pd.MultiIndex.from_product( .....: [pd.date_range("20130101", periods=10, freq="12H"), ["a", "b"]] .....: ), .....: ) .....: In [112]: dft2 Out[112]: A 2013-01-01 00:00:00 a -0.298694 b 0.823553 2013-01-01 12:00:00 a 0.943285 b -1.479399 2013-01-02 00:00:00 a -1.643342 ... ... 2013-01-04 12:00:00 b 0.069036 2013-01-05 00:00:00 a 0.122297 b 1.422060 2013-01-05 12:00:00 a 0.370079 b 1.016331 [20 rows x 1 columns] In [113]: dft2.loc["2013-01-05"] Out[113]: A 2013-01-05 00:00:00 a 0.122297 b 1.422060 2013-01-05 12:00:00 a 0.370079 b 1.016331 In [114]: idx = pd.IndexSlice In [115]: dft2 = dft2.swaplevel(0, 1).sort_index() In [116]: dft2.loc[idx[:, "2013-01-05"], :] Out[116]: A a 2013-01-05 00:00:00 0.122297 2013-01-05 12:00:00 0.370079 b 2013-01-05 00:00:00 1.422060 2013-01-05 12:00:00 1.016331
New in version 0.25.0.
Slicing with string indexing also honors UTC offset.
In [117]: df = pd.DataFrame([0], index=pd.DatetimeIndex(["2019-01-01"], tz="US/Pacific")) In [118]: df Out[118]: 0 2019-01-01 00:00:00-08:00 0 In [119]: df["2019-01-01 12:00:00+04:00":"2019-01-01 13:00:00+04:00"] Out[119]: 0 2019-01-01 00:00:00-08:00 0
The same string used as an indexing parameter can be treated either as a slice or as an exact match depending on the resolution of the index. If the string is less accurate than the index, it will be treated as a slice, otherwise as an exact match.
Consider a Series object with a minute resolution index:
In [120]: series_minute = pd.Series( .....: [1, 2, 3], .....: pd.DatetimeIndex( .....: ["2011-12-31 23:59:00", "2012-01-01 00:00:00", "2012-01-01 00:02:00"] .....: ), .....: ) .....: In [121]: series_minute.index.resolution Out[121]: 'minute'
A timestamp string less accurate than a minute gives a Series object.
In [122]: series_minute["2011-12-31 23"] Out[122]: 2011-12-31 23:59:00 1 dtype: int64
A timestamp string with minute resolution (or more accurate), gives a scalar instead, i.e. it is not casted to a slice.
In [123]: series_minute["2011-12-31 23:59"] Out[123]: 1 In [124]: series_minute["2011-12-31 23:59:00"] Out[124]: 1
If index resolution is second, then the minute-accurate timestamp gives a Series.
In [125]: series_second = pd.Series( .....: [1, 2, 3], .....: pd.DatetimeIndex( .....: ["2011-12-31 23:59:59", "2012-01-01 00:00:00", "2012-01-01 00:00:01"] .....: ), .....: ) .....: In [126]: series_second.index.resolution Out[126]: 'second' In [127]: series_second["2011-12-31 23:59"] Out[127]: 2011-12-31 23:59:59 1 dtype: int64
If the timestamp string is treated as a slice, it can be used to index DataFrame with .loc[] as well.
.loc[]
In [128]: dft_minute = pd.DataFrame( .....: {"a": [1, 2, 3], "b": [4, 5, 6]}, index=series_minute.index .....: ) .....: In [129]: dft_minute.loc["2011-12-31 23"] Out[129]: a b 2011-12-31 23:59:00 1 4
However, if the string is treated as an exact match, the selection in DataFrame’s [] will be column-wise and not row-wise, see Indexing Basics. For example dft_minute['2011-12-31 23:59'] will raise KeyError as '2012-12-31 23:59' has the same resolution as the index and there is no column with such name:
[]
dft_minute['2011-12-31 23:59']
KeyError
'2012-12-31 23:59'
To always have unambiguous selection, whether the row is treated as a slice or a single selection, use .loc.
In [130]: dft_minute.loc["2011-12-31 23:59"] Out[130]: a 1 b 4 Name: 2011-12-31 23:59:00, dtype: int64
Note also that DatetimeIndex resolution cannot be less precise than day.
In [131]: series_monthly = pd.Series( .....: [1, 2, 3], pd.DatetimeIndex(["2011-12", "2012-01", "2012-02"]) .....: ) .....: In [132]: series_monthly.index.resolution Out[132]: 'day' In [133]: series_monthly["2011-12"] # returns Series Out[133]: 2011-12-01 1 dtype: int64
As discussed in previous section, indexing a DatetimeIndex with a partial string depends on the “accuracy” of the period, in other words how specific the interval is in relation to the resolution of the index. In contrast, indexing with Timestamp or datetime objects is exact, because the objects have exact meaning. These also follow the semantics of including both endpoints.
These Timestamp and datetime objects have exact hours, minutes, and seconds, even though they were not explicitly specified (they are 0).
hours, minutes,
seconds
0
In [134]: dft[datetime.datetime(2013, 1, 1): datetime.datetime(2013, 2, 28)] Out[134]: A 2013-01-01 00:00:00 0.276232 2013-01-01 00:01:00 -1.087401 2013-01-01 00:02:00 -0.673690 2013-01-01 00:03:00 0.113648 2013-01-01 00:04:00 -1.478427 ... ... 2013-02-27 23:56:00 1.197749 2013-02-27 23:57:00 0.720521 2013-02-27 23:58:00 -0.072718 2013-02-27 23:59:00 -0.681192 2013-02-28 00:00:00 -0.557501 [83521 rows x 1 columns]
With no defaults.
In [135]: dft[ .....: datetime.datetime(2013, 1, 1, 10, 12, 0): datetime.datetime( .....: 2013, 2, 28, 10, 12, 0 .....: ) .....: ] .....: Out[135]: A 2013-01-01 10:12:00 0.565375 2013-01-01 10:13:00 0.068184 2013-01-01 10:14:00 0.788871 2013-01-01 10:15:00 -0.280343 2013-01-01 10:16:00 0.931536 ... ... 2013-02-28 10:08:00 0.148098 2013-02-28 10:09:00 -0.388138 2013-02-28 10:10:00 0.139348 2013-02-28 10:11:00 0.085288 2013-02-28 10:12:00 0.950146 [83521 rows x 1 columns]
A truncate() convenience function is provided that is similar to slicing. Note that truncate assumes a 0 value for any unspecified date component in a DatetimeIndex in contrast to slicing which returns any partially matching dates:
truncate()
truncate
In [136]: rng2 = pd.date_range("2011-01-01", "2012-01-01", freq="W") In [137]: ts2 = pd.Series(np.random.randn(len(rng2)), index=rng2) In [138]: ts2.truncate(before="2011-11", after="2011-12") Out[138]: 2011-11-06 0.437823 2011-11-13 -0.293083 2011-11-20 -0.059881 2011-11-27 1.252450 Freq: W-SUN, dtype: float64 In [139]: ts2["2011-11":"2011-12"] Out[139]: 2011-11-06 0.437823 2011-11-13 -0.293083 2011-11-20 -0.059881 2011-11-27 1.252450 2011-12-04 0.046611 2011-12-11 0.059478 2011-12-18 -0.286539 2011-12-25 0.841669 Freq: W-SUN, dtype: float64
Even complicated fancy indexing that breaks the DatetimeIndex frequency regularity will result in a DatetimeIndex, although frequency is lost:
In [140]: ts2[[0, 2, 6]].index Out[140]: DatetimeIndex(['2011-01-02', '2011-01-16', '2011-02-13'], dtype='datetime64[ns]', freq=None)
There are several time/date properties that one can access from Timestamp or a collection of timestamps like a DatetimeIndex.
Property
Description
The year of the datetime
The month of the datetime
The days of the datetime
The hour of the datetime
The minutes of the datetime
The seconds of the datetime
The microseconds of the datetime
The nanoseconds of the datetime
date
Returns datetime.date (does not contain timezone information)
time
Returns datetime.time (does not contain timezone information)
timetz
Returns datetime.time as local time with timezone information
dayofyear
The ordinal day of year
day_of_year
weekofyear
The week ordinal of the year
week
dayofweek
The number of the day of the week with Monday=0, Sunday=6
day_of_week
weekday
quarter
Quarter of the date: Jan-Mar = 1, Apr-Jun = 2, etc.
days_in_month
The number of days in the month of the datetime
is_month_start
Logical indicating if first day of month (defined by frequency)
is_month_end
Logical indicating if last day of month (defined by frequency)
is_quarter_start
Logical indicating if first day of quarter (defined by frequency)
is_quarter_end
Logical indicating if last day of quarter (defined by frequency)
is_year_start
Logical indicating if first day of year (defined by frequency)
is_year_end
Logical indicating if last day of year (defined by frequency)
is_leap_year
Logical indicating if the date belongs to a leap year
Furthermore, if you have a Series with datetimelike values, then you can access these properties via the .dt accessor, as detailed in the section on .dt accessors.
.dt
New in version 1.1.0.
You may obtain the year, week and day components of the ISO year from the ISO 8601 standard:
In [141]: idx = pd.date_range(start="2019-12-29", freq="D", periods=4) In [142]: idx.isocalendar() Out[142]: year week day 2019-12-29 2019 52 7 2019-12-30 2020 1 1 2019-12-31 2020 1 2 2020-01-01 2020 1 3 In [143]: idx.to_series().dt.isocalendar() Out[143]: year week day 2019-12-29 2019 52 7 2019-12-30 2020 1 1 2019-12-31 2020 1 2 2020-01-01 2020 1 3
In the preceding examples, frequency strings (e.g. 'D') were used to specify a frequency that defined:
'D'
how the date times in DatetimeIndex were spaced when using date_range()
the frequency of a Period or PeriodIndex
These frequency strings map to a DateOffset object and its subclasses. A DateOffset is similar to a Timedelta that represents a duration of time but follows specific calendar duration rules. For example, a Timedelta day will always increment datetimes by 24 hours, while a DateOffset day will increment datetimes to the same time the next day whether a day represents 23, 24 or 25 hours due to daylight savings time. However, all DateOffset subclasses that are an hour or smaller (Hour, Minute, Second, Milli, Micro, Nano) behave like Timedelta and respect absolute time.
datetimes
Hour
Minute
Second
Milli
Micro
Nano
The basic DateOffset acts similar to dateutil.relativedelta (relativedelta documentation) that shifts a date time by the corresponding calendar duration specified. The arithmetic operator (+) or the apply method can be used to perform the shift.
