Time series / date functionality

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.

For example, pandas supports:

Parsing time series information from various sources and formats

  1. In [1]: import datetime
  3. In [2]: dti = pd.to_datetime(['1/1/2018', np.datetime64('2018-01-01'),
  4.    ...:                       datetime.datetime(2018, 1, 1)])
  5.    ...: 
  7. In [3]: dti
  8. 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

  1. In [4]: dti = pd.date_range('2018-01-01', periods=3, freq='H')
  3. In [5]: dti
  4. Out[5]: 
  5. DatetimeIndex(['2018-01-01 00:00:00', '2018-01-01 01:00:00',
  6.                '2018-01-01 02:00:00'],
  7.               dtype='datetime64[ns]', freq='H')

Manipulating and converting date times with timezone information

  1. In [6]: dti = dti.tz_localize('UTC')
  3. In [7]: dti
  4. Out[7]: 
  5. DatetimeIndex(['2018-01-01 00:00:00+00:00', '2018-01-01 01:00:00+00:00',
  6.                '2018-01-01 02:00:00+00:00'],
  7.               dtype='datetime64[ns, UTC]', freq='H')
  9. In [8]: dti.tz_convert('US/Pacific')
  10. Out[8]: 
  11. DatetimeIndex(['2017-12-31 16:00:00-08:00', '2017-12-31 17:00:00-08:00',
  12.                '2017-12-31 18:00:00-08:00'],
  13.               dtype='datetime64[ns, US/Pacific]', freq='H')

Resampling or converting a time series to a particular frequency

  1. In [9]: idx = pd.date_range('2018-01-01', periods=5, freq='H')
  3. In [10]: ts = pd.Series(range(len(idx)), index=idx)
  5. In [11]: ts
  6. Out[11]: 
  7. 2018-01-01 00:00:00    0
  8. 2018-01-01 01:00:00    1
  9. 2018-01-01 02:00:00    2
  10. 2018-01-01 03:00:00    3
  11. 2018-01-01 04:00:00    4
  12. Freq: H, dtype: int64
  14. In [12]: ts.resample('2H').mean()
  15. Out[12]: 
  16. 2018-01-01 00:00:00    0.5
  17. 2018-01-01 02:00:00    2.5
  18. 2018-01-01 04:00:00    4.0
  19. Freq: 2H, dtype: float64

Performing date and time arithmetic with absolute or relative time increments

  1. In [13]: friday = pd.Timestamp('2018-01-05')
  3. In [14]: friday.day_name()
  4. Out[14]: 'Friday'
  6. # Add 1 day
  7. In [15]: saturday = friday + pd.Timedelta('1 day')
  9. In [16]: saturday.day_name()
  10. Out[16]: 'Saturday'
  12. # Add 1 business day (Friday --> Monday)
  13. In [17]: monday = friday + pd.offsets.BDay()
  15. In [18]: monday.day_name()
  16. Out[18]: 'Monday'

pandas provides a relatively compact and self-contained set of tools for performing the above tasks and more.

Overview

pandas captures 4 general time related concepts:

  1. Date times: A specific date and time with timezone support. Similar to datetime.datetime from the standard library.
  2. Time deltas: An absolute time duration. Similar to datetime.timedelta from the standard library.
  3. Time spans: A span of time defined by a point in time and its associated frequency.
  4. Date offsets: A relative time duration that respects calendar arithmetic. Similar to dateutil.relativedelta.relativedelta from the dateutil package.
Concept Scalar Class Array Class pandas Data Type Primary Creation Method
Date times Timestamp DatetimeIndex datetime64[ns] or datetime64[ns, tz] to_datetime or date_range
Time deltas Timedelta TimedeltaIndex timedelta64[ns] to_timedelta or timedelta_range
Time spans Period PeriodIndex period[freq] Period or period_range
Date offsets DateOffset None None DateOffset

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.

  1. In [19]: pd.Series(range(3), index=pd.date_range('2000', freq='D', periods=3))
  2. Out[19]: 
  3. 2000-01-01    0
  4. 2000-01-02    1
  5. 2000-01-03    2
  6. Freq: D, dtype: int64

However, Series and DataFrame can directly also support the time component as data itself.

  1. In [20]: pd.Series(pd.date_range('2000', freq='D', periods=3))
  2. Out[20]: 
  3. 0   2000-01-01
  4. 1   2000-01-02
  5. 2   2000-01-03
  6. 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.

  1. In [21]: pd.Series(pd.period_range('1/1/2011', freq='M', periods=3))
  2. Out[21]: 
  3. 0    2011-01
  4. 1    2011-02
  5. 2    2011-03
  6. dtype: period[M]
  8. In [22]: pd.Series([pd.DateOffset(1), pd.DateOffset(2)])
  9. Out[22]: 
  10. 0         <DateOffset>
  11. 1    <2 * DateOffsets>
  12. dtype: object
  14. In [23]: pd.Series(pd.date_range('1/1/2011', freq='M', periods=3))
  15. Out[23]: 
  16. 0   2011-01-31
  17. 1   2011-02-28
  18. 2   2011-03-31
  19. 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.

  1. In [24]: pd.Timestamp(pd.NaT)
  2. Out[24]: NaT
  4. In [25]: pd.Timedelta(pd.NaT)
  5. Out[25]: NaT
  7. In [26]: pd.Period(pd.NaT)
  8. Out[26]: NaT
  10. # Equality acts as np.nan would
  11. In [27]: pd.NaT == pd.NaT
  12. Out[27]: False

Timestamps vs. Time Spans

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.

  1. In [28]: pd.Timestamp(datetime.datetime(2012, 5, 1))
  2. Out[28]: Timestamp('2012-05-01 00:00:00')
  4. In [29]: pd.Timestamp('2012-05-01')
  5. Out[29]: Timestamp('2012-05-01 00:00:00')
  7. In [30]: pd.Timestamp(2012, 5, 1)
  8. 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:

  1. In [31]: pd.Period('2011-01')
  2. Out[31]: Period('2011-01', 'M')
  4. In [32]: pd.Period('2012-05', freq='D')
  5. 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.

  1. In [33]: dates = [pd.Timestamp('2012-05-01'),
  2.    ....:          pd.Timestamp('2012-05-02'),
  3.    ....:          pd.Timestamp('2012-05-03')]
  4.    ....: 
  6. In [34]: ts = pd.Series(np.random.randn(3), dates)
  8. In [35]: type(ts.index)
  9. Out[35]: pandas.core.indexes.datetimes.DatetimeIndex
  11. In [36]: ts.index
  12. Out[36]: DatetimeIndex(['2012-05-01', '2012-05-02', '2012-05-03'], dtype='datetime64[ns]', freq=None)
  14. In [37]: ts
  15. Out[37]: 
  16. 2012-05-01    0.469112
  17. 2012-05-02   -0.282863
  18. 2012-05-03   -1.509059
  19. dtype: float64
  21. In [38]: periods = [pd.Period('2012-01'), pd.Period('2012-02'), pd.Period('2012-03')]
  23. In [39]: ts = pd.Series(np.random.randn(3), periods)
  25. In [40]: type(ts.index)
  26. Out[40]: pandas.core.indexes.period.PeriodIndex
  28. In [41]: ts.index
  29. Out[41]: PeriodIndex(['2012-01', '2012-02', '2012-03'], dtype='period[M]', freq='M')
  31. In [42]: ts
  32. Out[42]: 
  33. 2012-01   -1.135632
  34. 2012-02    1.212112
  35. 2012-03   -0.173215
  36. 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.

Converting to timestamps

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:

  1. In [43]: pd.to_datetime(pd.Series(['Jul 31, 2009', '2010-01-10', None]))
  2. Out[43]: 
  3. 0   2009-07-31
  4. 1   2010-01-10
  5. 2          NaT
  6. dtype: datetime64[ns]
  8. In [44]: pd.to_datetime(['2005/11/23', '2010.12.31'])
  9. 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:

  1. In [45]: pd.to_datetime(['04-01-2012 10:00'], dayfirst=True)
  2. Out[45]: DatetimeIndex(['2012-01-04 10:00:00'], dtype='datetime64[ns]', freq=None)
  4. In [46]: pd.to_datetime(['14-01-2012', '01-14-2012'], dayfirst=True)
  5. Out[46]: DatetimeIndex(['2012-01-14', '2012-01-14'], dtype='datetime64[ns]', freq=None)

::: danger 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.

  1. In [47]: pd.to_datetime('2010/11/12')
  2. Out[47]: Timestamp('2010-11-12 00:00:00')
  4. In [48]: pd.Timestamp('2010/11/12')
  5. Out[48]: Timestamp('2010-11-12 00:00:00')

You can also use the DatetimeIndex constructor directly:

  1. In [49]: pd.DatetimeIndex(['2018-01-01', '2018-01-03', '2018-01-05'])
  2. 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:

  1. In [50]: pd.DatetimeIndex(['2018-01-01', '2018-01-03', '2018-01-05'], freq='infer')
  2. Out[50]: DatetimeIndex(['2018-01-01', '2018-01-03', '2018-01-05'], dtype='datetime64[ns]', freq='2D')

Providing a format argument

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.

  1. In [51]: pd.to_datetime('2010/11/12', format='%Y/%m/%d')
  2. Out[51]: Timestamp('2010-11-12 00:00:00')
  4. In [52]: pd.to_datetime('12-11-2010 00:00', format='%d-%m-%Y %H:%M')
  5. 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.

Assembling datetime from multiple DataFrame columns

New in version 0.18.1.

You can also pass a DataFrame of integer or string columns to assemble into a Series of Timestamps.

  1. In [53]: df = pd.DataFrame({'year': [2015, 2016],
  2.    ....:                    'month': [2, 3],
  3.    ....:                    'day': [4, 5],
  4.    ....:                    'hour': [2, 3]})
  5.    ....: 
  7. In [54]: pd.to_datetime(df)
  8. Out[54]: 
  9. 0   2015-02-04 02:00:00
  10. 1   2016-03-05 03:00:00
  11. dtype: datetime64[ns]

You can pass only the columns that you need to assemble.

  1. In [55]: pd.to_datetime(df[['year', 'month', 'day']])
  2. Out[55]: 
  3. 0   2015-02-04
  4. 1   2016-03-05
  5. dtype: datetime64[ns]

pd.to_datetime looks for standard designations of the datetime component in the column names, including:

  • required: year, month, day
  • optional: hour, minute, second, millisecond, microsecond, nanosecond

Invalid data

The default behavior, errors='raise', is to raise when unparseable:

  1. In [2]: pd.to_datetime(['2009/07/31', 'asd'], errors='raise')
  2. ValueError: Unknown string format

Pass errors='ignore' to return the original input when unparseable:

  1. In [56]: pd.to_datetime(['2009/07/31', 'asd'], errors='ignore')
  2. Out[56]: Index(['2009/07/31', 'asd'], dtype='object')

Pass errors='coerce' to convert unparseable data to NaT (not a time):

  1. In [57]: pd.to_datetime(['2009/07/31', 'asd'], errors='coerce')
  2. Out[57]: DatetimeIndex(['2009-07-31', 'NaT'], dtype='datetime64[ns]', freq=None)

Epoch timestamps

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.