dateutil.relativedelta
+
apply
# This particular day contains a day light savings time transition In [144]: ts = pd.Timestamp("2016-10-30 00:00:00", tz="Europe/Helsinki") # Respects absolute time In [145]: ts + pd.Timedelta(days=1) Out[145]: Timestamp('2016-10-30 23:00:00+0200', tz='Europe/Helsinki') # Respects calendar time In [146]: ts + pd.DateOffset(days=1) Out[146]: Timestamp('2016-10-31 00:00:00+0200', tz='Europe/Helsinki') In [147]: friday = pd.Timestamp("2018-01-05") In [148]: friday.day_name() Out[148]: 'Friday' # Add 2 business days (Friday --> Tuesday) In [149]: two_business_days = 2 * pd.offsets.BDay() In [150]: two_business_days.apply(friday) Out[150]: Timestamp('2018-01-09 00:00:00') In [151]: friday + two_business_days Out[151]: Timestamp('2018-01-09 00:00:00') In [152]: (friday + two_business_days).day_name() Out[152]: 'Tuesday'
Most DateOffsets have associated frequencies strings, or offset aliases, that can be passed into freq keyword arguments. The available date offsets and associated frequency strings can be found below:
DateOffsets
Date Offset
Frequency String
Generic offset class, defaults to absolute 24 hours
BDay or BusinessDay
BDay
BusinessDay
'B'
business day (weekday)
CDay or CustomBusinessDay
CDay
CustomBusinessDay
'C'
custom business day
Week
'W'
one week, optionally anchored on a day of the week
WeekOfMonth
'WOM'
the x-th day of the y-th week of each month
LastWeekOfMonth
'LWOM'
the x-th day of the last week of each month
MonthEnd
'M'
calendar month end
MonthBegin
'MS'
calendar month begin
BMonthEnd or BusinessMonthEnd
BMonthEnd
BusinessMonthEnd
'BM'
business month end
BMonthBegin or BusinessMonthBegin
BMonthBegin
BusinessMonthBegin
'BMS'
business month begin
CBMonthEnd or CustomBusinessMonthEnd
CBMonthEnd
CustomBusinessMonthEnd
'CBM'
custom business month end
CBMonthBegin or CustomBusinessMonthBegin
CBMonthBegin
CustomBusinessMonthBegin
'CBMS'
custom business month begin
SemiMonthEnd
'SM'
15th (or other day_of_month) and calendar month end
SemiMonthBegin
'SMS'
15th (or other day_of_month) and calendar month begin
QuarterEnd
'Q'
calendar quarter end
QuarterBegin
'QS'
calendar quarter begin
BQuarterEnd
'BQ
business quarter end
BQuarterBegin
'BQS'
business quarter begin
FY5253Quarter
'REQ'
retail (aka 52-53 week) quarter
YearEnd
'A'
calendar year end
YearBegin
'AS' or 'BYS'
'AS'
'BYS'
calendar year begin
BYearEnd
'BA'
business year end
BYearBegin
'BAS'
business year begin
FY5253
'RE'
retail (aka 52-53 week) year
Easter
Easter holiday
BusinessHour
'BH'
business hour
CustomBusinessHour
'CBH'
custom business hour
Day
one absolute day
'H'
one hour
'T' or 'min'
'T'
'min'
one minute
'S'
one second
'L' or 'ms'
'L'
'ms'
one millisecond
'U' or 'us'
'U'
'us'
one microsecond
'N'
one nanosecond
DateOffsets additionally have rollforward() and rollback() methods for moving a date forward or backward respectively to a valid offset date relative to the offset. For example, business offsets will roll dates that land on the weekends (Saturday and Sunday) forward to Monday since business offsets operate on the weekdays.
rollforward()
rollback()
In [153]: ts = pd.Timestamp("2018-01-06 00:00:00") In [154]: ts.day_name() Out[154]: 'Saturday' # BusinessHour's valid offset dates are Monday through Friday In [155]: offset = pd.offsets.BusinessHour(start="09:00") # Bring the date to the closest offset date (Monday) In [156]: offset.rollforward(ts) Out[156]: Timestamp('2018-01-08 09:00:00') # Date is brought to the closest offset date first and then the hour is added In [157]: ts + offset Out[157]: Timestamp('2018-01-08 10:00:00')
These operations preserve time (hour, minute, etc) information by default. To reset time to midnight, use normalize() before or after applying the operation (depending on whether you want the time information included in the operation).
normalize()
In [158]: ts = pd.Timestamp("2014-01-01 09:00") In [159]: day = pd.offsets.Day() In [160]: day.apply(ts) Out[160]: Timestamp('2014-01-02 09:00:00') In [161]: day.apply(ts).normalize() Out[161]: Timestamp('2014-01-02 00:00:00') In [162]: ts = pd.Timestamp("2014-01-01 22:00") In [163]: hour = pd.offsets.Hour() In [164]: hour.apply(ts) Out[164]: Timestamp('2014-01-01 23:00:00') In [165]: hour.apply(ts).normalize() Out[165]: Timestamp('2014-01-01 00:00:00') In [166]: hour.apply(pd.Timestamp("2014-01-01 23:30")).normalize() Out[166]: Timestamp('2014-01-02 00:00:00')
Some of the offsets can be “parameterized” when created to result in different behaviors. For example, the Week offset for generating weekly data accepts a weekday parameter which results in the generated dates always lying on a particular day of the week:
In [167]: d = datetime.datetime(2008, 8, 18, 9, 0) In [168]: d Out[168]: datetime.datetime(2008, 8, 18, 9, 0) In [169]: d + pd.offsets.Week() Out[169]: Timestamp('2008-08-25 09:00:00') In [170]: d + pd.offsets.Week(weekday=4) Out[170]: Timestamp('2008-08-22 09:00:00') In [171]: (d + pd.offsets.Week(weekday=4)).weekday() Out[171]: 4 In [172]: d - pd.offsets.Week() Out[172]: Timestamp('2008-08-11 09:00:00')
The normalize option will be effective for addition and subtraction.
normalize
In [173]: d + pd.offsets.Week(normalize=True) Out[173]: Timestamp('2008-08-25 00:00:00') In [174]: d - pd.offsets.Week(normalize=True) Out[174]: Timestamp('2008-08-11 00:00:00')
Another example is parameterizing YearEnd with the specific ending month:
In [175]: d + pd.offsets.YearEnd() Out[175]: Timestamp('2008-12-31 09:00:00') In [176]: d + pd.offsets.YearEnd(month=6) Out[176]: Timestamp('2009-06-30 09:00:00')
Offsets can be used with either a Series or DatetimeIndex to apply the offset to each element.
In [177]: rng = pd.date_range("2012-01-01", "2012-01-03") In [178]: s = pd.Series(rng) In [179]: rng Out[179]: DatetimeIndex(['2012-01-01', '2012-01-02', '2012-01-03'], dtype='datetime64[ns]', freq='D') In [180]: rng + pd.DateOffset(months=2) Out[180]: DatetimeIndex(['2012-03-01', '2012-03-02', '2012-03-03'], dtype='datetime64[ns]', freq=None) In [181]: s + pd.DateOffset(months=2) Out[181]: 0 2012-03-01 1 2012-03-02 2 2012-03-03 dtype: datetime64[ns] In [182]: s - pd.DateOffset(months=2) Out[182]: 0 2011-11-01 1 2011-11-02 2 2011-11-03 dtype: datetime64[ns]
If the offset class maps directly to a Timedelta (Day, Hour, Minute, Second, Micro, Milli, Nano) it can be used exactly like a Timedelta - see the Timedelta section for more examples.
In [183]: s - pd.offsets.Day(2) Out[183]: 0 2011-12-30 1 2011-12-31 2 2012-01-01 dtype: datetime64[ns] In [184]: td = s - pd.Series(pd.date_range("2011-12-29", "2011-12-31")) In [185]: td Out[185]: 0 3 days 1 3 days 2 3 days dtype: timedelta64[ns] In [186]: td + pd.offsets.Minute(15) Out[186]: 0 3 days 00:15:00 1 3 days 00:15:00 2 3 days 00:15:00 dtype: timedelta64[ns]
Note that some offsets (such as BQuarterEnd) do not have a vectorized implementation. They can still be used but may calculate significantly slower and will show a PerformanceWarning
PerformanceWarning
In [187]: rng + pd.offsets.BQuarterEnd() Out[187]: DatetimeIndex(['2012-03-30', '2012-03-30', '2012-03-30'], dtype='datetime64[ns]', freq=None)
The CDay or CustomBusinessDay class provides a parametric BusinessDay class which can be used to create customized business day calendars which account for local holidays and local weekend conventions.
As an interesting example, let’s look at Egypt where a Friday-Saturday weekend is observed.
In [188]: weekmask_egypt = "Sun Mon Tue Wed Thu" # They also observe International Workers' Day so let's # add that for a couple of years In [189]: holidays = [ .....: "2012-05-01", .....: datetime.datetime(2013, 5, 1), .....: np.datetime64("2014-05-01"), .....: ] .....: In [190]: bday_egypt = pd.offsets.CustomBusinessDay( .....: holidays=holidays, .....: weekmask=weekmask_egypt, .....: ) .....: In [191]: dt = datetime.datetime(2013, 4, 30) In [192]: dt + 2 * bday_egypt Out[192]: Timestamp('2013-05-05 00:00:00')
Let’s map to the weekday names:
In [193]: dts = pd.date_range(dt, periods=5, freq=bday_egypt) In [194]: pd.Series(dts.weekday, dts).map(pd.Series("Mon Tue Wed Thu Fri Sat Sun".split())) Out[194]: 2013-04-30 Tue 2013-05-02 Thu 2013-05-05 Sun 2013-05-06 Mon 2013-05-07 Tue Freq: C, dtype: object
Holiday calendars can be used to provide the list of holidays. See the holiday calendar section for more information.
In [195]: from pandas.tseries.holiday import USFederalHolidayCalendar In [196]: bday_us = pd.offsets.CustomBusinessDay(calendar=USFederalHolidayCalendar()) # Friday before MLK Day In [197]: dt = datetime.datetime(2014, 1, 17) # Tuesday after MLK Day (Monday is skipped because it's a holiday) In [198]: dt + bday_us Out[198]: Timestamp('2014-01-21 00:00:00')
Monthly offsets that respect a certain holiday calendar can be defined in the usual way.
In [199]: bmth_us = pd.offsets.CustomBusinessMonthBegin(calendar=USFederalHolidayCalendar()) # Skip new years In [200]: dt = datetime.datetime(2013, 12, 17) In [201]: dt + bmth_us Out[201]: Timestamp('2014-01-02 00:00:00') # Define date index with custom offset In [202]: pd.date_range(start="20100101", end="20120101", freq=bmth_us) Out[202]: DatetimeIndex(['2010-01-04', '2010-02-01', '2010-03-01', '2010-04-01', '2010-05-03', '2010-06-01', '2010-07-01', '2010-08-02', '2010-09-01', '2010-10-01', '2010-11-01', '2010-12-01', '2011-01-03', '2011-02-01', '2011-03-01', '2011-04-01', '2011-05-02', '2011-06-01', '2011-07-01', '2011-08-01', '2011-09-01', '2011-10-03', '2011-11-01', '2011-12-01'], dtype='datetime64[ns]', freq='CBMS')
The frequency string ‘C’ is used to indicate that a CustomBusinessDay DateOffset is used, it is important to note that since CustomBusinessDay is a parameterised type, instances of CustomBusinessDay may differ and this is not detectable from the ‘C’ frequency string. The user therefore needs to ensure that the ‘C’ frequency string is used consistently within the user’s application.
The BusinessHour class provides a business hour representation on BusinessDay, allowing to use specific start and end times.
By default, BusinessHour uses 9:00 - 17:00 as business hours. Adding BusinessHour will increment Timestamp by hourly frequency. If target Timestamp is out of business hours, move to the next business hour then increment it. If the result exceeds the business hours end, the remaining hours are added to the next business day.
In [203]: bh = pd.offsets.BusinessHour() In [204]: bh Out[204]: <BusinessHour: BH=09:00-17:00> # 2014-08-01 is Friday In [205]: pd.Timestamp("2014-08-01 10:00").weekday() Out[205]: 4 In [206]: pd.Timestamp("2014-08-01 10:00") + bh Out[206]: Timestamp('2014-08-01 11:00:00') # Below example is the same as: pd.Timestamp('2014-08-01 09:00') + bh In [207]: pd.Timestamp("2014-08-01 08:00") + bh Out[207]: Timestamp('2014-08-01 10:00:00') # If the results is on the end time, move to the next business day In [208]: pd.Timestamp("2014-08-01 16:00") + bh Out[208]: Timestamp('2014-08-04 09:00:00') # Remainings are added to the next day In [209]: pd.Timestamp("2014-08-01 16:30") + bh Out[209]: Timestamp('2014-08-04 09:30:00') # Adding 2 business hours In [210]: pd.Timestamp("2014-08-01 10:00") + pd.offsets.BusinessHour(2) Out[210]: Timestamp('2014-08-01 12:00:00') # Subtracting 3 business hours In [211]: pd.Timestamp("2014-08-01 10:00") + pd.offsets.BusinessHour(-3) Out[211]: Timestamp('2014-07-31 15:00:00')
You can also specify start and end time by keywords. The argument must be a str with an hour:minute representation or a datetime.time instance. Specifying seconds, microseconds and nanoseconds as business hour results in ValueError.
str
hour:minute
datetime.time
ValueError
In [212]: bh = pd.offsets.BusinessHour(start="11:00", end=datetime.time(20, 0)) In [213]: bh Out[213]: <BusinessHour: BH=11:00-20:00> In [214]: pd.Timestamp("2014-08-01 13:00") + bh Out[214]: Timestamp('2014-08-01 14:00:00') In [215]: pd.Timestamp("2014-08-01 09:00") + bh Out[215]: Timestamp('2014-08-01 12:00:00') In [216]: pd.Timestamp("2014-08-01 18:00") + bh Out[216]: Timestamp('2014-08-01 19:00:00')
Passing start time later than end represents midnight business hour. In this case, business hour exceeds midnight and overlap to the next day. Valid business hours are distinguished by whether it started from valid BusinessDay.
In [217]: bh = pd.offsets.BusinessHour(start="17:00", end="09:00") In [218]: bh Out[218]: <BusinessHour: BH=17:00-09:00> In [219]: pd.Timestamp("2014-08-01 17:00") + bh Out[219]: Timestamp('2014-08-01 18:00:00') In [220]: pd.Timestamp("2014-08-01 23:00") + bh Out[220]: Timestamp('2014-08-02 00:00:00') # Although 2014-08-02 is Saturday, # it is valid because it starts from 08-01 (Friday). In [221]: pd.Timestamp("2014-08-02 04:00") + bh Out[221]: Timestamp('2014-08-02 05:00:00') # Although 2014-08-04 is Monday, # it is out of business hours because it starts from 08-03 (Sunday). In [222]: pd.Timestamp("2014-08-04 04:00") + bh Out[222]: Timestamp('2014-08-04 18:00:00')
Applying BusinessHour.rollforward and rollback to out of business hours results in the next business hour start or previous day’s end. Different from other offsets, BusinessHour.rollforward may output different results from apply by definition.
BusinessHour.rollforward
rollback
This is because one day’s business hour end is equal to next day’s business hour start. For example, under the default business hours (9:00 - 17:00), there is no gap (0 minutes) between 2014-08-01 17:00 and 2014-08-04 09:00.
2014-08-01 17:00
2014-08-04 09:00
# This adjusts a Timestamp to business hour edge In [223]: pd.offsets.BusinessHour().rollback(pd.Timestamp("2014-08-02 15:00")) Out[223]: Timestamp('2014-08-01 17:00:00') In [224]: pd.offsets.BusinessHour().rollforward(pd.Timestamp("2014-08-02 15:00")) Out[224]: Timestamp('2014-08-04 09:00:00') # It is the same as BusinessHour().apply(pd.Timestamp('2014-08-01 17:00')). # And it is the same as BusinessHour().apply(pd.Timestamp('2014-08-04 09:00')) In [225]: pd.offsets.BusinessHour().apply(pd.Timestamp("2014-08-02 15:00")) Out[225]: Timestamp('2014-08-04 10:00:00') # BusinessDay results (for reference) In [226]: pd.offsets.BusinessHour().rollforward(pd.Timestamp("2014-08-02")) Out[226]: Timestamp('2014-08-04 09:00:00') # It is the same as BusinessDay().apply(pd.Timestamp('2014-08-01')) # The result is the same as rollworward because BusinessDay never overlap. In [227]: pd.offsets.BusinessHour().apply(pd.Timestamp("2014-08-02")) Out[227]: Timestamp('2014-08-04 10:00:00')
BusinessHour regards Saturday and Sunday as holidays. To use arbitrary holidays, you can use CustomBusinessHour offset, as explained in the following subsection.
The CustomBusinessHour is a mixture of BusinessHour and CustomBusinessDay which allows you to specify arbitrary holidays. CustomBusinessHour works as the same as BusinessHour except that it skips specified custom holidays.
In [228]: from pandas.tseries.holiday import USFederalHolidayCalendar In [229]: bhour_us = pd.offsets.CustomBusinessHour(calendar=USFederalHolidayCalendar()) # Friday before MLK Day In [230]: dt = datetime.datetime(2014, 1, 17, 15) In [231]: dt + bhour_us Out[231]: Timestamp('2014-01-17 16:00:00') # Tuesday after MLK Day (Monday is skipped because it's a holiday) In [232]: dt + bhour_us * 2 Out[232]: Timestamp('2014-01-21 09:00:00')
You can use keyword arguments supported by either BusinessHour and CustomBusinessDay.
In [233]: bhour_mon = pd.offsets.CustomBusinessHour(start="10:00", weekmask="Tue Wed Thu Fri") # Monday is skipped because it's a holiday, business hour starts from 10:00 In [234]: dt + bhour_mon * 2 Out[234]: Timestamp('2014-01-21 10:00:00')
A number of string aliases are given to useful common time series frequencies. We will refer to these aliases as offset aliases.
Alias
B
business day frequency
C
custom business day frequency
D
calendar day frequency
W
weekly frequency
M
month end frequency
SM
semi-month end frequency (15th and end of month)
BM
business month end frequency
CBM
custom business month end frequency
MS
month start frequency
SMS
semi-month start frequency (1st and 15th)
BMS
business month start frequency
CBMS
custom business month start frequency
Q
quarter end frequency
BQ
business quarter end frequency
QS
quarter start frequency
BQS
business quarter start frequency
A, Y
year end frequency
BA, BY
business year end frequency
AS, YS
year start frequency
BAS, BYS
business year start frequency
BH
business hour frequency
H
hourly frequency
T, min
minutely frequency
S
secondly frequency
L, ms
milliseconds
U, us
microseconds
N
nanoseconds
As we have seen previously, the alias and the offset instance are fungible in most functions:
In [235]: pd.date_range(start, periods=5, freq="B") Out[235]: DatetimeIndex(['2011-01-03', '2011-01-04', '2011-01-05', '2011-01-06', '2011-01-07'], dtype='datetime64[ns]', freq='B') In [236]: pd.date_range(start, periods=5, freq=pd.offsets.BDay()) Out[236]: DatetimeIndex(['2011-01-03', '2011-01-04', '2011-01-05', '2011-01-06', '2011-01-07'], dtype='datetime64[ns]', freq='B')
You can combine together day and intraday offsets:
In [237]: pd.date_range(start, periods=10, freq="2h20min") Out[237]: DatetimeIndex(['2011-01-01 00:00:00', '2011-01-01 02:20:00', '2011-01-01 04:40:00', '2011-01-01 07:00:00', '2011-01-01 09:20:00', '2011-01-01 11:40:00', '2011-01-01 14:00:00', '2011-01-01 16:20:00', '2011-01-01 18:40:00', '2011-01-01 21:00:00'], dtype='datetime64[ns]', freq='140T') In [238]: pd.date_range(start, periods=10, freq="1D10U") Out[238]: DatetimeIndex([ '2011-01-01 00:00:00', '2011-01-02 00:00:00.000010', '2011-01-03 00:00:00.000020', '2011-01-04 00:00:00.000030', '2011-01-05 00:00:00.000040', '2011-01-06 00:00:00.000050', '2011-01-07 00:00:00.000060', '2011-01-08 00:00:00.000070', '2011-01-09 00:00:00.000080', '2011-01-10 00:00:00.000090'], dtype='datetime64[ns]', freq='86400000010U')
For some frequencies you can specify an anchoring suffix:
W-SUN
weekly frequency (Sundays). Same as ‘W’
W-MON
weekly frequency (Mondays)
W-TUE
weekly frequency (Tuesdays)
W-WED
weekly frequency (Wednesdays)
W-THU
weekly frequency (Thursdays)
W-FRI
weekly frequency (Fridays)
W-SAT
weekly frequency (Saturdays)
(B)Q(S)-DEC
quarterly frequency, year ends in December. Same as ‘Q’
(B)Q(S)-JAN
quarterly frequency, year ends in January
(B)Q(S)-FEB
quarterly frequency, year ends in February
(B)Q(S)-MAR
quarterly frequency, year ends in March
(B)Q(S)-APR
quarterly frequency, year ends in April
(B)Q(S)-MAY
quarterly frequency, year ends in May
(B)Q(S)-JUN
quarterly frequency, year ends in June
(B)Q(S)-JUL
quarterly frequency, year ends in July
(B)Q(S)-AUG
quarterly frequency, year ends in August
(B)Q(S)-SEP
quarterly frequency, year ends in September
(B)Q(S)-OCT
quarterly frequency, year ends in October
(B)Q(S)-NOV
quarterly frequency, year ends in November
(B)A(S)-DEC
annual frequency, anchored end of December. Same as ‘A’
(B)A(S)-JAN
annual frequency, anchored end of January
(B)A(S)-FEB
annual frequency, anchored end of February
(B)A(S)-MAR
annual frequency, anchored end of March
(B)A(S)-APR
annual frequency, anchored end of April
(B)A(S)-MAY
annual frequency, anchored end of May
(B)A(S)-JUN
annual frequency, anchored end of June
(B)A(S)-JUL
annual frequency, anchored end of July
(B)A(S)-AUG
annual frequency, anchored end of August
(B)A(S)-SEP
annual frequency, anchored end of September
(B)A(S)-OCT
annual frequency, anchored end of October
(B)A(S)-NOV
annual frequency, anchored end of November
These can be used as arguments to date_range, bdate_range, constructors for DatetimeIndex, as well as various other timeseries-related functions in pandas.
For those offsets that are anchored to the start or end of specific frequency (MonthEnd, MonthBegin, WeekEnd, etc), the following rules apply to rolling forward and backwards.
WeekEnd
When n is not 0, if the given date is not on an anchor point, it snapped to the next(previous) anchor point, and moved |n|-1 additional steps forwards or backwards.
n
|n|-1
In [239]: pd.Timestamp("2014-01-02") + pd.offsets.MonthBegin(n=1) Out[239]: Timestamp('2014-02-01 00:00:00') In [240]: pd.Timestamp("2014-01-02") + pd.offsets.MonthEnd(n=1) Out[240]: Timestamp('2014-01-31 00:00:00') In [241]: pd.Timestamp("2014-01-02") - pd.offsets.MonthBegin(n=1) Out[241]: Timestamp('2014-01-01 00:00:00') In [242]: pd.Timestamp("2014-01-02") - pd.offsets.MonthEnd(n=1) Out[242]: Timestamp('2013-12-31 00:00:00') In [243]: pd.Timestamp("2014-01-02") + pd.offsets.MonthBegin(n=4) Out[243]: Timestamp('2014-05-01 00:00:00') In [244]: pd.Timestamp("2014-01-02") - pd.offsets.MonthBegin(n=4) Out[244]: Timestamp('2013-10-01 00:00:00')
If the given date is on an anchor point, it is moved |n| points forwards or backwards.
|n|
In [245]: pd.Timestamp("2014-01-01") + pd.offsets.MonthBegin(n=1) Out[245]: Timestamp('2014-02-01 00:00:00') In [246]: pd.Timestamp("2014-01-31") + pd.offsets.MonthEnd(n=1) Out[246]: Timestamp('2014-02-28 00:00:00') In [247]: pd.Timestamp("2014-01-01") - pd.offsets.MonthBegin(n=1) Out[247]: Timestamp('2013-12-01 00:00:00') In [248]: pd.Timestamp("2014-01-31") - pd.offsets.MonthEnd(n=1) Out[248]: Timestamp('2013-12-31 00:00:00') In [249]: pd.Timestamp("2014-01-01") + pd.offsets.MonthBegin(n=4) Out[249]: Timestamp('2014-05-01 00:00:00') In [250]: pd.Timestamp("2014-01-31") - pd.offsets.MonthBegin(n=4) Out[250]: Timestamp('2013-10-01 00:00:00')
For the case when n=0, the date is not moved if on an anchor point, otherwise it is rolled forward to the next anchor point.
n=0
In [251]: pd.Timestamp("2014-01-02") + pd.offsets.MonthBegin(n=0) Out[251]: Timestamp('2014-02-01 00:00:00') In [252]: pd.Timestamp("2014-01-02") + pd.offsets.MonthEnd(n=0) Out[252]: Timestamp('2014-01-31 00:00:00') In [253]: pd.Timestamp("2014-01-01") + pd.offsets.MonthBegin(n=0) Out[253]: Timestamp('2014-01-01 00:00:00') In [254]: pd.Timestamp("2014-01-31") + pd.offsets.MonthEnd(n=0) Out[254]: Timestamp('2014-01-31 00:00:00')
Holidays and calendars provide a simple way to define holiday rules to be used with CustomBusinessDay or in other analysis that requires a predefined set of holidays. The AbstractHolidayCalendar class provides all the necessary methods to return a list of holidays and only rules need to be defined in a specific holiday calendar class. Furthermore, the start_date and end_date class attributes determine over what date range holidays are generated. These should be overwritten on the AbstractHolidayCalendar class to have the range apply to all calendar subclasses. USFederalHolidayCalendar is the only calendar that exists and primarily serves as an example for developing other calendars.