  1. In [58]: pd.to_datetime([1349720105, 1349806505, 1349892905,
  2.    ....:                 1349979305, 1350065705], unit='s')
  3.    ....: 
  4. Out[58]: 
  5. DatetimeIndex(['2012-10-08 18:15:05', '2012-10-09 18:15:05',
  6.                '2012-10-10 18:15:05', '2012-10-11 18:15:05',
  7.                '2012-10-12 18:15:05'],
  8.               dtype='datetime64[ns]', freq=None)
  10. In [59]: pd.to_datetime([1349720105100, 1349720105200, 1349720105300,
  11.    ....:                 1349720105400, 1349720105500], unit='ms')
  12.    ....: 
  13. Out[59]: 
  14. DatetimeIndex(['2012-10-08 18:15:05.100000', '2012-10-08 18:15:05.200000',
  15.                '2012-10-08 18:15:05.300000', '2012-10-08 18:15:05.400000',
  16.                '2012-10-08 18:15:05.500000'],
  17.               dtype='datetime64[ns]', freq=None)

Constructing a Timestamp or DatetimeIndex with an epoch timestamp with the tz argument specified will currently localize the epoch timestamps to UTC first then convert the result to the specified time zone. However, this behavior is deprecated, and if you have epochs in wall time in another timezone, it is recommended to read the epochs as timezone-naive timestamps and then localize to the appropriate timezone:

  1. In [60]: pd.Timestamp(1262347200000000000).tz_localize('US/Pacific')
  2. Out[60]: Timestamp('2010-01-01 12:00:00-0800', tz='US/Pacific')
  4. In [61]: pd.DatetimeIndex([1262347200000000000]).tz_localize('US/Pacific')
  5. Out[61]: DatetimeIndex(['2010-01-01 12:00:00-08:00'], dtype='datetime64[ns, US/Pacific]', freq=None)

::: tip Note

Epoch times will be rounded to the nearest nanosecond.

:::

::: danger Warning

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).

  1. In [62]: pd.to_datetime([1490195805.433, 1490195805.433502912], unit='s')
  2. Out[62]: DatetimeIndex(['2017-03-22 15:16:45.433000088', '2017-03-22 15:16:45.433502913'], dtype='datetime64[ns]', freq=None)
  4. In [63]: pd.to_datetime(1490195805433502912, unit='ns')
  5. Out[63]: Timestamp('2017-03-22 15:16:45.433502912')

:::

Using the origin Parameter

From timestamps to epoch

To invert the operation from above, namely, to convert from a Timestamp to a ‘unix’ epoch:

  1. In [64]: stamps = pd.date_range('2012-10-08 18:15:05', periods=4, freq='D')
  3. In [65]: stamps
  4. Out[65]: 
  5. DatetimeIndex(['2012-10-08 18:15:05', '2012-10-09 18:15:05',
  6.                '2012-10-10 18:15:05', '2012-10-11 18:15:05'],
  7.               dtype='datetime64[ns]', freq='D')

We subtract the epoch (midnight at January 1, 1970 UTC) and then floor divide by the “unit” (1 second).

  1. In [66]: (stamps - pd.Timestamp("1970-01-01")) // pd.Timedelta('1s')
  2. Out[66]: Int64Index([1349720105, 1349806505, 1349892905, 1349979305], dtype='int64')

Using the origin Parameter

New in version 0.20.0.

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:

  1. In [67]: pd.to_datetime([1, 2, 3], unit='D', origin=pd.Timestamp('1960-01-01'))
  2. 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.

  1. In [68]: pd.to_datetime([1, 2, 3], unit='D')
  2. Out[68]: DatetimeIndex(['1970-01-02', '1970-01-03', '1970-01-04'], dtype='datetime64[ns]', freq=None)

Generating ranges of timestamps

To generate an index with timestamps, you can use either the DatetimeIndex or Index constructor and pass in a list of datetime objects:

  1. In [69]: dates = [datetime.datetime(2012, 5, 1),
  2.    ....:          datetime.datetime(2012, 5, 2),
  3.    ....:          datetime.datetime(2012, 5, 3)]
  4.    ....: 
  6. # Note the frequency information
  7. In [70]: index = pd.DatetimeIndex(dates)
  9. In [71]: index
  10. Out[71]: DatetimeIndex(['2012-05-01', '2012-05-02', '2012-05-03'], dtype='datetime64[ns]', freq=None)
  12. # Automatically converted to DatetimeIndex
  13. In [72]: index = pd.Index(dates)
  15. In [73]: index
  16. 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:

  1. In [74]: start = datetime.datetime(2011, 1, 1)
  3. In [75]: end = datetime.datetime(2012, 1, 1)
  5. In [76]: index = pd.date_range(start, end)
  7. In [77]: index
  8. Out[77]: 
  9. DatetimeIndex(['2011-01-01', '2011-01-02', '2011-01-03', '2011-01-04',
  10.                '2011-01-05', '2011-01-06', '2011-01-07', '2011-01-08',
  11.                '2011-01-09', '2011-01-10',
  12.                ...
  13.                '2011-12-23', '2011-12-24', '2011-12-25', '2011-12-26',
  14.                '2011-12-27', '2011-12-28', '2011-12-29', '2011-12-30',
  15.                '2011-12-31', '2012-01-01'],
  16.               dtype='datetime64[ns]', length=366, freq='D')
  18. In [78]: index = pd.bdate_range(start, end)
  20. In [79]: index
  21. Out[79]: 
  22. DatetimeIndex(['2011-01-03', '2011-01-04', '2011-01-05', '2011-01-06',
  23.                '2011-01-07', '2011-01-10', '2011-01-11', '2011-01-12',
  24.                '2011-01-13', '2011-01-14',
  25.                ...
  26.                '2011-12-19', '2011-12-20', '2011-12-21', '2011-12-22',
  27.                '2011-12-23', '2011-12-26', '2011-12-27', '2011-12-28',
  28.                '2011-12-29', '2011-12-30'],
  29.               dtype='datetime64[ns]', length=260, freq='B')

Convenience functions like date_range and bdate_range can utilize a variety of frequency aliases:

  1. In [80]: pd.date_range(start, periods=1000, freq='M')
  2. Out[80]: 
  3. DatetimeIndex(['2011-01-31', '2011-02-28', '2011-03-31', '2011-04-30',
  4.                '2011-05-31', '2011-06-30', '2011-07-31', '2011-08-31',
  5.                '2011-09-30', '2011-10-31',
  6.                ...
  7.                '2093-07-31', '2093-08-31', '2093-09-30', '2093-10-31',
  8.                '2093-11-30', '2093-12-31', '2094-01-31', '2094-02-28',
  9.                '2094-03-31', '2094-04-30'],
  10.               dtype='datetime64[ns]', length=1000, freq='M')
  12. In [81]: pd.bdate_range(start, periods=250, freq='BQS')
  13. Out[81]: 
  14. DatetimeIndex(['2011-01-03', '2011-04-01', '2011-07-01', '2011-10-03',
  15.                '2012-01-02', '2012-04-02', '2012-07-02', '2012-10-01',
  16.                '2013-01-01', '2013-04-01',
  17.                ...
  18.                '2071-01-01', '2071-04-01', '2071-07-01', '2071-10-01',
  19.                '2072-01-01', '2072-04-01', '2072-07-01', '2072-10-03',
  20.                '2073-01-02', '2073-04-03'],
  21.               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:

  1. In [82]: pd.date_range(start, end, freq='BM')
  2. Out[82]: 
  3. DatetimeIndex(['2011-01-31', '2011-02-28', '2011-03-31', '2011-04-29',
  4.                '2011-05-31', '2011-06-30', '2011-07-29', '2011-08-31',
  5.                '2011-09-30', '2011-10-31', '2011-11-30', '2011-12-30'],
  6.               dtype='datetime64[ns]', freq='BM')
  8. In [83]: pd.date_range(start, end, freq='W')
  9. Out[83]: 
  10. DatetimeIndex(['2011-01-02', '2011-01-09', '2011-01-16', '2011-01-23',
  11.                '2011-01-30', '2011-02-06', '2011-02-13', '2011-02-20',
  12.                '2011-02-27', '2011-03-06', '2011-03-13', '2011-03-20',
  13.                '2011-03-27', '2011-04-03', '2011-04-10', '2011-04-17',
  14.                '2011-04-24', '2011-05-01', '2011-05-08', '2011-05-15',
  15.                '2011-05-22', '2011-05-29', '2011-06-05', '2011-06-12',
  16.                '2011-06-19', '2011-06-26', '2011-07-03', '2011-07-10',
  17.                '2011-07-17', '2011-07-24', '2011-07-31', '2011-08-07',
  18.                '2011-08-14', '2011-08-21', '2011-08-28', '2011-09-04',
  19.                '2011-09-11', '2011-09-18', '2011-09-25', '2011-10-02',
  20.                '2011-10-09', '2011-10-16', '2011-10-23', '2011-10-30',
  21.                '2011-11-06', '2011-11-13', '2011-11-20', '2011-11-27',
  22.                '2011-12-04', '2011-12-11', '2011-12-18', '2011-12-25',
  23.                '2012-01-01'],
  24.               dtype='datetime64[ns]', freq='W-SUN')
  26. In [84]: pd.bdate_range(end=end, periods=20)
  27. Out[84]: 
  28. DatetimeIndex(['2011-12-05', '2011-12-06', '2011-12-07', '2011-12-08',
  29.                '2011-12-09', '2011-12-12', '2011-12-13', '2011-12-14',
  30.                '2011-12-15', '2011-12-16', '2011-12-19', '2011-12-20',
  31.                '2011-12-21', '2011-12-22', '2011-12-23', '2011-12-26',
  32.                '2011-12-27', '2011-12-28', '2011-12-29', '2011-12-30'],
  33.               dtype='datetime64[ns]', freq='B')
  35. In [85]: pd.bdate_range(start=start, periods=20)
  36. Out[85]: 
  37. DatetimeIndex(['2011-01-03', '2011-01-04', '2011-01-05', '2011-01-06',
  38.                '2011-01-07', '2011-01-10', '2011-01-11', '2011-01-12',
  39.                '2011-01-13', '2011-01-14', '2011-01-17', '2011-01-18',
  40.                '2011-01-19', '2011-01-20', '2011-01-21', '2011-01-24',
  41.                '2011-01-25', '2011-01-26', '2011-01-27', '2011-01-28'],
  42.               dtype='datetime64[ns]', freq='B')

New in version 0.23.0.

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:

  1. In [86]: pd.date_range('2018-01-01', '2018-01-05', periods=5)
  2. Out[86]: 
  3. DatetimeIndex(['2018-01-01', '2018-01-02', '2018-01-03', '2018-01-04',
  4.                '2018-01-05'],
  5.               dtype='datetime64[ns]', freq=None)
  7. In [87]: pd.date_range('2018-01-01', '2018-01-05', periods=10)
  8. Out[87]: 
  9. DatetimeIndex(['2018-01-01 00:00:00', '2018-01-01 10:40:00',
  10.                '2018-01-01 21:20:00', '2018-01-02 08:00:00',
  11.                '2018-01-02 18:40:00', '2018-01-03 05:20:00',
  12.                '2018-01-03 16:00:00', '2018-01-04 02:40:00',
  13.                '2018-01-04 13:20:00', '2018-01-05 00:00:00'],
  14.               dtype='datetime64[ns]', freq=None)

Custom frequency ranges

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.

  1. In [88]: weekmask = 'Mon Wed Fri'
  3. In [89]: holidays = [datetime.datetime(2011, 1, 5), datetime.datetime(2011, 3, 14)]
  5. In [90]: pd.bdate_range(start, end, freq='C', weekmask=weekmask, holidays=holidays)
  6. Out[90]: 
  7. DatetimeIndex(['2011-01-03', '2011-01-07', '2011-01-10', '2011-01-12',
  8.                '2011-01-14', '2011-01-17', '2011-01-19', '2011-01-21',
  9.                '2011-01-24', '2011-01-26',
  10.                ...
  11.                '2011-12-09', '2011-12-12', '2011-12-14', '2011-12-16',
  12.                '2011-12-19', '2011-12-21', '2011-12-23', '2011-12-26',
  13.                '2011-12-28', '2011-12-30'],
  14.               dtype='datetime64[ns]', length=154, freq='C')
  16. In [91]: pd.bdate_range(start, end, freq='CBMS', weekmask=weekmask)
  17. Out[91]: 
  18. DatetimeIndex(['2011-01-03', '2011-02-02', '2011-03-02', '2011-04-01',
  19.                '2011-05-02', '2011-06-01', '2011-07-01', '2011-08-01',
  20.                '2011-09-02', '2011-10-03', '2011-11-02', '2011-12-02'],
  21.               dtype='datetime64[ns]', freq='CBMS')

Custom business days

Timestamp limitations

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:

  1. In [92]: pd.Timestamp.min
  2. Out[92]: Timestamp('1677-09-21 00:12:43.145225')
  4. In [93]: pd.Timestamp.max
  5. Out[93]: Timestamp('2262-04-11 23:47:16.854775807')

Representing out-of-bounds spans

Indexing

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 and tshift method on pandas objects.
  • 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.

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

::: tip Note

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.

  1. In [94]: rng = pd.date_range(start, end, freq='BM')
  3. In [95]: ts = pd.Series(np.random.randn(len(rng)), index=rng)
  5. In [96]: ts.index
  6. Out[96]: 
  7. DatetimeIndex(['2011-01-31', '2011-02-28', '2011-03-31', '2011-04-29',
  8.                '2011-05-31', '2011-06-30', '2011-07-29', '2011-08-31',
  9.                '2011-09-30', '2011-10-31', '2011-11-30', '2011-12-30'],
  10.               dtype='datetime64[ns]', freq='BM')
  12. In [97]: ts[:5].index
  13. Out[97]: 
  14. DatetimeIndex(['2011-01-31', '2011-02-28', '2011-03-31', '2011-04-29',
  15.                '2011-05-31'],
  16.               dtype='datetime64[ns]', freq='BM')
  18. In [98]: ts[::2].index
  19. Out[98]: 
  20. DatetimeIndex(['2011-01-31', '2011-03-31', '2011-05-31', '2011-07-29',
  21.                '2011-09-30', '2011-11-30'],
  22.               dtype='datetime64[ns]', freq='2BM')

Partial string indexing

Dates and strings that parse to timestamps can be passed as indexing parameters:

  1. In [99]: ts['1/31/2011']
  2. Out[99]: 0.11920871129693428
  4. In [100]: ts[datetime.datetime(2011, 12, 25):]
  5. Out[100]: 
  6. 2011-12-30    0.56702
  7. Freq: BM, dtype: float64
  9. In [101]: ts['10/31/2011':'12/31/2011']
  10. Out[101]: 
  11. 2011-10-31    0.271860
  12. 2011-11-30   -0.424972
  13. 2011-12-30    0.567020
  14. 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:

  1. In [102]: ts['2011']
  2. Out[102]: 
  3. 2011-01-31    0.119209
  4. 2011-02-28   -1.044236
  5. 2011-03-31   -0.861849
  6. 2011-04-29   -2.104569
  7. 2011-05-31   -0.494929
  8. 2011-06-30    1.071804
  9. 2011-07-29    0.721555
  10. 2011-08-31   -0.706771
  11. 2011-09-30   -1.039575
  12. 2011-10-31    0.271860
  13. 2011-11-30   -0.424972
  14. 2011-12-30    0.567020
  15. Freq: BM, dtype: float64
  17. In [103]: ts['2011-6']
  18. Out[103]: 
  19. 2011-06-30    1.071804
  20. 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:

  1. In [104]: dft = pd.DataFrame(np.random.randn(100000, 1), columns=['A'],
  2.    .....:                    index=pd.date_range('20130101', periods=100000, freq='T'))
  3.    .....: 
  5. In [105]: dft
  6. Out[105]: 
  7.                             A
  8. 2013-01-01 00:00:00  0.276232
  9. 2013-01-01 00:01:00 -1.087401
  10. 2013-01-01 00:02:00 -0.673690
  11. 2013-01-01 00:03:00  0.113648
  12. 2013-01-01 00:04:00 -1.478427
  13. ...                       ...
  14. 2013-03-11 10:35:00 -0.747967
  15. 2013-03-11 10:36:00 -0.034523
  16. 2013-03-11 10:37:00 -0.201754
  17. 2013-03-11 10:38:00 -1.509067
  18. 2013-03-11 10:39:00 -1.693043
  20. [100000 rows x 1 columns]
  22. In [106]: dft['2013']
  23. Out[106]: 
  24.                             A
  25. 2013-01-01 00:00:00  0.276232
  26. 2013-01-01 00:01:00 -1.087401
  27. 2013-01-01 00:02:00 -0.673690
  28. 2013-01-01 00:03:00  0.113648
  29. 2013-01-01 00:04:00 -1.478427
  30. ...                       ...
  31. 2013-03-11 10:35:00 -0.747967
  32. 2013-03-11 10:36:00 -0.034523
  33. 2013-03-11 10:37:00 -0.201754
  34. 2013-03-11 10:38:00 -1.509067
  35. 2013-03-11 10:39:00 -1.693043
  37. [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:

  1. In [107]: dft['2013-1':'2013-2']
  2. Out[107]: 
  3.                             A
  4. 2013-01-01 00:00:00  0.276232
  5. 2013-01-01 00:01:00 -1.087401
  6. 2013-01-01 00:02:00 -0.673690
  7. 2013-01-01 00:03:00  0.113648
  8. 2013-01-01 00:04:00 -1.478427
  9. ...                       ...
  10. 2013-02-28 23:55:00  0.850929
  11. 2013-02-28 23:56:00  0.976712
  12. 2013-02-28 23:57:00 -2.693884
  13. 2013-02-28 23:58:00 -1.575535
  14. 2013-02-28 23:59:00 -1.573517
  16. [84960 rows x 1 columns]

This specifies a stop time that includes all of the times on the last day:

  1. In [108]: dft['2013-1':'2013-2-28']
  2. Out[108]: 
  3.                             A
  4. 2013-01-01 00:00:00  0.276232
  5. 2013-01-01 00:01:00 -1.087401
  6. 2013-01-01 00:02:00 -0.673690
  7. 2013-01-01 00:03:00  0.113648
  8. 2013-01-01 00:04:00 -1.478427
  9. ...                       ...
  10. 2013-02-28 23:55:00  0.850929
  11. 2013-02-28 23:56:00  0.976712
  12. 2013-02-28 23:57:00 -2.693884
  13. 2013-02-28 23:58:00 -1.575535
  14. 2013-02-28 23:59:00 -1.573517
  16. [84960 rows x 1 columns]

This specifies an exact stop time (and is not the same as the above):

  1. In [109]: dft['2013-1':'2013-2-28 00:00:00']
  2. Out[109]: 
  3.                             A
  4. 2013-01-01 00:00:00  0.276232
  5. 2013-01-01 00:01:00 -1.087401
  6. 2013-01-01 00:02:00 -0.673690
  7. 2013-01-01 00:03:00  0.113648
  8. 2013-01-01 00:04:00 -1.478427
  9. ...                       ...
  10. 2013-02-27 23:56:00  1.197749
  11. 2013-02-27 23:57:00  0.720521
  12. 2013-02-27 23:58:00 -0.072718
  13. 2013-02-27 23:59:00 -0.681192
  14. 2013-02-28 00:00:00 -0.557501
  16. [83521 rows x 1 columns]

We are stopping on the included end-point as it is part of the index:

  1. In [110]: dft['2013-1-15':'2013-1-15 12:30:00']
  2. Out[110]: 
  3.                             A
  4. 2013-01-15 00:00:00 -0.984810
  5. 2013-01-15 00:01:00  0.941451
  6. 2013-01-15 00:02:00  1.559365
  7. 2013-01-15 00:03:00  1.034374
  8. 2013-01-15 00:04:00 -1.480656
  9. ...                       ...
  10. 2013-01-15 12:26:00  0.371454
  11. 2013-01-15 12:27:00 -0.930806
  12. 2013-01-15 12:28:00 -0.069177
  13. 2013-01-15 12:29:00  0.066510
  14. 2013-01-15 12:30:00 -0.003945
  16. [751 rows x 1 columns]

New in version 0.18.0.