AbstractHolidayCalendar
rules
start_date
end_date
USFederalHolidayCalendar
For holidays that occur on fixed dates (e.g., US Memorial Day or July 4th) an observance rule determines when that holiday is observed if it falls on a weekend or some other non-observed day. Defined observance rules are:
Rule
nearest_workday
move Saturday to Friday and Sunday to Monday
sunday_to_monday
move Sunday to following Monday
next_monday_or_tuesday
move Saturday to Monday and Sunday/Monday to Tuesday
previous_friday
move Saturday and Sunday to previous Friday”
next_monday
move Saturday and Sunday to following Monday
An example of how holidays and holiday calendars are defined:
In [255]: from pandas.tseries.holiday import ( .....: Holiday, .....: USMemorialDay, .....: AbstractHolidayCalendar, .....: nearest_workday, .....: MO, .....: ) .....: In [256]: class ExampleCalendar(AbstractHolidayCalendar): .....: rules = [ .....: USMemorialDay, .....: Holiday("July 4th", month=7, day=4, observance=nearest_workday), .....: Holiday( .....: "Columbus Day", .....: month=10, .....: day=1, .....: offset=pd.DateOffset(weekday=MO(2)), .....: ), .....: ] .....: In [257]: cal = ExampleCalendar() In [258]: cal.holidays(datetime.datetime(2012, 1, 1), datetime.datetime(2012, 12, 31)) Out[258]: DatetimeIndex(['2012-05-28', '2012-07-04', '2012-10-08'], dtype='datetime64[ns]', freq=None)
weekday=MO(2) is same as 2 * Week(weekday=2)
Using this calendar, creating an index or doing offset arithmetic skips weekends and holidays (i.e., Memorial Day/July 4th). For example, the below defines a custom business day offset using the ExampleCalendar. Like any other offset, it can be used to create a DatetimeIndex or added to datetime or Timestamp objects.
ExampleCalendar
In [259]: pd.date_range( .....: start="7/1/2012", end="7/10/2012", freq=pd.offsets.CDay(calendar=cal) .....: ).to_pydatetime() .....: Out[259]: array([datetime.datetime(2012, 7, 2, 0, 0), datetime.datetime(2012, 7, 3, 0, 0), datetime.datetime(2012, 7, 5, 0, 0), datetime.datetime(2012, 7, 6, 0, 0), datetime.datetime(2012, 7, 9, 0, 0), datetime.datetime(2012, 7, 10, 0, 0)], dtype=object) In [260]: offset = pd.offsets.CustomBusinessDay(calendar=cal) In [261]: datetime.datetime(2012, 5, 25) + offset Out[261]: Timestamp('2012-05-29 00:00:00') In [262]: datetime.datetime(2012, 7, 3) + offset Out[262]: Timestamp('2012-07-05 00:00:00') In [263]: datetime.datetime(2012, 7, 3) + 2 * offset Out[263]: Timestamp('2012-07-06 00:00:00') In [264]: datetime.datetime(2012, 7, 6) + offset Out[264]: Timestamp('2012-07-09 00:00:00')
Ranges are defined by the start_date and end_date class attributes of AbstractHolidayCalendar. The defaults are shown below.
In [265]: AbstractHolidayCalendar.start_date Out[265]: Timestamp('1970-01-01 00:00:00') In [266]: AbstractHolidayCalendar.end_date Out[266]: Timestamp('2200-12-31 00:00:00')
These dates can be overwritten by setting the attributes as datetime/Timestamp/string.
In [267]: AbstractHolidayCalendar.start_date = datetime.datetime(2012, 1, 1) In [268]: AbstractHolidayCalendar.end_date = datetime.datetime(2012, 12, 31) In [269]: cal.holidays() Out[269]: DatetimeIndex(['2012-05-28', '2012-07-04', '2012-10-08'], dtype='datetime64[ns]', freq=None)
Every calendar class is accessible by name using the get_calendar function which returns a holiday class instance. Any imported calendar class will automatically be available by this function. Also, HolidayCalendarFactory provides an easy interface to create calendars that are combinations of calendars or calendars with additional rules.
get_calendar
HolidayCalendarFactory
In [270]: from pandas.tseries.holiday import get_calendar, HolidayCalendarFactory, USLaborDay In [271]: cal = get_calendar("ExampleCalendar") In [272]: cal.rules Out[272]: [Holiday: Memorial Day (month=5, day=31, offset=<DateOffset: weekday=MO(-1)>), Holiday: July 4th (month=7, day=4, observance=<function nearest_workday at 0x7fa919c66f70>), Holiday: Columbus Day (month=10, day=1, offset=<DateOffset: weekday=MO(+2)>)] In [273]: new_cal = HolidayCalendarFactory("NewExampleCalendar", cal, USLaborDay) In [274]: new_cal.rules Out[274]: [Holiday: Labor Day (month=9, day=1, offset=<DateOffset: weekday=MO(+1)>), Holiday: Memorial Day (month=5, day=31, offset=<DateOffset: weekday=MO(-1)>), Holiday: July 4th (month=7, day=4, observance=<function nearest_workday at 0x7fa919c66f70>), Holiday: Columbus Day (month=10, day=1, offset=<DateOffset: weekday=MO(+2)>)]
One may want to shift or lag the values in a time series back and forward in time. The method for this is shift(), which is available on all of the pandas objects.
shift()
In [275]: ts = pd.Series(range(len(rng)), index=rng) In [276]: ts = ts[:5] In [277]: ts.shift(1) Out[277]: 2012-01-01 NaN 2012-01-02 0.0 2012-01-03 1.0 Freq: D, dtype: float64
The shift method accepts an freq argument which can accept a DateOffset class or other timedelta-like object or also an offset alias.
When freq is specified, shift method changes all the dates in the index rather than changing the alignment of the data and the index:
In [278]: ts.shift(5, freq="D") Out[278]: 2012-01-06 0 2012-01-07 1 2012-01-08 2 Freq: D, dtype: int64 In [279]: ts.shift(5, freq=pd.offsets.BDay()) Out[279]: 2012-01-06 0 2012-01-09 1 2012-01-10 2 dtype: int64 In [280]: ts.shift(5, freq="BM") Out[280]: 2012-05-31 0 2012-05-31 1 2012-05-31 2 dtype: int64
Note that with when freq is specified, the leading entry is no longer NaN because the data is not being realigned.
The primary function for changing frequencies is the asfreq() method. For a DatetimeIndex, this is basically just a thin, but convenient wrapper around reindex() which generates a date_range and calls reindex.
asfreq()
reindex()
reindex
In [281]: dr = pd.date_range("1/1/2010", periods=3, freq=3 * pd.offsets.BDay()) In [282]: ts = pd.Series(np.random.randn(3), index=dr) In [283]: ts Out[283]: 2010-01-01 1.494522 2010-01-06 -0.778425 2010-01-11 -0.253355 Freq: 3B, dtype: float64 In [284]: ts.asfreq(pd.offsets.BDay()) Out[284]: 2010-01-01 1.494522 2010-01-04 NaN 2010-01-05 NaN 2010-01-06 -0.778425 2010-01-07 NaN 2010-01-08 NaN 2010-01-11 -0.253355 Freq: B, dtype: float64
asfreq provides a further convenience so you can specify an interpolation method for any gaps that may appear after the frequency conversion.
asfreq
In [285]: ts.asfreq(pd.offsets.BDay(), method="pad") Out[285]: 2010-01-01 1.494522 2010-01-04 1.494522 2010-01-05 1.494522 2010-01-06 -0.778425 2010-01-07 -0.778425 2010-01-08 -0.778425 2010-01-11 -0.253355 Freq: B, dtype: float64
Related to asfreq and reindex is fillna(), which is documented in the missing data section.
fillna()
DatetimeIndex can be converted to an array of Python native datetime.datetime objects using the to_pydatetime method.
to_pydatetime
pandas has a simple, powerful, and efficient functionality for performing resampling operations during frequency conversion (e.g., converting secondly data into 5-minutely data). This is extremely common in, but not limited to, financial applications.
resample() is a time-based groupby, followed by a reduction method on each of its groups. See some cookbook examples for some advanced strategies.
resample()
The resample() method can be used directly from DataFrameGroupBy objects, see the groupby docs.
DataFrameGroupBy
In [286]: rng = pd.date_range("1/1/2012", periods=100, freq="S") In [287]: ts = pd.Series(np.random.randint(0, 500, len(rng)), index=rng) In [288]: ts.resample("5Min").sum() Out[288]: 2012-01-01 25103 Freq: 5T, dtype: int64
The resample function is very flexible and allows you to specify many different parameters to control the frequency conversion and resampling operation.
resample
Any function available via dispatching is available as a method of the returned object, including sum, mean, std, sem, max, min, median, first, last, ohlc:
sum
mean
std
sem
max
min
median
first
last
ohlc
In [289]: ts.resample("5Min").mean() Out[289]: 2012-01-01 251.03 Freq: 5T, dtype: float64 In [290]: ts.resample("5Min").ohlc() Out[290]: open high low close 2012-01-01 308 460 9 205 In [291]: ts.resample("5Min").max() Out[291]: 2012-01-01 460 Freq: 5T, dtype: int64
For downsampling, closed can be set to ‘left’ or ‘right’ to specify which end of the interval is closed:
closed
In [292]: ts.resample("5Min", closed="right").mean() Out[292]: 2011-12-31 23:55:00 308.000000 2012-01-01 00:00:00 250.454545 Freq: 5T, dtype: float64 In [293]: ts.resample("5Min", closed="left").mean() Out[293]: 2012-01-01 251.03 Freq: 5T, dtype: float64
Parameters like label are used to manipulate the resulting labels. label specifies whether the result is labeled with the beginning or the end of the interval.
label
In [294]: ts.resample("5Min").mean() # by default label='left' Out[294]: 2012-01-01 251.03 Freq: 5T, dtype: float64 In [295]: ts.resample("5Min", label="left").mean() Out[295]: 2012-01-01 251.03 Freq: 5T, dtype: float64
The default values for label and closed is ‘left’ for all frequency offsets except for ‘M’, ‘A’, ‘Q’, ‘BM’, ‘BA’, ‘BQ’, and ‘W’ which all have a default of ‘right’.
This might unintendedly lead to looking ahead, where the value for a later time is pulled back to a previous time as in the following example with the BusinessDay frequency:
In [296]: s = pd.date_range("2000-01-01", "2000-01-05").to_series() In [297]: s.iloc[2] = pd.NaT In [298]: s.dt.day_name() Out[298]: 2000-01-01 Saturday 2000-01-02 Sunday 2000-01-03 NaN 2000-01-04 Tuesday 2000-01-05 Wednesday Freq: D, dtype: object # default: label='left', closed='left' In [299]: s.resample("B").last().dt.day_name() Out[299]: 1999-12-31 Sunday 2000-01-03 NaN 2000-01-04 Tuesday 2000-01-05 Wednesday Freq: B, dtype: object
Notice how the value for Sunday got pulled back to the previous Friday. To get the behavior where the value for Sunday is pushed to Monday, use instead
In [300]: s.resample("B", label="right", closed="right").last().dt.day_name() Out[300]: 2000-01-03 Sunday 2000-01-04 Tuesday 2000-01-05 Wednesday Freq: B, dtype: object
The axis parameter can be set to 0 or 1 and allows you to resample the specified axis for a DataFrame.
axis
kind can be set to ‘timestamp’ or ‘period’ to convert the resulting index to/from timestamp and time span representations. By default resample retains the input representation.
kind
convention can be set to ‘start’ or ‘end’ when resampling period data (detail below). It specifies how low frequency periods are converted to higher frequency periods.
convention
For upsampling, you can specify a way to upsample and the limit parameter to interpolate over the gaps that are created:
limit
# from secondly to every 250 milliseconds In [301]: ts[:2].resample("250L").asfreq() Out[301]: 2012-01-01 00:00:00.000 308.0 2012-01-01 00:00:00.250 NaN 2012-01-01 00:00:00.500 NaN 2012-01-01 00:00:00.750 NaN 2012-01-01 00:00:01.000 204.0 Freq: 250L, dtype: float64 In [302]: ts[:2].resample("250L").ffill() Out[302]: 2012-01-01 00:00:00.000 308 2012-01-01 00:00:00.250 308 2012-01-01 00:00:00.500 308 2012-01-01 00:00:00.750 308 2012-01-01 00:00:01.000 204 Freq: 250L, dtype: int64 In [303]: ts[:2].resample("250L").ffill(limit=2) Out[303]: 2012-01-01 00:00:00.000 308.0 2012-01-01 00:00:00.250 308.0 2012-01-01 00:00:00.500 308.0 2012-01-01 00:00:00.750 NaN 2012-01-01 00:00:01.000 204.0 Freq: 250L, dtype: float64
Sparse timeseries are the ones where you have a lot fewer points relative to the amount of time you are looking to resample. Naively upsampling a sparse series can potentially generate lots of intermediate values. When you don’t want to use a method to fill these values, e.g. fill_method is None, then intermediate values will be filled with NaN.
fill_method
NaN
Since resample is a time-based groupby, the following is a method to efficiently resample only the groups that are not all NaN.