DatetimeIndex partial string indexing also works on a DataFrame with a MultiIndex:

  1. In [111]: dft2 = pd.DataFrame(np.random.randn(20, 1),
  2.    .....:                     columns=['A'],
  3.    .....:                     index=pd.MultiIndex.from_product(
  4.    .....:                         [pd.date_range('20130101', periods=10, freq='12H'),
  5.    .....:                          ['a', 'b']]))
  6.    .....: 
  8. In [112]: dft2
  9. Out[112]: 
  10.                               A
  11. 2013-01-01 00:00:00 a -0.298694
  12.                     b  0.823553
  13. 2013-01-01 12:00:00 a  0.943285
  14.                     b -1.479399
  15. 2013-01-02 00:00:00 a -1.643342
  16. ...                         ...
  17. 2013-01-04 12:00:00 b  0.069036
  18. 2013-01-05 00:00:00 a  0.122297
  19.                     b  1.422060
  20. 2013-01-05 12:00:00 a  0.370079
  21.                     b  1.016331
  23. [20 rows x 1 columns]
  25. In [113]: dft2.loc['2013-01-05']
  26. Out[113]: 
  27.                               A
  28. 2013-01-05 00:00:00 a  0.122297
  29.                     b  1.422060
  30. 2013-01-05 12:00:00 a  0.370079
  31.                     b  1.016331
  33. In [114]: idx = pd.IndexSlice
  35. In [115]: dft2 = dft2.swaplevel(0, 1).sort_index()
  37. In [116]: dft2.loc[idx[:, '2013-01-05'], :]
  38. Out[116]: 
  39.                               A
  40. a 2013-01-05 00:00:00  0.122297
  41.   2013-01-05 12:00:00  0.370079
  42. b 2013-01-05 00:00:00  1.422060
  43.   2013-01-05 12:00:00  1.016331

New in version 0.25.0.

Slicing with string indexing also honors UTC offset.

  1. In [117]: df = pd.DataFrame([0], index=pd.DatetimeIndex(['2019-01-01'], tz='US/Pacific'))
  3. In [118]: df
  4. Out[118]: 
  5.                            0
  6. 2019-01-01 00:00:00-08:00  0
  8. In [119]: df['2019-01-01 12:00:00+04:00':'2019-01-01 13:00:00+04:00']
  9. Out[119]: 
  10.                            0
  11. 2019-01-01 00:00:00-08:00  0

Slice vs. exact match

Changed in version 0.20.0.

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:

  1. In [120]: series_minute = pd.Series([1, 2, 3],
  2.    .....:                           pd.DatetimeIndex(['2011-12-31 23:59:00',
  3.    .....:                                             '2012-01-01 00:00:00',
  4.    .....:                                             '2012-01-01 00:02:00']))
  5.    .....: 
  7. In [121]: series_minute.index.resolution
  8. Out[121]: 'minute'

A timestamp string less accurate than a minute gives a Series object.

  1. In [122]: series_minute['2011-12-31 23']
  2. Out[122]: 
  3. 2011-12-31 23:59:00    1
  4. dtype: int64

A timestamp string with minute resolution (or more accurate), gives a scalar instead, i.e. it is not casted to a slice.

  1. In [123]: series_minute['2011-12-31 23:59']
  2. Out[123]: 1
  4. In [124]: series_minute['2011-12-31 23:59:00']
  5. Out[124]: 1

If index resolution is second, then the minute-accurate timestamp gives a Series.

  1. In [125]: series_second = pd.Series([1, 2, 3],
  2.    .....:                           pd.DatetimeIndex(['2011-12-31 23:59:59',
  3.    .....:                                             '2012-01-01 00:00:00',
  4.    .....:                                             '2012-01-01 00:00:01']))
  5.    .....: 
  7. In [126]: series_second.index.resolution
  8. Out[126]: 'second'
  10. In [127]: series_second['2011-12-31 23:59']
  11. Out[127]: 
  12. 2011-12-31 23:59:59    1
  13. dtype: int64

If the timestamp string is treated as a slice, it can be used to index DataFrame with [] as well.

  1. In [128]: dft_minute = pd.DataFrame({'a': [1, 2, 3], 'b': [4, 5, 6]},
  2.    .....:                           index=series_minute.index)
  3.    .....: 
  5. In [129]: dft_minute['2011-12-31 23']
  6. Out[129]: 
  7.                      a  b
  8. 2011-12-31 23:59:00  1  4

::: danger Warning

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:

To always have unambiguous selection, whether the row is treated as a slice or a single selection, use .loc.

  1. In [130]: dft_minute.loc['2011-12-31 23:59']
  2. Out[130]: 
  3. a    1
  4. b    4
  5. Name: 2011-12-31 23:59:00, dtype: int64

:::

Note also that DatetimeIndex resolution cannot be less precise than day.

  1. In [131]: series_monthly = pd.Series([1, 2, 3],
  2.    .....:                            pd.DatetimeIndex(['2011-12', '2012-01', '2012-02']))
  3.    .....: 
  5. In [132]: series_monthly.index.resolution
  6. Out[132]: 'day'
  8. In [133]: series_monthly['2011-12']  # returns Series
  9. Out[133]: 
  10. 2011-12-01    1
  11. dtype: int64

Exact indexing

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).

  1. In [134]: dft[datetime.datetime(2013, 1, 1):datetime.datetime(2013, 2, 28)]
  2. Out[134]: 
  3.                             A
  4. 2013-01-01 00:00:00  0.276232
  5. 2013-01-01 00:01:00 -1.087401
  6. 2013-01-01 00:02:00 -0.673690
  7. 2013-01-01 00:03:00  0.113648
  8. 2013-01-01 00:04:00 -1.478427
  9. ...                       ...
  10. 2013-02-27 23:56:00  1.197749
  11. 2013-02-27 23:57:00  0.720521
  12. 2013-02-27 23:58:00 -0.072718
  13. 2013-02-27 23:59:00 -0.681192
  14. 2013-02-28 00:00:00 -0.557501
  16. [83521 rows x 1 columns]

With no defaults.

  1. In [135]: dft[datetime.datetime(2013, 1, 1, 10, 12, 0):
  2.    .....:     datetime.datetime(2013, 2, 28, 10, 12, 0)]
  3.    .....: 
  4. Out[135]: 
  5.                             A
  6. 2013-01-01 10:12:00  0.565375
  7. 2013-01-01 10:13:00  0.068184
  8. 2013-01-01 10:14:00  0.788871
  9. 2013-01-01 10:15:00 -0.280343
  10. 2013-01-01 10:16:00  0.931536
  11. ...                       ...
  12. 2013-02-28 10:08:00  0.148098
  13. 2013-02-28 10:09:00 -0.388138
  14. 2013-02-28 10:10:00  0.139348
  15. 2013-02-28 10:11:00  0.085288
  16. 2013-02-28 10:12:00  0.950146
  18. [83521 rows x 1 columns]

Truncating & fancy indexing

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:

  1. In [136]: rng2 = pd.date_range('2011-01-01', '2012-01-01', freq='W')
  3. In [137]: ts2 = pd.Series(np.random.randn(len(rng2)), index=rng2)
  5. In [138]: ts2.truncate(before='2011-11', after='2011-12')
  6. Out[138]: 
  7. 2011-11-06    0.437823
  8. 2011-11-13   -0.293083
  9. 2011-11-20   -0.059881
  10. 2011-11-27    1.252450
  11. Freq: W-SUN, dtype: float64
  13. In [139]: ts2['2011-11':'2011-12']
  14. Out[139]: 
  15. 2011-11-06    0.437823
  16. 2011-11-13   -0.293083
  17. 2011-11-20   -0.059881
  18. 2011-11-27    1.252450
  19. 2011-12-04    0.046611
  20. 2011-12-11    0.059478
  21. 2011-12-18   -0.286539
  22. 2011-12-25    0.841669
  23. Freq: W-SUN, dtype: float64

Even complicated fancy indexing that breaks the DatetimeIndex frequency regularity will result in a DatetimeIndex, although frequency is lost:

  1. In [140]: ts2[[0, 2, 6]].index
  2. Out[140]: DatetimeIndex(['2011-01-02', '2011-01-16', '2011-02-13'], dtype='datetime64[ns]', freq=None)

Time/date components

There are several time/date properties that one can access from Timestamp or a collection of timestamps like a DatetimeIndex.

Property Description
year The year of the datetime
month The month of the datetime
day The days of the datetime
hour The hour of the datetime
minute The minutes of the datetime
second The seconds of the datetime
microsecond The microseconds of the datetime
nanosecond 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
weekofyear The week ordinal of the year
week The week ordinal of the year
dayofweek The number of the day of the week with Monday=0, Sunday=6
weekday The number of the day of the week with Monday=0, Sunday=6
weekday_name The name of the day in a week (ex: Friday)
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.

DateOffset objects

In the preceding examples, frequency strings (e.g. 'D') were used to specify a frequency that defined:

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.

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.

  1. # This particular day contains a day light savings time transition
  2. In [141]: ts = pd.Timestamp('2016-10-30 00:00:00', tz='Europe/Helsinki')
  4. # Respects absolute time
  5. In [142]: ts + pd.Timedelta(days=1)
  6. Out[142]: Timestamp('2016-10-30 23:00:00+0200', tz='Europe/Helsinki')
  8. # Respects calendar time
  9. In [143]: ts + pd.DateOffset(days=1)
  10. Out[143]: Timestamp('2016-10-31 00:00:00+0200', tz='Europe/Helsinki')
  12. In [144]: friday = pd.Timestamp('2018-01-05')
  14. In [145]: friday.day_name()
  15. Out[145]: 'Friday'
  17. # Add 2 business days (Friday --> Tuesday)
  18. In [146]: two_business_days = 2 * pd.offsets.BDay()
  20. In [147]: two_business_days.apply(friday)
  21. Out[147]: Timestamp('2018-01-09 00:00:00')
  23. In [148]: friday + two_business_days
  24. Out[148]: Timestamp('2018-01-09 00:00:00')
  26. In [149]: (friday + two_business_days).day_name()
  27. Out[149]: '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:

Date Offset Frequency String Description
DateOffset None Generic offset class, defaults to 1 calendar day
BDay or BusinessDay ‘B’ business day (weekday)
CDay or 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 ‘BM’ business month end
BMonthBegin or BusinessMonthBegin ‘BMS’ business month begin
CBMonthEnd or CustomBusinessMonthEnd ‘CBM’ custom business month end
CBMonthBegin or 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’ calendar year begin
BYearEnd ‘BA’ business year end
BYearBegin ‘BAS’ business year begin
FY5253 ‘RE’ retail (aka 52-53 week) year
Easter None Easter holiday
BusinessHour ‘BH’ business hour
CustomBusinessHour ‘CBH’ custom business hour
Day ‘D’ one absolute day
Hour ‘H’ one hour
Minute ‘T’ or ‘min’ one minute
Second ‘S’ one second
Milli ‘L’ or ‘ms’ one millisecond
Micro ‘U’ or ‘us’ one microsecond
Nano ‘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.

  1. In [150]: ts = pd.Timestamp('2018-01-06 00:00:00')
  3. In [151]: ts.day_name()
  4. Out[151]: 'Saturday'
  6. # BusinessHour's valid offset dates are Monday through Friday
  7. In [152]: offset = pd.offsets.BusinessHour(start='09:00')
  9. # Bring the date to the closest offset date (Monday)
  10. In [153]: offset.rollforward(ts)
  11. Out[153]: Timestamp('2018-01-08 09:00:00')
  13. # Date is brought to the closest offset date first and then the hour is added
  14. In [154]: ts + offset
  15. Out[154]: 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).