In [304]: rng = pd.date_range("2014-1-1", periods=100, freq="D") + pd.Timedelta("1s") In [305]: ts = pd.Series(range(100), index=rng)
If we want to resample to the full range of the series:
In [306]: ts.resample("3T").sum() Out[306]: 2014-01-01 00:00:00 0 2014-01-01 00:03:00 0 2014-01-01 00:06:00 0 2014-01-01 00:09:00 0 2014-01-01 00:12:00 0 .. 2014-04-09 23:48:00 0 2014-04-09 23:51:00 0 2014-04-09 23:54:00 0 2014-04-09 23:57:00 0 2014-04-10 00:00:00 99 Freq: 3T, Length: 47521, dtype: int64
We can instead only resample those groups where we have points as follows:
In [307]: from functools import partial In [308]: from pandas.tseries.frequencies import to_offset In [309]: def round(t, freq): .....: freq = to_offset(freq) .....: return pd.Timestamp((t.value // freq.delta.value) * freq.delta.value) .....: In [310]: ts.groupby(partial(round, freq="3T")).sum() Out[310]: 2014-01-01 0 2014-01-02 1 2014-01-03 2 2014-01-04 3 2014-01-05 4 .. 2014-04-06 95 2014-04-07 96 2014-04-08 97 2014-04-09 98 2014-04-10 99 Length: 100, dtype: int64
Similar to the aggregating API, groupby API, and the window API, a Resampler can be selectively resampled.
Resampler
Resampling a DataFrame, the default will be to act on all columns with the same function.
In [311]: df = pd.DataFrame( .....: np.random.randn(1000, 3), .....: index=pd.date_range("1/1/2012", freq="S", periods=1000), .....: columns=["A", "B", "C"], .....: ) .....: In [312]: r = df.resample("3T") In [313]: r.mean() Out[313]: A B C 2012-01-01 00:00:00 -0.033823 -0.121514 -0.081447 2012-01-01 00:03:00 0.056909 0.146731 -0.024320 2012-01-01 00:06:00 -0.058837 0.047046 -0.052021 2012-01-01 00:09:00 0.063123 -0.026158 -0.066533 2012-01-01 00:12:00 0.186340 -0.003144 0.074752 2012-01-01 00:15:00 -0.085954 -0.016287 -0.050046
We can select a specific column or columns using standard getitem.
In [314]: r["A"].mean() Out[314]: 2012-01-01 00:00:00 -0.033823 2012-01-01 00:03:00 0.056909 2012-01-01 00:06:00 -0.058837 2012-01-01 00:09:00 0.063123 2012-01-01 00:12:00 0.186340 2012-01-01 00:15:00 -0.085954 Freq: 3T, Name: A, dtype: float64 In [315]: r[["A", "B"]].mean() Out[315]: A B 2012-01-01 00:00:00 -0.033823 -0.121514 2012-01-01 00:03:00 0.056909 0.146731 2012-01-01 00:06:00 -0.058837 0.047046 2012-01-01 00:09:00 0.063123 -0.026158 2012-01-01 00:12:00 0.186340 -0.003144 2012-01-01 00:15:00 -0.085954 -0.016287
You can pass a list or dict of functions to do aggregation with, outputting a DataFrame:
In [316]: r["A"].agg([np.sum, np.mean, np.std]) Out[316]: sum mean std 2012-01-01 00:00:00 -6.088060 -0.033823 1.043263 2012-01-01 00:03:00 10.243678 0.056909 1.058534 2012-01-01 00:06:00 -10.590584 -0.058837 0.949264 2012-01-01 00:09:00 11.362228 0.063123 1.028096 2012-01-01 00:12:00 33.541257 0.186340 0.884586 2012-01-01 00:15:00 -8.595393 -0.085954 1.035476
On a resampled DataFrame, you can pass a list of functions to apply to each column, which produces an aggregated result with a hierarchical index:
In [317]: r.agg([np.sum, np.mean]) Out[317]: A B C sum mean sum mean sum mean 2012-01-01 00:00:00 -6.088060 -0.033823 -21.872530 -0.121514 -14.660515 -0.081447 2012-01-01 00:03:00 10.243678 0.056909 26.411633 0.146731 -4.377642 -0.024320 2012-01-01 00:06:00 -10.590584 -0.058837 8.468289 0.047046 -9.363825 -0.052021 2012-01-01 00:09:00 11.362228 0.063123 -4.708526 -0.026158 -11.975895 -0.066533 2012-01-01 00:12:00 33.541257 0.186340 -0.565895 -0.003144 13.455299 0.074752 2012-01-01 00:15:00 -8.595393 -0.085954 -1.628689 -0.016287 -5.004580 -0.050046
By passing a dict to aggregate you can apply a different aggregation to the columns of a DataFrame:
aggregate
In [318]: r.agg({"A": np.sum, "B": lambda x: np.std(x, ddof=1)}) Out[318]: A B 2012-01-01 00:00:00 -6.088060 1.001294 2012-01-01 00:03:00 10.243678 1.074597 2012-01-01 00:06:00 -10.590584 0.987309 2012-01-01 00:09:00 11.362228 0.944953 2012-01-01 00:12:00 33.541257 1.095025 2012-01-01 00:15:00 -8.595393 1.035312
The function names can also be strings. In order for a string to be valid it must be implemented on the resampled object:
In [319]: r.agg({"A": "sum", "B": "std"}) Out[319]: A B 2012-01-01 00:00:00 -6.088060 1.001294 2012-01-01 00:03:00 10.243678 1.074597 2012-01-01 00:06:00 -10.590584 0.987309 2012-01-01 00:09:00 11.362228 0.944953 2012-01-01 00:12:00 33.541257 1.095025 2012-01-01 00:15:00 -8.595393 1.035312
Furthermore, you can also specify multiple aggregation functions for each column separately.
In [320]: r.agg({"A": ["sum", "std"], "B": ["mean", "std"]}) Out[320]: A B sum std mean std 2012-01-01 00:00:00 -6.088060 1.043263 -0.121514 1.001294 2012-01-01 00:03:00 10.243678 1.058534 0.146731 1.074597 2012-01-01 00:06:00 -10.590584 0.949264 0.047046 0.987309 2012-01-01 00:09:00 11.362228 1.028096 -0.026158 0.944953 2012-01-01 00:12:00 33.541257 0.884586 -0.003144 1.095025 2012-01-01 00:15:00 -8.595393 1.035476 -0.016287 1.035312
If a DataFrame does not have a datetimelike index, but instead you want to resample based on datetimelike column in the frame, it can passed to the on keyword.
on
In [321]: df = pd.DataFrame( .....: {"date": pd.date_range("2015-01-01", freq="W", periods=5), "a": np.arange(5)}, .....: index=pd.MultiIndex.from_arrays( .....: [[1, 2, 3, 4, 5], pd.date_range("2015-01-01", freq="W", periods=5)], .....: names=["v", "d"], .....: ), .....: ) .....: In [322]: df Out[322]: date a v d 1 2015-01-04 2015-01-04 0 2 2015-01-11 2015-01-11 1 3 2015-01-18 2015-01-18 2 4 2015-01-25 2015-01-25 3 5 2015-02-01 2015-02-01 4 In [323]: df.resample("M", on="date").sum() Out[323]: a date 2015-01-31 6 2015-02-28 4
Similarly, if you instead want to resample by a datetimelike level of MultiIndex, its name or location can be passed to the level keyword.
level
In [324]: df.resample("M", level="d").sum() Out[324]: a d 2015-01-31 6 2015-02-28 4
With the Resampler object in hand, iterating through the grouped data is very natural and functions similarly to itertools.groupby():
itertools.groupby()
In [325]: small = pd.Series( .....: range(6), .....: index=pd.to_datetime( .....: [ .....: "2017-01-01T00:00:00", .....: "2017-01-01T00:30:00", .....: "2017-01-01T00:31:00", .....: "2017-01-01T01:00:00", .....: "2017-01-01T03:00:00", .....: "2017-01-01T03:05:00", .....: ] .....: ), .....: ) .....: In [326]: resampled = small.resample("H") In [327]: for name, group in resampled: .....: print("Group: ", name) .....: print("-" * 27) .....: print(group, end="\n\n") .....: Group: 2017-01-01 00:00:00 --------------------------- 2017-01-01 00:00:00 0 2017-01-01 00:30:00 1 2017-01-01 00:31:00 2 dtype: int64 Group: 2017-01-01 01:00:00 --------------------------- 2017-01-01 01:00:00 3 dtype: int64 Group: 2017-01-01 02:00:00 --------------------------- Series([], dtype: int64) Group: 2017-01-01 03:00:00 --------------------------- 2017-01-01 03:00:00 4 2017-01-01 03:05:00 5 dtype: int64
See Iterating through groups or Resampler.__iter__ for more.
Resampler.__iter__
offset
The bins of the grouping are adjusted based on the beginning of the day of the time series starting point. This works well with frequencies that are multiples of a day (like 30D) or that divide a day evenly (like 90s or 1min). This can create inconsistencies with some frequencies that do not meet this criteria. To change this behavior you can specify a fixed Timestamp with the argument origin.
30D
90s
1min
In [328]: start, end = "2000-10-01 23:30:00", "2000-10-02 00:30:00" In [329]: middle = "2000-10-02 00:00:00" In [330]: rng = pd.date_range(start, end, freq="7min") In [331]: ts = pd.Series(np.arange(len(rng)) * 3, index=rng) In [332]: ts Out[332]: 2000-10-01 23:30:00 0 2000-10-01 23:37:00 3 2000-10-01 23:44:00 6 2000-10-01 23:51:00 9 2000-10-01 23:58:00 12 2000-10-02 00:05:00 15 2000-10-02 00:12:00 18 2000-10-02 00:19:00 21 2000-10-02 00:26:00 24 Freq: 7T, dtype: int64
Here we can see that, when using origin with its default value ('start_day'), the result after '2000-10-02 00:00:00' are not identical depending on the start of time series:
'start_day'
'2000-10-02 00:00:00'
In [333]: ts.resample("17min", origin="start_day").sum() Out[333]: 2000-10-01 23:14:00 0 2000-10-01 23:31:00 9 2000-10-01 23:48:00 21 2000-10-02 00:05:00 54 2000-10-02 00:22:00 24 Freq: 17T, dtype: int64 In [334]: ts[middle:end].resample("17min", origin="start_day").sum() Out[334]: 2000-10-02 00:00:00 33 2000-10-02 00:17:00 45 Freq: 17T, dtype: int64
Here we can see that, when setting origin to 'epoch', the result after '2000-10-02 00:00:00' are identical depending on the start of time series:
'epoch'
In [335]: ts.resample("17min", origin="epoch").sum() Out[335]: 2000-10-01 23:18:00 0 2000-10-01 23:35:00 18 2000-10-01 23:52:00 27 2000-10-02 00:09:00 39 2000-10-02 00:26:00 24 Freq: 17T, dtype: int64 In [336]: ts[middle:end].resample("17min", origin="epoch").sum() Out[336]: 2000-10-01 23:52:00 15 2000-10-02 00:09:00 39 2000-10-02 00:26:00 24 Freq: 17T, dtype: int64
If needed you can use a custom timestamp for origin:
In [337]: ts.resample("17min", origin="2001-01-01").sum() Out[337]: 2000-10-01 23:30:00 9 2000-10-01 23:47:00 21 2000-10-02 00:04:00 54 2000-10-02 00:21:00 24 Freq: 17T, dtype: int64 In [338]: ts[middle:end].resample("17min", origin=pd.Timestamp("2001-01-01")).sum() Out[338]: 2000-10-02 00:04:00 54 2000-10-02 00:21:00 24 Freq: 17T, dtype: int64
If needed you can just adjust the bins with an offset Timedelta that would be added to the default origin. Those two examples are equivalent for this time series:
In [339]: ts.resample("17min", origin="start").sum() Out[339]: 2000-10-01 23:30:00 9 2000-10-01 23:47:00 21 2000-10-02 00:04:00 54 2000-10-02 00:21:00 24 Freq: 17T, dtype: int64 In [340]: ts.resample("17min", offset="23h30min").sum() Out[340]: 2000-10-01 23:30:00 9 2000-10-01 23:47:00 21 2000-10-02 00:04:00 54 2000-10-02 00:21:00 24 Freq: 17T, dtype: int64
Note the use of 'start' for origin on the last example. In that case, origin will be set to the first value of the timeseries.