  1. In [155]: ts = pd.Timestamp('2014-01-01 09:00')
  3. In [156]: day = pd.offsets.Day()
  5. In [157]: day.apply(ts)
  6. Out[157]: Timestamp('2014-01-02 09:00:00')
  8. In [158]: day.apply(ts).normalize()
  9. Out[158]: Timestamp('2014-01-02 00:00:00')
  11. In [159]: ts = pd.Timestamp('2014-01-01 22:00')
  13. In [160]: hour = pd.offsets.Hour()
  15. In [161]: hour.apply(ts)
  16. Out[161]: Timestamp('2014-01-01 23:00:00')
  18. In [162]: hour.apply(ts).normalize()
  19. Out[162]: Timestamp('2014-01-01 00:00:00')
  21. In [163]: hour.apply(pd.Timestamp("2014-01-01 23:30")).normalize()
  22. Out[163]: Timestamp('2014-01-02 00:00:00')

Parametric offsets

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:

  1. In [164]: d = datetime.datetime(2008, 8, 18, 9, 0)
  3. In [165]: d
  4. Out[165]: datetime.datetime(2008, 8, 18, 9, 0)
  6. In [166]: d + pd.offsets.Week()
  7. Out[166]: Timestamp('2008-08-25 09:00:00')
  9. In [167]: d + pd.offsets.Week(weekday=4)
  10. Out[167]: Timestamp('2008-08-22 09:00:00')
  12. In [168]: (d + pd.offsets.Week(weekday=4)).weekday()
  13. Out[168]: 4
  15. In [169]: d - pd.offsets.Week()
  16. Out[169]: Timestamp('2008-08-11 09:00:00')

The normalize option will be effective for addition and subtraction.

  1. In [170]: d + pd.offsets.Week(normalize=True)
  2. Out[170]: Timestamp('2008-08-25 00:00:00')
  4. In [171]: d - pd.offsets.Week(normalize=True)
  5. Out[171]: Timestamp('2008-08-11 00:00:00')

Another example is parameterizing YearEnd with the specific ending month:

  1. In [172]: d + pd.offsets.YearEnd()
  2. Out[172]: Timestamp('2008-12-31 09:00:00')
  4. In [173]: d + pd.offsets.YearEnd(month=6)
  5. Out[173]: Timestamp('2009-06-30 09:00:00')

Using offsets with Series / DatetimeIndex

Offsets can be used with either a Series or DatetimeIndex to apply the offset to each element.

  1. In [174]: rng = pd.date_range('2012-01-01', '2012-01-03')
  3. In [175]: s = pd.Series(rng)
  5. In [176]: rng
  6. Out[176]: DatetimeIndex(['2012-01-01', '2012-01-02', '2012-01-03'], dtype='datetime64[ns]', freq='D')
  8. In [177]: rng + pd.DateOffset(months=2)
  9. Out[177]: DatetimeIndex(['2012-03-01', '2012-03-02', '2012-03-03'], dtype='datetime64[ns]', freq='D')
  11. In [178]: s + pd.DateOffset(months=2)
  12. Out[178]: 
  13. 0   2012-03-01
  14. 1   2012-03-02
  15. 2   2012-03-03
  16. dtype: datetime64[ns]
  18. In [179]: s - pd.DateOffset(months=2)
  19. Out[179]: 
  20. 0   2011-11-01
  21. 1   2011-11-02
  22. 2   2011-11-03
  23. 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.

  1. In [180]: s - pd.offsets.Day(2)
  2. Out[180]: 
  3. 0   2011-12-30
  4. 1   2011-12-31
  5. 2   2012-01-01
  6. dtype: datetime64[ns]
  8. In [181]: td = s - pd.Series(pd.date_range('2011-12-29', '2011-12-31'))
  10. In [182]: td
  11. Out[182]: 
  12. 0   3 days
  13. 1   3 days
  14. 2   3 days
  15. dtype: timedelta64[ns]
  17. In [183]: td + pd.offsets.Minute(15)
  18. Out[183]: 
  19. 0   3 days 00:15:00
  20. 1   3 days 00:15:00
  21. 2   3 days 00:15:00
  22. 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

  1. In [184]: rng + pd.offsets.BQuarterEnd()
  2. Out[184]: DatetimeIndex(['2012-03-30', '2012-03-30', '2012-03-30'], dtype='datetime64[ns]', freq='D')

Custom business days

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.

  1. In [185]: weekmask_egypt = 'Sun Mon Tue Wed Thu'
  3. # They also observe International Workers' Day so let's
  4. # add that for a couple of years
  5. In [186]: holidays = ['2012-05-01',
  6.    .....:             datetime.datetime(2013, 5, 1),
  7.    .....:             np.datetime64('2014-05-01')]
  8.    .....: 
  10. In [187]: bday_egypt = pd.offsets.CustomBusinessDay(holidays=holidays,
  11.    .....:                                           weekmask=weekmask_egypt)
  12.    .....: 
  14. In [188]: dt = datetime.datetime(2013, 4, 30)
  16. In [189]: dt + 2 * bday_egypt
  17. Out[189]: Timestamp('2013-05-05 00:00:00')

Let’s map to the weekday names:

  1. In [190]: dts = pd.date_range(dt, periods=5, freq=bday_egypt)
  3. In [191]: pd.Series(dts.weekday, dts).map(
  4.    .....:     pd.Series('Mon Tue Wed Thu Fri Sat Sun'.split()))
  5.    .....: 
  6. Out[191]: 
  7. 2013-04-30    Tue
  8. 2013-05-02    Thu
  9. 2013-05-05    Sun
  10. 2013-05-06    Mon
  11. 2013-05-07    Tue
  12. Freq: C, dtype: object

Holiday calendars can be used to provide the list of holidays. See the holiday calendar section for more information.

  1. In [192]: from pandas.tseries.holiday import USFederalHolidayCalendar
  3. In [193]: bday_us = pd.offsets.CustomBusinessDay(calendar=USFederalHolidayCalendar())
  5. # Friday before MLK Day
  6. In [194]: dt = datetime.datetime(2014, 1, 17)
  8. # Tuesday after MLK Day (Monday is skipped because it's a holiday)
  9. In [195]: dt + bday_us
  10. Out[195]: Timestamp('2014-01-21 00:00:00')

Monthly offsets that respect a certain holiday calendar can be defined in the usual way.

  1. In [196]: bmth_us = pd.offsets.CustomBusinessMonthBegin(
  2.    .....:     calendar=USFederalHolidayCalendar())
  3.    .....: 
  5. # Skip new years
  6. In [197]: dt = datetime.datetime(2013, 12, 17)
  8. In [198]: dt + bmth_us
  9. Out[198]: Timestamp('2014-01-02 00:00:00')
  11. # Define date index with custom offset
  12. In [199]: pd.date_range(start='20100101', end='20120101', freq=bmth_us)
  13. Out[199]: 
  14. DatetimeIndex(['2010-01-04', '2010-02-01', '2010-03-01', '2010-04-01',
  15.                '2010-05-03', '2010-06-01', '2010-07-01', '2010-08-02',
  16.                '2010-09-01', '2010-10-01', '2010-11-01', '2010-12-01',
  17.                '2011-01-03', '2011-02-01', '2011-03-01', '2011-04-01',
  18.                '2011-05-02', '2011-06-01', '2011-07-01', '2011-08-01',
  19.                '2011-09-01', '2011-10-03', '2011-11-01', '2011-12-01'],
  20.               dtype='datetime64[ns]', freq='CBMS')

::: tip Note

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.

:::

Business hour

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.

  1. In [200]: bh = pd.offsets.BusinessHour()
  3. In [201]: bh
  4. Out[201]: <BusinessHour: BH=09:00-17:00>
  6. # 2014-08-01 is Friday
  7. In [202]: pd.Timestamp('2014-08-01 10:00').weekday()
  8. Out[202]: 4
  10. In [203]: pd.Timestamp('2014-08-01 10:00') + bh
  11. Out[203]: Timestamp('2014-08-01 11:00:00')
  13. # Below example is the same as: pd.Timestamp('2014-08-01 09:00') + bh
  14. In [204]: pd.Timestamp('2014-08-01 08:00') + bh
  15. Out[204]: Timestamp('2014-08-01 10:00:00')
  17. # If the results is on the end time, move to the next business day
  18. In [205]: pd.Timestamp('2014-08-01 16:00') + bh
  19. Out[205]: Timestamp('2014-08-04 09:00:00')
  21. # Remainings are added to the next day
  22. In [206]: pd.Timestamp('2014-08-01 16:30') + bh
  23. Out[206]: Timestamp('2014-08-04 09:30:00')
  25. # Adding 2 business hours
  26. In [207]: pd.Timestamp('2014-08-01 10:00') + pd.offsets.BusinessHour(2)
  27. Out[207]: Timestamp('2014-08-01 12:00:00')
  29. # Subtracting 3 business hours
  30. In [208]: pd.Timestamp('2014-08-01 10:00') + pd.offsets.BusinessHour(-3)
  31. Out[208]: 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.

  1. In [209]: bh = pd.offsets.BusinessHour(start='11:00', end=datetime.time(20, 0))
  3. In [210]: bh
  4. Out[210]: <BusinessHour: BH=11:00-20:00>
  6. In [211]: pd.Timestamp('2014-08-01 13:00') + bh
  7. Out[211]: Timestamp('2014-08-01 14:00:00')
  9. In [212]: pd.Timestamp('2014-08-01 09:00') + bh
  10. Out[212]: Timestamp('2014-08-01 12:00:00')
  12. In [213]: pd.Timestamp('2014-08-01 18:00') + bh
  13. Out[213]: 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.

  1. In [214]: bh = pd.offsets.BusinessHour(start='17:00', end='09:00')
  3. In [215]: bh
  4. Out[215]: <BusinessHour: BH=17:00-09:00>
  6. In [216]: pd.Timestamp('2014-08-01 17:00') + bh
  7. Out[216]: Timestamp('2014-08-01 18:00:00')
  9. In [217]: pd.Timestamp('2014-08-01 23:00') + bh
  10. Out[217]: Timestamp('2014-08-02 00:00:00')
  12. # Although 2014-08-02 is Saturday,
  13. # it is valid because it starts from 08-01 (Friday).
  14. In [218]: pd.Timestamp('2014-08-02 04:00') + bh
  15. Out[218]: Timestamp('2014-08-02 05:00:00')
  17. # Although 2014-08-04 is Monday,
  18. # it is out of business hours because it starts from 08-03 (Sunday).
  19. In [219]: pd.Timestamp('2014-08-04 04:00') + bh
  20. Out[219]: 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.

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.

  1. # This adjusts a Timestamp to business hour edge
  2. In [220]: pd.offsets.BusinessHour().rollback(pd.Timestamp('2014-08-02 15:00'))
  3. Out[220]: Timestamp('2014-08-01 17:00:00')
  5. In [221]: pd.offsets.BusinessHour().rollforward(pd.Timestamp('2014-08-02 15:00'))
  6. Out[221]: Timestamp('2014-08-04 09:00:00')
  8. # It is the same as BusinessHour().apply(pd.Timestamp('2014-08-01 17:00')).
  9. # And it is the same as BusinessHour().apply(pd.Timestamp('2014-08-04 09:00'))
  10. In [222]: pd.offsets.BusinessHour().apply(pd.Timestamp('2014-08-02 15:00'))
  11. Out[222]: Timestamp('2014-08-04 10:00:00')
  13. # BusinessDay results (for reference)
  14. In [223]: pd.offsets.BusinessHour().rollforward(pd.Timestamp('2014-08-02'))
  15. Out[223]: Timestamp('2014-08-04 09:00:00')
  17. # It is the same as BusinessDay().apply(pd.Timestamp('2014-08-01'))
  18. # The result is the same as rollworward because BusinessDay never overlap.
  19. In [224]: pd.offsets.BusinessHour().apply(pd.Timestamp('2014-08-02'))
  20. Out[224]: 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.

Custom business hour

New in version 0.18.1.

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.

  1. In [225]: from pandas.tseries.holiday import USFederalHolidayCalendar
  3. In [226]: bhour_us = pd.offsets.CustomBusinessHour(calendar=USFederalHolidayCalendar())
  5. # Friday before MLK Day
  6. In [227]: dt = datetime.datetime(2014, 1, 17, 15)
  8. In [228]: dt + bhour_us
  9. Out[228]: Timestamp('2014-01-17 16:00:00')
  11. # Tuesday after MLK Day (Monday is skipped because it's a holiday)
  12. In [229]: dt + bhour_us * 2
  13. Out[229]: Timestamp('2014-01-21 09:00:00')

You can use keyword arguments supported by either BusinessHour and CustomBusinessDay.

  1. In [230]: bhour_mon = pd.offsets.CustomBusinessHour(start='10:00',
  2.    .....:                                           weekmask='Tue Wed Thu Fri')
  3.    .....: 
  5. # Monday is skipped because it's a holiday, business hour starts from 10:00
  6. In [231]: dt + bhour_mon * 2
  7. Out[231]: Timestamp('2014-01-21 10:00:00')

Offset aliases

A number of string aliases are given to useful common time series frequencies. We will refer to these aliases as offset aliases.

Alias Description
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

Combining aliases

As we have seen previously, the alias and the offset instance are fungible in most functions:

  1. In [232]: pd.date_range(start, periods=5, freq='B')
  2. Out[232]: 
  3. DatetimeIndex(['2011-01-03', '2011-01-04', '2011-01-05', '2011-01-06',
  4.                '2011-01-07'],
  5.               dtype='datetime64[ns]', freq='B')
  7. In [233]: pd.date_range(start, periods=5, freq=pd.offsets.BDay())
  8. Out[233]: 
  9. DatetimeIndex(['2011-01-03', '2011-01-04', '2011-01-05', '2011-01-06',
  10.                '2011-01-07'],
  11.               dtype='datetime64[ns]', freq='B')

You can combine together day and intraday offsets:

  1. In [234]: pd.date_range(start, periods=10, freq='2h20min')
  2. Out[234]: 
  3. DatetimeIndex(['2011-01-01 00:00:00', '2011-01-01 02:20:00',
  4.                '2011-01-01 04:40:00', '2011-01-01 07:00:00',
  5.                '2011-01-01 09:20:00', '2011-01-01 11:40:00',
  6.                '2011-01-01 14:00:00', '2011-01-01 16:20:00',
  7.                '2011-01-01 18:40:00', '2011-01-01 21:00:00'],
  8.               dtype='datetime64[ns]', freq='140T')
  10. In [235]: pd.date_range(start, periods=10, freq='1D10U')
  11. Out[235]: 
  12. DatetimeIndex([       '2011-01-01 00:00:00', '2011-01-02 00:00:00.000010',
  13.                '2011-01-03 00:00:00.000020', '2011-01-04 00:00:00.000030',
  14.                '2011-01-05 00:00:00.000040', '2011-01-06 00:00:00.000050',
  15.                '2011-01-07 00:00:00.000060', '2011-01-08 00:00:00.000070',
  16.                '2011-01-09 00:00:00.000080', '2011-01-10 00:00:00.000090'],
  17.               dtype='datetime64[ns]', freq='86400000010U')

Anchored offsets

For some frequencies you can specify an anchoring suffix:

Alias Description
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.

Anchored offset semantics

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.

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.

  1. In [236]: pd.Timestamp('2014-01-02') + pd.offsets.MonthBegin(n=1)
  2. Out[236]: Timestamp('2014-02-01 00:00:00')
  4. In [237]: pd.Timestamp('2014-01-02') + pd.offsets.MonthEnd(n=1)
  5. Out[237]: Timestamp('2014-01-31 00:00:00')
  7. In [238]: pd.Timestamp('2014-01-02') - pd.offsets.MonthBegin(n=1)
  8. Out[238]: Timestamp('2014-01-01 00:00:00')
  10. In [239]: pd.Timestamp('2014-01-02') - pd.offsets.MonthEnd(n=1)
  11. Out[239]: Timestamp('2013-12-31 00:00:00')
  13. In [240]: pd.Timestamp('2014-01-02') + pd.offsets.MonthBegin(n=4)
  14. Out[240]: Timestamp('2014-05-01 00:00:00')
  16. In [241]: pd.Timestamp('2014-01-02') - pd.offsets.MonthBegin(n=4)
  17. Out[241]: Timestamp('2013-10-01 00:00:00')

If the given date is on an anchor point, it is moved |n| points forwards or backwards.

  1. In [242]: pd.Timestamp('2014-01-01') + pd.offsets.MonthBegin(n=1)
  2. Out[242]: Timestamp('2014-02-01 00:00:00')
  4. In [243]: pd.Timestamp('2014-01-31') + pd.offsets.MonthEnd(n=1)
  5. Out[243]: Timestamp('2014-02-28 00:00:00')
  7. In [244]: pd.Timestamp('2014-01-01') - pd.offsets.MonthBegin(n=1)
  8. Out[244]: Timestamp('2013-12-01 00:00:00')
  10. In [245]: pd.Timestamp('2014-01-31') - pd.offsets.MonthEnd(n=1)
  11. Out[245]: Timestamp('2013-12-31 00:00:00')
  13. In [246]: pd.Timestamp('2014-01-01') + pd.offsets.MonthBegin(n=4)
  14. Out[246]: Timestamp('2014-05-01 00:00:00')
  16. In [247]: pd.Timestamp('2014-01-31') - pd.offsets.MonthBegin(n=4)
  17. Out[247]: 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.

  1. In [248]: pd.Timestamp('2014-01-02') + pd.offsets.MonthBegin(n=0)
  2. Out[248]: Timestamp('2014-02-01 00:00:00')
  4. In [249]: pd.Timestamp('2014-01-02') + pd.offsets.MonthEnd(n=0)
  5. Out[249]: Timestamp('2014-01-31 00:00:00')
  7. In [250]: pd.Timestamp('2014-01-01') + pd.offsets.MonthBegin(n=0)
  8. Out[250]: Timestamp('2014-01-01 00:00:00')
  10. In [251]: pd.Timestamp('2014-01-31') + pd.offsets.MonthEnd(n=0)
  11. Out[251]: Timestamp('2014-01-31 00:00:00')

Holidays / holiday calendars

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.

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 Description
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:

  1. In [252]: from pandas.tseries.holiday import Holiday, USMemorialDay,\
  2.    .....:     AbstractHolidayCalendar, nearest_workday, MO
  3.    .....: 
  5. In [253]: class ExampleCalendar(AbstractHolidayCalendar):
  6.    .....:     rules = [
  7.    .....:         USMemorialDay,
  8.    .....:         Holiday('July 4th', month=7, day=4, observance=nearest_workday),
  9.    .....:         Holiday('Columbus Day', month=10, day=1,
  10.    .....:                 offset=pd.DateOffset(weekday=MO(2)))]
  11.    .....: 
  13. In [254]: cal = ExampleCalendar()
  15. In [255]: cal.holidays(datetime.datetime(2012, 1, 1), datetime.datetime(2012, 12, 31))
  16. Out[255]: DatetimeIndex(['2012-05-28', '2012-07-04', '2012-10-08'], dtype='datetime64[ns]', freq=None)
hint: 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.

  1. In [256]: pd.date_range(start='7/1/2012', end='7/10/2012',
  2.    .....:               freq=pd.offsets.CDay(calendar=cal)).to_pydatetime()
  3.    .....: 
  4. Out[256]: 
  5. array([datetime.datetime(2012, 7, 2, 0, 0),
  6.        datetime.datetime(2012, 7, 3, 0, 0),
  7.        datetime.datetime(2012, 7, 5, 0, 0),
  8.        datetime.datetime(2012, 7, 6, 0, 0),
  9.        datetime.datetime(2012, 7, 9, 0, 0),
  10.        datetime.datetime(2012, 7, 10, 0, 0)], dtype=object)
  12. In [257]: offset = pd.offsets.CustomBusinessDay(calendar=cal)
  14. In [258]: datetime.datetime(2012, 5, 25) + offset
  15. Out[258]: Timestamp('2012-05-29 00:00:00')
  17. In [259]: datetime.datetime(2012, 7, 3) + offset
  18. Out[259]: Timestamp('2012-07-05 00:00:00')
  20. In [260]: datetime.datetime(2012, 7, 3) + 2 * offset
  21. Out[260]: Timestamp('2012-07-06 00:00:00')
  23. In [261]: datetime.datetime(2012, 7, 6) + offset
  24. Out[261]: 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.

  1. In [262]: AbstractHolidayCalendar.start_date
  2. Out[262]: Timestamp('1970-01-01 00:00:00')
  4. In [263]: AbstractHolidayCalendar.end_date
  5. Out[263]: Timestamp('2030-12-31 00:00:00')

These dates can be overwritten by setting the attributes as datetime/Timestamp/string.

  1. In [264]: AbstractHolidayCalendar.start_date = datetime.datetime(2012, 1, 1)
  3. In [265]: AbstractHolidayCalendar.end_date = datetime.datetime(2012, 12, 31)
  5. In [266]: cal.holidays()
  6. Out[266]: 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.