'start'
Regular intervals of time are represented by Period objects in pandas while sequences of Period objects are collected in a PeriodIndex, which can be created with the convenience function period_range.
A Period represents a span of time (e.g., a day, a month, a quarter, etc). You can specify the span via freq keyword using a frequency alias like below. Because freq represents a span of Period, it cannot be negative like “-3D”.
In [341]: pd.Period("2012", freq="A-DEC") Out[341]: Period('2012', 'A-DEC') In [342]: pd.Period("2012-1-1", freq="D") Out[342]: Period('2012-01-01', 'D') In [343]: pd.Period("2012-1-1 19:00", freq="H") Out[343]: Period('2012-01-01 19:00', 'H') In [344]: pd.Period("2012-1-1 19:00", freq="5H") Out[344]: Period('2012-01-01 19:00', '5H')
Adding and subtracting integers from periods shifts the period by its own frequency. Arithmetic is not allowed between Period with different freq (span).
In [345]: p = pd.Period("2012", freq="A-DEC") In [346]: p + 1 Out[346]: Period('2013', 'A-DEC') In [347]: p - 3 Out[347]: Period('2009', 'A-DEC') In [348]: p = pd.Period("2012-01", freq="2M") In [349]: p + 2 Out[349]: Period('2012-05', '2M') In [350]: p - 1 Out[350]: Period('2011-11', '2M') In [351]: p == pd.Period("2012-01", freq="3M") --------------------------------------------------------------------------- IncompatibleFrequency Traceback (most recent call last) <ipython-input-351-747e98b17e20> in <module> ----> 1 p == pd.Period("2012-01", freq="3M") /pandas/pandas/_libs/tslibs/period.pyx in pandas._libs.tslibs.period._Period.__richcmp__() IncompatibleFrequency: Input has different freq=3M from Period(freq=2M)
If Period freq is daily or higher (D, H, T, S, L, U, N), offsets and timedelta-like can be added if the result can have the same freq. Otherwise, ValueError will be raised.
T
L
U
offsets
In [352]: p = pd.Period("2014-07-01 09:00", freq="H") In [353]: p + pd.offsets.Hour(2) Out[353]: Period('2014-07-01 11:00', 'H') In [354]: p + datetime.timedelta(minutes=120) Out[354]: Period('2014-07-01 11:00', 'H') In [355]: p + np.timedelta64(7200, "s") Out[355]: Period('2014-07-01 11:00', 'H')
In [1]: p + pd.offsets.Minute(5) Traceback ... ValueError: Input has different freq from Period(freq=H)
If Period has other frequencies, only the same offsets can be added. Otherwise, ValueError will be raised.
In [356]: p = pd.Period("2014-07", freq="M") In [357]: p + pd.offsets.MonthEnd(3) Out[357]: Period('2014-10', 'M')
In [1]: p + pd.offsets.MonthBegin(3) Traceback ... ValueError: Input has different freq from Period(freq=M)
Taking the difference of Period instances with the same frequency will return the number of frequency units between them:
In [358]: pd.Period("2012", freq="A-DEC") - pd.Period("2002", freq="A-DEC") Out[358]: <10 * YearEnds: month=12>
Regular sequences of Period objects can be collected in a PeriodIndex, which can be constructed using the period_range convenience function:
In [359]: prng = pd.period_range("1/1/2011", "1/1/2012", freq="M") In [360]: prng Out[360]: PeriodIndex(['2011-01', '2011-02', '2011-03', '2011-04', '2011-05', '2011-06', '2011-07', '2011-08', '2011-09', '2011-10', '2011-11', '2011-12', '2012-01'], dtype='period[M]', freq='M')
The PeriodIndex constructor can also be used directly:
In [361]: pd.PeriodIndex(["2011-1", "2011-2", "2011-3"], freq="M") Out[361]: PeriodIndex(['2011-01', '2011-02', '2011-03'], dtype='period[M]', freq='M')
Passing multiplied frequency outputs a sequence of Period which has multiplied span.
In [362]: pd.period_range(start="2014-01", freq="3M", periods=4) Out[362]: PeriodIndex(['2014-01', '2014-04', '2014-07', '2014-10'], dtype='period[3M]', freq='3M')
If start or end are Period objects, they will be used as anchor endpoints for a PeriodIndex with frequency matching that of the PeriodIndex constructor.
In [363]: pd.period_range( .....: start=pd.Period("2017Q1", freq="Q"), end=pd.Period("2017Q2", freq="Q"), freq="M" .....: ) .....: Out[363]: PeriodIndex(['2017-03', '2017-04', '2017-05', '2017-06'], dtype='period[M]', freq='M')
Just like DatetimeIndex, a PeriodIndex can also be used to index pandas objects:
In [364]: ps = pd.Series(np.random.randn(len(prng)), prng) In [365]: ps Out[365]: 2011-01 -2.916901 2011-02 0.514474 2011-03 1.346470 2011-04 0.816397 2011-05 2.258648 2011-06 0.494789 2011-07 0.301239 2011-08 0.464776 2011-09 -1.393581 2011-10 0.056780 2011-11 0.197035 2011-12 2.261385 2012-01 -0.329583 Freq: M, dtype: float64
PeriodIndex supports addition and subtraction with the same rule as Period.
In [366]: idx = pd.period_range("2014-07-01 09:00", periods=5, freq="H") In [367]: idx Out[367]: PeriodIndex(['2014-07-01 09:00', '2014-07-01 10:00', '2014-07-01 11:00', '2014-07-01 12:00', '2014-07-01 13:00'], dtype='period[H]', freq='H') In [368]: idx + pd.offsets.Hour(2) Out[368]: PeriodIndex(['2014-07-01 11:00', '2014-07-01 12:00', '2014-07-01 13:00', '2014-07-01 14:00', '2014-07-01 15:00'], dtype='period[H]', freq='H') In [369]: idx = pd.period_range("2014-07", periods=5, freq="M") In [370]: idx Out[370]: PeriodIndex(['2014-07', '2014-08', '2014-09', '2014-10', '2014-11'], dtype='period[M]', freq='M') In [371]: idx + pd.offsets.MonthEnd(3) Out[371]: PeriodIndex(['2014-10', '2014-11', '2014-12', '2015-01', '2015-02'], dtype='period[M]', freq='M')
PeriodIndex has its own dtype named period, refer to Period Dtypes.
period
PeriodIndex has a custom period dtype. This is a pandas extension dtype similar to the timezone aware dtype (datetime64[ns, tz]).
The period dtype holds the freq attribute and is represented with period[freq] like period[D] or period[M], using frequency strings.
period[D]
period[M]
In [372]: pi = pd.period_range("2016-01-01", periods=3, freq="M") In [373]: pi Out[373]: PeriodIndex(['2016-01', '2016-02', '2016-03'], dtype='period[M]', freq='M') In [374]: pi.dtype Out[374]: period[M]
The period dtype can be used in .astype(...). It allows one to change the freq of a PeriodIndex like .asfreq() and convert a DatetimeIndex to PeriodIndex like to_period():
.astype(...)
.asfreq()
to_period()
# change monthly freq to daily freq In [375]: pi.astype("period[D]") Out[375]: PeriodIndex(['2016-01-31', '2016-02-29', '2016-03-31'], dtype='period[D]', freq='D') # convert to DatetimeIndex In [376]: pi.astype("datetime64[ns]") Out[376]: DatetimeIndex(['2016-01-01', '2016-02-01', '2016-03-01'], dtype='datetime64[ns]', freq='MS') # convert to PeriodIndex In [377]: dti = pd.date_range("2011-01-01", freq="M", periods=3) In [378]: dti Out[378]: DatetimeIndex(['2011-01-31', '2011-02-28', '2011-03-31'], dtype='datetime64[ns]', freq='M') In [379]: dti.astype("period[M]") Out[379]: PeriodIndex(['2011-01', '2011-02', '2011-03'], dtype='period[M]', freq='M')
PeriodIndex now supports partial string slicing with non-monotonic indexes.
You can pass in dates and strings to Series and DataFrame with PeriodIndex, in the same manner as DatetimeIndex. For details, refer to DatetimeIndex Partial String Indexing.
In [380]: ps["2011-01"] Out[380]: -2.9169013294054507 In [381]: ps[datetime.datetime(2011, 12, 25):] Out[381]: 2011-12 2.261385 2012-01 -0.329583 Freq: M, dtype: float64 In [382]: ps["10/31/2011":"12/31/2011"] Out[382]: 2011-10 0.056780 2011-11 0.197035 2011-12 2.261385 Freq: M, dtype: float64
Passing a string representing a lower frequency than PeriodIndex returns partial sliced data.
In [383]: ps["2011"] Out[383]: 2011-01 -2.916901 2011-02 0.514474 2011-03 1.346470 2011-04 0.816397 2011-05 2.258648 2011-06 0.494789 2011-07 0.301239 2011-08 0.464776 2011-09 -1.393581 2011-10 0.056780 2011-11 0.197035 2011-12 2.261385 Freq: M, dtype: float64 In [384]: dfp = pd.DataFrame( .....: np.random.randn(600, 1), .....: columns=["A"], .....: index=pd.period_range("2013-01-01 9:00", periods=600, freq="T"), .....: ) .....: In [385]: dfp Out[385]: A 2013-01-01 09:00 -0.538468 2013-01-01 09:01 -1.365819 2013-01-01 09:02 -0.969051 2013-01-01 09:03 -0.331152 2013-01-01 09:04 -0.245334 ... ... 2013-01-01 18:55 0.522460 2013-01-01 18:56 0.118710 2013-01-01 18:57 0.167517 2013-01-01 18:58 0.922883 2013-01-01 18:59 1.721104 [600 rows x 1 columns] In [386]: dfp.loc["2013-01-01 10H"] Out[386]: A 2013-01-01 10:00 -0.308975 2013-01-01 10:01 0.542520 2013-01-01 10:02 1.061068 2013-01-01 10:03 0.754005 2013-01-01 10:04 0.352933 ... ... 2013-01-01 10:55 -0.865621 2013-01-01 10:56 -1.167818 2013-01-01 10:57 -2.081748 2013-01-01 10:58 -0.527146 2013-01-01 10:59 0.802298 [60 rows x 1 columns]
As with DatetimeIndex, the endpoints will be included in the result. The example below slices data starting from 10:00 to 11:59.
In [387]: dfp["2013-01-01 10H":"2013-01-01 11H"] Out[387]: A 2013-01-01 10:00 -0.308975 2013-01-01 10:01 0.542520 2013-01-01 10:02 1.061068 2013-01-01 10:03 0.754005 2013-01-01 10:04 0.352933 ... ... 2013-01-01 11:55 -0.590204 2013-01-01 11:56 1.539990 2013-01-01 11:57 -1.224826 2013-01-01 11:58 0.578798 2013-01-01 11:59 -0.685496 [120 rows x 1 columns]
The frequency of Period and PeriodIndex can be converted via the asfreq method. Let’s start with the fiscal year 2011, ending in December:
In [388]: p = pd.Period("2011", freq="A-DEC") In [389]: p Out[389]: Period('2011', 'A-DEC')
We can convert it to a monthly frequency. Using the how parameter, we can specify whether to return the starting or ending month:
how
In [390]: p.asfreq("M", how="start") Out[390]: Period('2011-01', 'M') In [391]: p.asfreq("M", how="end") Out[391]: Period('2011-12', 'M')
The shorthands ‘s’ and ‘e’ are provided for convenience:
In [392]: p.asfreq("M", "s") Out[392]: Period('2011-01', 'M') In [393]: p.asfreq("M", "e") Out[393]: Period('2011-12', 'M')
Converting to a “super-period” (e.g., annual frequency is a super-period of quarterly frequency) automatically returns the super-period that includes the input period:
In [394]: p = pd.Period("2011-12", freq="M") In [395]: p.asfreq("A-NOV") Out[395]: Period('2012', 'A-NOV')
Note that since we converted to an annual frequency that ends the year in November, the monthly period of December 2011 is actually in the 2012 A-NOV period.
Period conversions with anchored frequencies are particularly useful for working with various quarterly data common to economics, business, and other fields. Many organizations define quarters relative to the month in which their fiscal year starts and ends. Thus, first quarter of 2011 could start in 2010 or a few months into 2011. Via anchored frequencies, pandas works for all quarterly frequencies Q-JAN through Q-DEC.
Q-JAN
Q-DEC
Q-DEC define regular calendar quarters:
In [396]: p = pd.Period("2012Q1", freq="Q-DEC") In [397]: p.asfreq("D", "s") Out[397]: Period('2012-01-01', 'D') In [398]: p.asfreq("D", "e") Out[398]: Period('2012-03-31', 'D')
Q-MAR defines fiscal year end in March:
Q-MAR
In [399]: p = pd.Period("2011Q4", freq="Q-MAR") In [400]: p.asfreq("D", "s") Out[400]: Period('2011-01-01', 'D') In [401]: p.asfreq("D", "e") Out[401]: Period('2011-03-31', 'D')
Timestamped data can be converted to PeriodIndex-ed data using to_period and vice-versa using to_timestamp:
to_period
to_timestamp
In [402]: rng = pd.date_range("1/1/2012", periods=5, freq="M") In [403]: ts = pd.Series(np.random.randn(len(rng)), index=rng) In [404]: ts Out[404]: 2012-01-31 1.931253 2012-02-29 -0.184594 2012-03-31 0.249656 2012-04-30 -0.978151 2012-05-31 -0.873389 Freq: M, dtype: float64 In [405]: ps = ts.to_period() In [406]: ps Out[406]: 2012-01 1.931253 2012-02 -0.184594 2012-03 0.249656 2012-04 -0.978151 2012-05 -0.873389 Freq: M, dtype: float64 In [407]: ps.to_timestamp() Out[407]: 2012-01-01 1.931253 2012-02-01 -0.184594 2012-03-01 0.249656 2012-04-01 -0.978151 2012-05-01 -0.873389 Freq: MS, dtype: float64
Remember that ‘s’ and ‘e’ can be used to return the timestamps at the start or end of the period:
In [408]: ps.to_timestamp("D", how="s") Out[408]: 2012-01-01 1.931253 2012-02-01 -0.184594 2012-03-01 0.249656 2012-04-01 -0.978151 2012-05-01 -0.873389 Freq: MS, dtype: float64
Converting between period and timestamp enables some convenient arithmetic functions to be used. In the following example, we convert a quarterly frequency with year ending in November to 9am of the end of the month following the quarter end:
In [409]: prng = pd.period_range("1990Q1", "2000Q4", freq="Q-NOV") In [410]: ts = pd.Series(np.random.randn(len(prng)), prng) In [411]: ts.index = (prng.asfreq("M", "e") + 1).asfreq("H", "s") + 9 In [412]: ts.head() Out[412]: 1990-03-01 09:00 -0.109291 1990-06-01 09:00 -0.637235 1990-09-01 09:00 -1.735925 1990-12-01 09:00 2.096946 1991-03-01 09:00 -1.039926 Freq: H, dtype: float64
If you have data that is outside of the Timestamp bounds, see Timestamp limitations, then you can use a PeriodIndex and/or Series of Periods to do computations.
Periods
In [413]: span = pd.period_range("1215-01-01", "1381-01-01", freq="D") In [414]: span Out[414]: PeriodIndex(['1215-01-01', '1215-01-02', '1215-01-03', '1215-01-04', '1215-01-05', '1215-01-06', '1215-01-07', '1215-01-08', '1215-01-09', '1215-01-10', ... '1380-12-23', '1380-12-24', '1380-12-25', '1380-12-26', '1380-12-27', '1380-12-28', '1380-12-29', '1380-12-30', '1380-12-31', '1381-01-01'], dtype='period[D]', length=60632, freq='D')
To convert from an int64 based YYYYMMDD representation.
int64
In [415]: s = pd.Series([20121231, 20141130, 99991231]) In [416]: s Out[416]: 0 20121231 1 20141130 2 99991231 dtype: int64 In [417]: def conv(x): .....: return pd.Period(year=x // 10000, month=x // 100 % 100, day=x % 100, freq="D") .....: In [418]: s.apply(conv) Out[418]: 0 2012-12-31 1 2014-11-30 2 9999-12-31 dtype: period[D] In [419]: s.apply(conv)[2] Out[419]: Period('9999-12-31', 'D')
These can easily be converted to a PeriodIndex:
In [420]: span = pd.PeriodIndex(s.apply(conv)) In [421]: span Out[421]: PeriodIndex(['2012-12-31', '2014-11-30', '9999-12-31'], dtype='period[D]', freq='D')
pandas provides rich support for working with timestamps in different time zones using the pytz and dateutil libraries or datetime.timezone objects from the standard library.
pytz
datetime.timezone
By default, pandas objects are time zone unaware:
In [422]: rng = pd.date_range("3/6/2012 00:00", periods=15, freq="D") In [423]: rng.tz is None Out[423]: True
To localize these dates to a time zone (assign a particular time zone to a naive date), you can use the tz_localize method or the tz keyword argument in date_range(), Timestamp, or DatetimeIndex. You can either pass pytz or dateutil time zone objects or Olson time zone database strings. Olson time zone strings will return pytz time zone objects by default. To return dateutil time zone objects, append dateutil/ before the string.
tz_localize
dateutil/
In pytz you can find a list of common (and less common) time zones using from pytz import common_timezones, all_timezones.
from pytz import common_timezones, all_timezones
dateutil uses the OS time zones so there isn’t a fixed list available. For common zones, the names are the same as pytz.
In [424]: import dateutil # pytz In [425]: rng_pytz = pd.date_range("3/6/2012 00:00", periods=3, freq="D", tz="Europe/London") In [426]: rng_pytz.tz Out[426]: <DstTzInfo 'Europe/London' LMT-1 day, 23:59:00 STD> # dateutil In [427]: rng_dateutil = pd.date_range("3/6/2012 00:00", periods=3, freq="D") In [428]: rng_dateutil = rng_dateutil.tz_localize("dateutil/Europe/London") In [429]: rng_dateutil.tz Out[429]: tzfile('/usr/share/zoneinfo/Europe/London') # dateutil - utc special case In [430]: rng_utc = pd.date_range( .....: "3/6/2012 00:00", .....: periods=3, .....: freq="D", .....: tz=dateutil.tz.tzutc(), .....: ) .....: In [431]: rng_utc.tz Out[431]: tzutc()
# datetime.timezone In [432]: rng_utc = pd.date_range( .....: "3/6/2012 00:00", .....: periods=3, .....: freq="D", .....: tz=datetime.timezone.utc, .....: ) .....: In [433]: rng_utc.tz Out[433]: datetime.timezone.utc
Note that the UTC time zone is a special case in dateutil and should be constructed explicitly as an instance of dateutil.tz.tzutc. You can also construct other time zones objects explicitly first.
UTC
dateutil.tz.tzutc
In [434]: import pytz # pytz In [435]: tz_pytz = pytz.timezone("Europe/London") In [436]: rng_pytz = pd.date_range("3/6/2012 00:00", periods=3, freq="D") In [437]: rng_pytz = rng_pytz.tz_localize(tz_pytz) In [438]: rng_pytz.tz == tz_pytz Out[438]: True # dateutil In [439]: tz_dateutil = dateutil.tz.gettz("Europe/London") In [440]: rng_dateutil = pd.date_range("3/6/2012 00:00", periods=3, freq="D", tz=tz_dateutil) In [441]: rng_dateutil.tz == tz_dateutil Out[441]: True
To convert a time zone aware pandas object from one time zone to another, you can use the tz_convert method.
tz_convert
In [442]: rng_pytz.tz_convert("US/Eastern") Out[442]: DatetimeIndex(['2012-03-05 19:00:00-05:00', '2012-03-06 19:00:00-05:00', '2012-03-07 19:00:00-05:00'], dtype='datetime64[ns, US/Eastern]', freq=None)
When using pytz time zones, DatetimeIndex will construct a different time zone object than a Timestamp for the same time zone input. A DatetimeIndex can hold a collection of Timestamp objects that may have different UTC offsets and cannot be succinctly represented by one pytz time zone instance while one Timestamp represents one point in time with a specific UTC offset.
In [443]: dti = pd.date_range("2019-01-01", periods=3, freq="D", tz="US/Pacific") In [444]: dti.tz Out[444]: <DstTzInfo 'US/Pacific' LMT-1 day, 16:07:00 STD> In [445]: ts = pd.Timestamp("2019-01-01", tz="US/Pacific") In [446]: ts.tz Out[446]: <DstTzInfo 'US/Pacific' PST-1 day, 16:00:00 STD>
Be wary of conversions between libraries. For some time zones, pytz and dateutil have different definitions of the zone. This is more of a problem for unusual time zones than for ‘standard’ zones like US/Eastern.
US/Eastern
Be aware that a time zone definition across versions of time zone libraries may not be considered equal. This may cause problems when working with stored data that is localized using one version and operated on with a different version. See here for how to handle such a situation.
For pytz time zones, it is incorrect to pass a time zone object directly into the datetime.datetime constructor (e.g., datetime.datetime(2011, 1, 1, tz=pytz.timezone('US/Eastern')). Instead, the datetime needs to be localized using the localize method on the pytz time zone object.
datetime.datetime(2011, 1, 1, tz=pytz.timezone('US/Eastern'))
localize
Be aware that for times in the future, correct conversion between time zones (and UTC) cannot be guaranteed by any time zone library because a timezone’s offset from UTC may be changed by the respective government.
If you are using dates beyond 2038-01-18, due to current deficiencies in the underlying libraries caused by the year 2038 problem, daylight saving time (DST) adjustments to timezone aware dates will not be applied. If and when the underlying libraries are fixed, the DST transitions will be applied.