  1. In [267]: from pandas.tseries.holiday import get_calendar, HolidayCalendarFactory,\
  2.    .....:     USLaborDay
  3.    .....: 
  5. In [268]: cal = get_calendar('ExampleCalendar')
  7. In [269]: cal.rules
  8. Out[269]: 
  9. [Holiday: Memorial Day (month=5, day=31, offset=<DateOffset: weekday=MO(-1)>),
  10.  Holiday: July 4th (month=7, day=4, observance=<function nearest_workday at 0x7f65d1933ea0>),
  11.  Holiday: Columbus Day (month=10, day=1, offset=<DateOffset: weekday=MO(+2)>)]
  13. In [270]: new_cal = HolidayCalendarFactory('NewExampleCalendar', cal, USLaborDay)
  15. In [271]: new_cal.rules
  16. Out[271]: 
  17. [Holiday: Labor Day (month=9, day=1, offset=<DateOffset: weekday=MO(+1)>),
  18.  Holiday: Memorial Day (month=5, day=31, offset=<DateOffset: weekday=MO(-1)>),
  19.  Holiday: July 4th (month=7, day=4, observance=<function nearest_workday at 0x7f65d1933ea0>),
  20.  Holiday: Columbus Day (month=10, day=1, offset=<DateOffset: weekday=MO(+2)>)]

Time Series-Related Instance Methods

Shifting / lagging

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.

  1. In [272]: ts = pd.Series(range(len(rng)), index=rng)
  3. In [273]: ts = ts[:5]
  5. In [274]: ts.shift(1)
  6. Out[274]: 
  7. 2012-01-01    NaN
  8. 2012-01-02    0.0
  9. 2012-01-03    1.0
  10. 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:

  1. In [275]: ts.shift(5, freq=pd.offsets.BDay())
  2. Out[275]: 
  3. 2012-01-06    0
  4. 2012-01-09    1
  5. 2012-01-10    2
  6. Freq: B, dtype: int64
  8. In [276]: ts.shift(5, freq='BM')
  9. Out[276]: 
  10. 2012-05-31    0
  11. 2012-05-31    1
  12. 2012-05-31    2
  13. Freq: D, dtype: int64

Rather than changing the alignment of the data and the index, DataFrame and Series objects also have a tshift() convenience method that changes all the dates in the index by a specified number of offsets:

  1. In [277]: ts.tshift(5, freq='D')
  2. Out[277]: 
  3. 2012-01-06    0
  4. 2012-01-07    1
  5. 2012-01-08    2
  6. Freq: D, dtype: int64

Note that with tshift, the leading entry is no longer NaN because the data is not being realigned.

Frequency conversion

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.

  1. In [278]: dr = pd.date_range('1/1/2010', periods=3, freq=3 * pd.offsets.BDay())
  3. In [279]: ts = pd.Series(np.random.randn(3), index=dr)
  5. In [280]: ts
  6. Out[280]: 
  7. 2010-01-01    1.494522
  8. 2010-01-06   -0.778425
  9. 2010-01-11   -0.253355
  10. Freq: 3B, dtype: float64
  12. In [281]: ts.asfreq(pd.offsets.BDay())
  13. Out[281]: 
  14. 2010-01-01    1.494522
  15. 2010-01-04         NaN
  16. 2010-01-05         NaN
  17. 2010-01-06   -0.778425
  18. 2010-01-07         NaN
  19. 2010-01-08         NaN
  20. 2010-01-11   -0.253355
  21. 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.

  1. In [282]: ts.asfreq(pd.offsets.BDay(), method='pad')
  2. Out[282]: 
  3. 2010-01-01    1.494522
  4. 2010-01-04    1.494522
  5. 2010-01-05    1.494522
  6. 2010-01-06   -0.778425
  7. 2010-01-07   -0.778425
  8. 2010-01-08   -0.778425
  9. 2010-01-11   -0.253355
  10. Freq: B, dtype: float64

Filling forward / backward

Related to asfreq and reindex is fillna(), which is documented in the missing data section.

Converting to Python datetimes

DatetimeIndex can be converted to an array of Python native datetime.datetime objects using the to_pydatetime method.

Resampling

::: danger Warning

The interface to .resample has changed in 0.18.0 to be more groupby-like and hence more flexible. See the whatsnew docs for a comparison with prior versions.

:::

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.

Starting in version 0.18.1, the resample() function can be used directly from DataFrameGroupBy objects, see the groupby docs.

::: tip Note

.resample() is similar to using a rolling() operation with a time-based offset, see a discussion here.

:::

Basics

  1. In [283]: rng = pd.date_range('1/1/2012', periods=100, freq='S')
  3. In [284]: ts = pd.Series(np.random.randint(0, 500, len(rng)), index=rng)
  5. In [285]: ts.resample('5Min').sum()
  6. Out[285]: 
  7. 2012-01-01    25103
  8. 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.

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:

  1. In [286]: ts.resample('5Min').mean()
  2. Out[286]: 
  3. 2012-01-01    251.03
  4. Freq: 5T, dtype: float64
  6. In [287]: ts.resample('5Min').ohlc()
  7. Out[287]: 
  8.             open  high  low  close
  9. 2012-01-01   308   460    9    205
  11. In [288]: ts.resample('5Min').max()
  12. Out[288]: 
  13. 2012-01-01    460
  14. Freq: 5T, dtype: int64

For downsampling, closed can be set to ‘left’ or ‘right’ to specify which end of the interval is closed:

  1. In [289]: ts.resample('5Min', closed='right').mean()
  2. Out[289]: 
  3. 2011-12-31 23:55:00    308.000000
  4. 2012-01-01 00:00:00    250.454545
  5. Freq: 5T, dtype: float64
  7. In [290]: ts.resample('5Min', closed='left').mean()
  8. Out[290]: 
  9. 2012-01-01    251.03
  10. Freq: 5T, dtype: float64

Parameters like label and loffset are used to manipulate the resulting labels. label specifies whether the result is labeled with the beginning or the end of the interval. loffset performs a time adjustment on the output labels.

  1. In [291]: ts.resample('5Min').mean()  # by default label='left'
  2. Out[291]: 
  3. 2012-01-01    251.03
  4. Freq: 5T, dtype: float64
  6. In [292]: ts.resample('5Min', label='left').mean()
  7. Out[292]: 
  8. 2012-01-01    251.03
  9. Freq: 5T, dtype: float64
  11. In [293]: ts.resample('5Min', label='left', loffset='1s').mean()
  12. Out[293]: 
  13. 2012-01-01 00:00:01    251.03
  14. dtype: float64

::: danger Warning

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:

  1. In [294]: s = pd.date_range('2000-01-01', '2000-01-05').to_series()
  3. In [295]: s.iloc[2] = pd.NaT
  5. In [296]: s.dt.weekday_name
  6. Out[296]: 
  7. 2000-01-01     Saturday
  8. 2000-01-02       Sunday
  9. 2000-01-03          NaN
  10. 2000-01-04      Tuesday
  11. 2000-01-05    Wednesday
  12. Freq: D, dtype: object
  14. # default: label='left', closed='left'
  15. In [297]: s.resample('B').last().dt.weekday_name
  16. Out[297]: 
  17. 1999-12-31       Sunday
  18. 2000-01-03          NaN
  19. 2000-01-04      Tuesday
  20. 2000-01-05    Wednesday
  21. 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

  1. In [298]: s.resample('B', label='right', closed='right').last().dt.weekday_name
  2. Out[298]: 
  3. 2000-01-03       Sunday
  4. 2000-01-04      Tuesday
  5. 2000-01-05    Wednesday
  6. 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.

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.

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.

Upsampling

For upsampling, you can specify a way to upsample and the limit parameter to interpolate over the gaps that are created:

  1. # from secondly to every 250 milliseconds
  2. In [299]: ts[:2].resample('250L').asfreq()
  3. Out[299]: 
  4. 2012-01-01 00:00:00.000    308.0
  5. 2012-01-01 00:00:00.250      NaN
  6. 2012-01-01 00:00:00.500      NaN
  7. 2012-01-01 00:00:00.750      NaN
  8. 2012-01-01 00:00:01.000    204.0
  9. Freq: 250L, dtype: float64
  11. In [300]: ts[:2].resample('250L').ffill()
  12. Out[300]: 
  13. 2012-01-01 00:00:00.000    308
  14. 2012-01-01 00:00:00.250    308
  15. 2012-01-01 00:00:00.500    308
  16. 2012-01-01 00:00:00.750    308
  17. 2012-01-01 00:00:01.000    204
  18. Freq: 250L, dtype: int64
  20. In [301]: ts[:2].resample('250L').ffill(limit=2)
  21. Out[301]: 
  22. 2012-01-01 00:00:00.000    308.0
  23. 2012-01-01 00:00:00.250    308.0
  24. 2012-01-01 00:00:00.500    308.0
  25. 2012-01-01 00:00:00.750      NaN
  26. 2012-01-01 00:00:01.000    204.0
  27. Freq: 250L, dtype: float64

Sparse resampling

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.

Since resample is a time-based groupby, the following is a method to efficiently resample only the groups that are not all NaN.

  1. In [302]: rng = pd.date_range('2014-1-1', periods=100, freq='D') + pd.Timedelta('1s')
  3. In [303]: ts = pd.Series(range(100), index=rng)

If we want to resample to the full range of the series:

  1. In [304]: ts.resample('3T').sum()
  2. Out[304]: 
  3. 2014-01-01 00:00:00     0
  4. 2014-01-01 00:03:00     0
  5. 2014-01-01 00:06:00     0
  6. 2014-01-01 00:09:00     0
  7. 2014-01-01 00:12:00     0
  8.                        ..
  9. 2014-04-09 23:48:00     0
  10. 2014-04-09 23:51:00     0
  11. 2014-04-09 23:54:00     0
  12. 2014-04-09 23:57:00     0
  13. 2014-04-10 00:00:00    99
  14. Freq: 3T, Length: 47521, dtype: int64

We can instead only resample those groups where we have points as follows:

  1. In [305]: from functools import partial
  3. In [306]: from pandas.tseries.frequencies import to_offset
  5. In [307]: def round(t, freq):
  6.    .....:     freq = to_offset(freq)
  7.    .....:     return pd.Timestamp((t.value // freq.delta.value) * freq.delta.value)
  8.    .....: 
  10. In [308]: ts.groupby(partial(round, freq='3T')).sum()
  11. Out[308]: 
  12. 2014-01-01     0
  13. 2014-01-02     1
  14. 2014-01-03     2
  15. 2014-01-04     3
  16. 2014-01-05     4
  17.               ..
  18. 2014-04-06    95
  19. 2014-04-07    96
  20. 2014-04-08    97
  21. 2014-04-09    98
  22. 2014-04-10    99
  23. Length: 100, dtype: int64

Aggregation

Similar to the aggregating API, groupby API, and the window functions API, a Resampler can be selectively resampled.

Resampling a DataFrame, the default will be to act on all columns with the same function.