For example, for two dates that are in British Summer Time (and so would normally be GMT+1), both the following asserts evaluate as true:
In [447]: d_2037 = "2037-03-31T010101" In [448]: d_2038 = "2038-03-31T010101" In [449]: DST = "Europe/London" In [450]: assert pd.Timestamp(d_2037, tz=DST) != pd.Timestamp(d_2037, tz="GMT") In [451]: assert pd.Timestamp(d_2038, tz=DST) == pd.Timestamp(d_2038, tz="GMT")
Under the hood, all timestamps are stored in UTC. Values from a time zone aware DatetimeIndex or Timestamp will have their fields (day, hour, minute, etc.) localized to the time zone. However, timestamps with the same UTC value are still considered to be equal even if they are in different time zones:
In [452]: rng_eastern = rng_utc.tz_convert("US/Eastern") In [453]: rng_berlin = rng_utc.tz_convert("Europe/Berlin") In [454]: rng_eastern[2] Out[454]: Timestamp('2012-03-07 19:00:00-0500', tz='US/Eastern', freq='D') In [455]: rng_berlin[2] Out[455]: Timestamp('2012-03-08 01:00:00+0100', tz='Europe/Berlin', freq='D') In [456]: rng_eastern[2] == rng_berlin[2] Out[456]: True
Operations between Series in different time zones will yield UTC Series, aligning the data on the UTC timestamps:
In [457]: ts_utc = pd.Series(range(3), pd.date_range("20130101", periods=3, tz="UTC")) In [458]: eastern = ts_utc.tz_convert("US/Eastern") In [459]: berlin = ts_utc.tz_convert("Europe/Berlin") In [460]: result = eastern + berlin In [461]: result Out[461]: 2013-01-01 00:00:00+00:00 0 2013-01-02 00:00:00+00:00 2 2013-01-03 00:00:00+00:00 4 Freq: D, dtype: int64 In [462]: result.index Out[462]: DatetimeIndex(['2013-01-01 00:00:00+00:00', '2013-01-02 00:00:00+00:00', '2013-01-03 00:00:00+00:00'], dtype='datetime64[ns, UTC]', freq='D')
To remove time zone information, use tz_localize(None) or tz_convert(None). tz_localize(None) will remove the time zone yielding the local time representation. tz_convert(None) will remove the time zone after converting to UTC time.
tz_localize(None)
tz_convert(None)
In [463]: didx = pd.date_range(start="2014-08-01 09:00", freq="H", periods=3, tz="US/Eastern") In [464]: didx Out[464]: DatetimeIndex(['2014-08-01 09:00:00-04:00', '2014-08-01 10:00:00-04:00', '2014-08-01 11:00:00-04:00'], dtype='datetime64[ns, US/Eastern]', freq='H') In [465]: didx.tz_localize(None) Out[465]: DatetimeIndex(['2014-08-01 09:00:00', '2014-08-01 10:00:00', '2014-08-01 11:00:00'], dtype='datetime64[ns]', freq=None) In [466]: didx.tz_convert(None) Out[466]: DatetimeIndex(['2014-08-01 13:00:00', '2014-08-01 14:00:00', '2014-08-01 15:00:00'], dtype='datetime64[ns]', freq='H') # tz_convert(None) is identical to tz_convert('UTC').tz_localize(None) In [467]: didx.tz_convert("UTC").tz_localize(None) Out[467]: DatetimeIndex(['2014-08-01 13:00:00', '2014-08-01 14:00:00', '2014-08-01 15:00:00'], dtype='datetime64[ns]', freq=None)
For ambiguous times, pandas supports explicitly specifying the keyword-only fold argument. Due to daylight saving time, one wall clock time can occur twice when shifting from summer to winter time; fold describes whether the datetime-like corresponds to the first (0) or the second time (1) the wall clock hits the ambiguous time. Fold is supported only for constructing from naive datetime.datetime (see datetime documentation for details) or from Timestamp or for constructing from components (see below). Only dateutil timezones are supported (see dateutil documentation for dateutil methods that deal with ambiguous datetimes) as pytz timezones do not support fold (see pytz documentation for details on how pytz deals with ambiguous datetimes). To localize an ambiguous datetime with pytz, please use Timestamp.tz_localize(). In general, we recommend to rely on Timestamp.tz_localize() when localizing ambiguous datetimes if you need direct control over how they are handled.
Timestamp.tz_localize()
In [468]: pd.Timestamp( .....: datetime.datetime(2019, 10, 27, 1, 30, 0, 0), .....: tz="dateutil/Europe/London", .....: fold=0, .....: ) .....: Out[468]: Timestamp('2019-10-27 01:30:00+0100', tz='dateutil//usr/share/zoneinfo/Europe/London') In [469]: pd.Timestamp( .....: year=2019, .....: month=10, .....: day=27, .....: hour=1, .....: minute=30, .....: tz="dateutil/Europe/London", .....: fold=1, .....: ) .....: Out[469]: Timestamp('2019-10-27 01:30:00+0000', tz='dateutil//usr/share/zoneinfo/Europe/London')
tz_localize may not be able to determine the UTC offset of a timestamp because daylight savings time (DST) in a local time zone causes some times to occur twice within one day (“clocks fall back”). The following options are available:
'raise': Raises a pytz.AmbiguousTimeError (the default behavior)
'raise'
pytz.AmbiguousTimeError
'infer': Attempt to determine the correct offset base on the monotonicity of the timestamps
'infer'
'NaT': Replaces ambiguous times with NaT
'NaT'
bool: True represents a DST time, False represents non-DST time. An array-like of bool values is supported for a sequence of times.
bool
True
False
In [470]: rng_hourly = pd.DatetimeIndex( .....: ["11/06/2011 00:00", "11/06/2011 01:00", "11/06/2011 01:00", "11/06/2011 02:00"] .....: ) .....:
This will fail as there are ambiguous times ('11/06/2011 01:00')
'11/06/2011 01:00'
In [2]: rng_hourly.tz_localize('US/Eastern') AmbiguousTimeError: Cannot infer dst time from Timestamp('2011-11-06 01:00:00'), try using the 'ambiguous' argument
Handle these ambiguous times by specifying the following.
In [471]: rng_hourly.tz_localize("US/Eastern", ambiguous="infer") Out[471]: DatetimeIndex(['2011-11-06 00:00:00-04:00', '2011-11-06 01:00:00-04:00', '2011-11-06 01:00:00-05:00', '2011-11-06 02:00:00-05:00'], dtype='datetime64[ns, US/Eastern]', freq=None) In [472]: rng_hourly.tz_localize("US/Eastern", ambiguous="NaT") Out[472]: DatetimeIndex(['2011-11-06 00:00:00-04:00', 'NaT', 'NaT', '2011-11-06 02:00:00-05:00'], dtype='datetime64[ns, US/Eastern]', freq=None) In [473]: rng_hourly.tz_localize("US/Eastern", ambiguous=[True, True, False, False]) Out[473]: DatetimeIndex(['2011-11-06 00:00:00-04:00', '2011-11-06 01:00:00-04:00', '2011-11-06 01:00:00-05:00', '2011-11-06 02:00:00-05:00'], dtype='datetime64[ns, US/Eastern]', freq=None)
A DST transition may also shift the local time ahead by 1 hour creating nonexistent local times (“clocks spring forward”). The behavior of localizing a timeseries with nonexistent times can be controlled by the nonexistent argument. The following options are available:
nonexistent
'raise': Raises a pytz.NonExistentTimeError (the default behavior)
pytz.NonExistentTimeError
'NaT': Replaces nonexistent times with NaT
'shift_forward': Shifts nonexistent times forward to the closest real time
'shift_forward'
'shift_backward': Shifts nonexistent times backward to the closest real time
'shift_backward'
timedelta object: Shifts nonexistent times by the timedelta duration
In [474]: dti = pd.date_range(start="2015-03-29 02:30:00", periods=3, freq="H") # 2:30 is a nonexistent time
Localization of nonexistent times will raise an error by default.
In [2]: dti.tz_localize('Europe/Warsaw') NonExistentTimeError: 2015-03-29 02:30:00
Transform nonexistent times to NaT or shift the times.
In [475]: dti Out[475]: DatetimeIndex(['2015-03-29 02:30:00', '2015-03-29 03:30:00', '2015-03-29 04:30:00'], dtype='datetime64[ns]', freq='H') In [476]: dti.tz_localize("Europe/Warsaw", nonexistent="shift_forward") Out[476]: DatetimeIndex(['2015-03-29 03:00:00+02:00', '2015-03-29 03:30:00+02:00', '2015-03-29 04:30:00+02:00'], dtype='datetime64[ns, Europe/Warsaw]', freq=None) In [477]: dti.tz_localize("Europe/Warsaw", nonexistent="shift_backward") Out[477]: DatetimeIndex(['2015-03-29 01:59:59.999999999+01:00', '2015-03-29 03:30:00+02:00', '2015-03-29 04:30:00+02:00'], dtype='datetime64[ns, Europe/Warsaw]', freq=None) In [478]: dti.tz_localize("Europe/Warsaw", nonexistent=pd.Timedelta(1, unit="H")) Out[478]: DatetimeIndex(['2015-03-29 03:30:00+02:00', '2015-03-29 03:30:00+02:00', '2015-03-29 04:30:00+02:00'], dtype='datetime64[ns, Europe/Warsaw]', freq=None) In [479]: dti.tz_localize("Europe/Warsaw", nonexistent="NaT") Out[479]: DatetimeIndex(['NaT', '2015-03-29 03:30:00+02:00', '2015-03-29 04:30:00+02:00'], dtype='datetime64[ns, Europe/Warsaw]', freq=None)
A Series with time zone naive values is represented with a dtype of datetime64[ns].
In [480]: s_naive = pd.Series(pd.date_range("20130101", periods=3)) In [481]: s_naive Out[481]: 0 2013-01-01 1 2013-01-02 2 2013-01-03 dtype: datetime64[ns]
A Series with a time zone aware values is represented with a dtype of datetime64[ns, tz] where tz is the time zone
In [482]: s_aware = pd.Series(pd.date_range("20130101", periods=3, tz="US/Eastern")) In [483]: s_aware Out[483]: 0 2013-01-01 00:00:00-05:00 1 2013-01-02 00:00:00-05:00 2 2013-01-03 00:00:00-05:00 dtype: datetime64[ns, US/Eastern]
Both of these Series time zone information can be manipulated via the .dt accessor, see the dt accessor section.
For example, to localize and convert a naive stamp to time zone aware.
In [484]: s_naive.dt.tz_localize("UTC").dt.tz_convert("US/Eastern") Out[484]: 0 2012-12-31 19:00:00-05:00 1 2013-01-01 19:00:00-05:00 2 2013-01-02 19:00:00-05:00 dtype: datetime64[ns, US/Eastern]
Time zone information can also be manipulated using the astype method. This method can localize and convert time zone naive timestamps or convert time zone aware timestamps.
astype
# localize and convert a naive time zone In [485]: s_naive.astype("datetime64[ns, US/Eastern]") Out[485]: 0 2012-12-31 19:00:00-05:00 1 2013-01-01 19:00:00-05:00 2 2013-01-02 19:00:00-05:00 dtype: datetime64[ns, US/Eastern] # make an aware tz naive In [486]: s_aware.astype("datetime64[ns]") Out[486]: 0 2013-01-01 05:00:00 1 2013-01-02 05:00:00 2 2013-01-03 05:00:00 dtype: datetime64[ns] # convert to a new time zone In [487]: s_aware.astype("datetime64[ns, CET]") Out[487]: 0 2013-01-01 06:00:00+01:00 1 2013-01-02 06:00:00+01:00 2 2013-01-03 06:00:00+01:00 dtype: datetime64[ns, CET]
Using Series.to_numpy() on a Series, returns a NumPy array of the data. NumPy does not currently support time zones (even though it is printing in the local time zone!), therefore an object array of Timestamps is returned for time zone aware data:
Series.to_numpy()
In [488]: s_naive.to_numpy() Out[488]: array(['2013-01-01T00:00:00.000000000', '2013-01-02T00:00:00.000000000', '2013-01-03T00:00:00.000000000'], dtype='datetime64[ns]') In [489]: s_aware.to_numpy() Out[489]: array([Timestamp('2013-01-01 00:00:00-0500', tz='US/Eastern', freq='D'), Timestamp('2013-01-02 00:00:00-0500', tz='US/Eastern', freq='D'), Timestamp('2013-01-03 00:00:00-0500', tz='US/Eastern', freq='D')], dtype=object)
By converting to an object array of Timestamps, it preserves the time zone information. For example, when converting back to a Series:
In [490]: pd.Series(s_aware.to_numpy()) Out[490]: 0 2013-01-01 00:00:00-05:00 1 2013-01-02 00:00:00-05:00 2 2013-01-03 00:00:00-05:00 dtype: datetime64[ns, US/Eastern]
However, if you want an actual NumPy datetime64[ns] array (with the values converted to UTC) instead of an array of objects, you can specify the dtype argument:
dtype
In [491]: s_aware.to_numpy(dtype="datetime64[ns]") Out[491]: array(['2013-01-01T05:00:00.000000000', '2013-01-02T05:00:00.000000000', '2013-01-03T05:00:00.000000000'], dtype='datetime64[ns]')