  1. In [309]: df = pd.DataFrame(np.random.randn(1000, 3),
  2.    .....:                   index=pd.date_range('1/1/2012', freq='S', periods=1000),
  3.    .....:                   columns=['A', 'B', 'C'])
  4.    .....: 
  6. In [310]: r = df.resample('3T')
  8. In [311]: r.mean()
  9. Out[311]: 
  10.                             A         B         C
  11. 2012-01-01 00:00:00 -0.033823 -0.121514 -0.081447
  12. 2012-01-01 00:03:00  0.056909  0.146731 -0.024320
  13. 2012-01-01 00:06:00 -0.058837  0.047046 -0.052021
  14. 2012-01-01 00:09:00  0.063123 -0.026158 -0.066533
  15. 2012-01-01 00:12:00  0.186340 -0.003144  0.074752
  16. 2012-01-01 00:15:00 -0.085954 -0.016287 -0.050046

We can select a specific column or columns using standard getitem.

  1. In [312]: r['A'].mean()
  2. Out[312]: 
  3. 2012-01-01 00:00:00   -0.033823
  4. 2012-01-01 00:03:00    0.056909
  5. 2012-01-01 00:06:00   -0.058837
  6. 2012-01-01 00:09:00    0.063123
  7. 2012-01-01 00:12:00    0.186340
  8. 2012-01-01 00:15:00   -0.085954
  9. Freq: 3T, Name: A, dtype: float64
  11. In [313]: r[['A', 'B']].mean()
  12. Out[313]: 
  13.                             A         B
  14. 2012-01-01 00:00:00 -0.033823 -0.121514
  15. 2012-01-01 00:03:00  0.056909  0.146731
  16. 2012-01-01 00:06:00 -0.058837  0.047046
  17. 2012-01-01 00:09:00  0.063123 -0.026158
  18. 2012-01-01 00:12:00  0.186340 -0.003144
  19. 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:

  1. In [314]: r['A'].agg([np.sum, np.mean, np.std])
  2. Out[314]: 
  3.                            sum      mean       std
  4. 2012-01-01 00:00:00  -6.088060 -0.033823  1.043263
  5. 2012-01-01 00:03:00  10.243678  0.056909  1.058534
  6. 2012-01-01 00:06:00 -10.590584 -0.058837  0.949264
  7. 2012-01-01 00:09:00  11.362228  0.063123  1.028096
  8. 2012-01-01 00:12:00  33.541257  0.186340  0.884586
  9. 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:

  1. In [315]: r.agg([np.sum, np.mean])
  2. Out[315]: 
  3.                              A                    B                    C          
  4.                            sum      mean        sum      mean        sum      mean
  5. 2012-01-01 00:00:00  -6.088060 -0.033823 -21.872530 -0.121514 -14.660515 -0.081447
  6. 2012-01-01 00:03:00  10.243678  0.056909  26.411633  0.146731  -4.377642 -0.024320
  7. 2012-01-01 00:06:00 -10.590584 -0.058837   8.468289  0.047046  -9.363825 -0.052021
  8. 2012-01-01 00:09:00  11.362228  0.063123  -4.708526 -0.026158 -11.975895 -0.066533
  9. 2012-01-01 00:12:00  33.541257  0.186340  -0.565895 -0.003144  13.455299  0.074752
  10. 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:

  1. In [316]: r.agg({'A': np.sum,
  2.    .....:        'B': lambda x: np.std(x, ddof=1)})
  3.    .....: 
  4. Out[316]: 
  5.                              A         B
  6. 2012-01-01 00:00:00  -6.088060  1.001294
  7. 2012-01-01 00:03:00  10.243678  1.074597
  8. 2012-01-01 00:06:00 -10.590584  0.987309
  9. 2012-01-01 00:09:00  11.362228  0.944953
  10. 2012-01-01 00:12:00  33.541257  1.095025
  11. 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:

  1. In [317]: r.agg({'A': 'sum', 'B': 'std'})
  2. Out[317]: 
  3.                              A         B
  4. 2012-01-01 00:00:00  -6.088060  1.001294
  5. 2012-01-01 00:03:00  10.243678  1.074597
  6. 2012-01-01 00:06:00 -10.590584  0.987309
  7. 2012-01-01 00:09:00  11.362228  0.944953
  8. 2012-01-01 00:12:00  33.541257  1.095025
  9. 2012-01-01 00:15:00  -8.595393  1.035312

Furthermore, you can also specify multiple aggregation functions for each column separately.

  1. In [318]: r.agg({'A': ['sum', 'std'], 'B': ['mean', 'std']})
  2. Out[318]: 
  3.                              A                   B          
  4.                            sum       std      mean       std
  5. 2012-01-01 00:00:00  -6.088060  1.043263 -0.121514  1.001294
  6. 2012-01-01 00:03:00  10.243678  1.058534  0.146731  1.074597
  7. 2012-01-01 00:06:00 -10.590584  0.949264  0.047046  0.987309
  8. 2012-01-01 00:09:00  11.362228  1.028096 -0.026158  0.944953
  9. 2012-01-01 00:12:00  33.541257  0.884586 -0.003144  1.095025
  10. 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.

  1. In [319]: df = pd.DataFrame({'date': pd.date_range('2015-01-01', freq='W', periods=5),
  2.    .....:                    'a': np.arange(5)},
  3.    .....:                   index=pd.MultiIndex.from_arrays([
  4.    .....:                       [1, 2, 3, 4, 5],
  5.    .....:                       pd.date_range('2015-01-01', freq='W', periods=5)],
  6.    .....:                       names=['v', 'd']))
  7.    .....: 
  9. In [320]: df
  10. Out[320]: 
  11.                    date  a
  12. v d                       
  13. 1 2015-01-04 2015-01-04  0
  14. 2 2015-01-11 2015-01-11  1
  15. 3 2015-01-18 2015-01-18  2
  16. 4 2015-01-25 2015-01-25  3
  17. 5 2015-02-01 2015-02-01  4
  19. In [321]: df.resample('M', on='date').sum()
  20. Out[321]: 
  21.             a
  22. date         
  23. 2015-01-31  6
  24. 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.

  1. In [322]: df.resample('M', level='d').sum()
  2. Out[322]: 
  3.             a
  4. d            
  5. 2015-01-31  6
  6. 2015-02-28  4

Iterating through groups

With the Resampler object in hand, iterating through the grouped data is very natural and functions similarly to itertools.groupby():

  1. In [323]: small = pd.Series(
  2.    .....:     range(6),
  3.    .....:     index=pd.to_datetime(['2017-01-01T00:00:00',
  4.    .....:                           '2017-01-01T00:30:00',
  5.    .....:                           '2017-01-01T00:31:00',
  6.    .....:                           '2017-01-01T01:00:00',
  7.    .....:                           '2017-01-01T03:00:00',
  8.    .....:                           '2017-01-01T03:05:00'])
  9.    .....: )
  10.    .....: 
  12. In [324]: resampled = small.resample('H')
  14. In [325]: for name, group in resampled:
  15.    .....:     print("Group: ", name)
  16.    .....:     print("-" * 27)
  17.    .....:     print(group, end="\n\n")
  18.    .....: 
  19. Group:  2017-01-01 00:00:00
  20. ---------------------------
  21. 2017-01-01 00:00:00    0
  22. 2017-01-01 00:30:00    1
  23. 2017-01-01 00:31:00    2
  24. dtype: int64
  26. Group:  2017-01-01 01:00:00
  27. ---------------------------
  28. 2017-01-01 01:00:00    3
  29. dtype: int64
  31. Group:  2017-01-01 02:00:00
  32. ---------------------------
  33. Series([], dtype: int64)
  35. Group:  2017-01-01 03:00:00
  36. ---------------------------
  37. 2017-01-01 03:00:00    4
  38. 2017-01-01 03:05:00    5
  39. dtype: int64

See Iterating through groups or Resampler.__iter__ for more.

Time span representation

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.

Period

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”.

  1. In [326]: pd.Period('2012', freq='A-DEC')
  2. Out[326]: Period('2012', 'A-DEC')
  4. In [327]: pd.Period('2012-1-1', freq='D')
  5. Out[327]: Period('2012-01-01', 'D')
  7. In [328]: pd.Period('2012-1-1 19:00', freq='H')
  8. Out[328]: Period('2012-01-01 19:00', 'H')
  10. In [329]: pd.Period('2012-1-1 19:00', freq='5H')
  11. Out[329]: 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).

  1. In [330]: p = pd.Period('2012', freq='A-DEC')
  3. In [331]: p + 1
  4. Out[331]: Period('2013', 'A-DEC')
  6. In [332]: p - 3
  7. Out[332]: Period('2009', 'A-DEC')
  9. In [333]: p = pd.Period('2012-01', freq='2M')
  11. In [334]: p + 2
  12. Out[334]: Period('2012-05', '2M')
  14. In [335]: p - 1
  15. Out[335]: Period('2011-11', '2M')
  17. In [336]: p == pd.Period('2012-01', freq='3M')
  18. ---------------------------------------------------------------------------
  19. IncompatibleFrequency                     Traceback (most recent call last)
  20. <ipython-input-336-4b67dc0b596c> in <module>
  21. ----> 1 p == pd.Period('2012-01', freq='3M')
  23. /pandas/pandas/_libs/tslibs/period.pyx in pandas._libs.tslibs.period._Period.__richcmp__()
  25. 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.

  1. In [337]: p = pd.Period('2014-07-01 09:00', freq='H')
  3. In [338]: p + pd.offsets.Hour(2)
  4. Out[338]: Period('2014-07-01 11:00', 'H')
  6. In [339]: p + datetime.timedelta(minutes=120)
  7. Out[339]: Period('2014-07-01 11:00', 'H')
  9. In [340]: p + np.timedelta64(7200, 's')
  10. Out[340]: Period('2014-07-01 11:00', 'H')
  1. In [1]: p + pd.offsets.Minute(5)
  2. Traceback
  3.    ...
  4. 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.

  1. In [341]: p = pd.Period('2014-07', freq='M')
  3. In [342]: p + pd.offsets.MonthEnd(3)
  4. Out[342]: Period('2014-10', 'M')
  1. In [1]: p + pd.offsets.MonthBegin(3)
  2. Traceback
  3.    ...
  4. 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:

  1. In [343]: pd.Period('2012', freq='A-DEC') - pd.Period('2002', freq='A-DEC')
  2. Out[343]: <10 * YearEnds: month=12>

PeriodIndex and period_range

Regular sequences of Period objects can be collected in a PeriodIndex, which can be constructed using the period_range convenience function:

  1. In [344]: prng = pd.period_range('1/1/2011', '1/1/2012', freq='M')
  3. In [345]: prng
  4. Out[345]: 
  5. PeriodIndex(['2011-01', '2011-02', '2011-03', '2011-04', '2011-05', '2011-06',
  6.              '2011-07', '2011-08', '2011-09', '2011-10', '2011-11', '2011-12',
  7.              '2012-01'],
  8.             dtype='period[M]', freq='M')

The PeriodIndex constructor can also be used directly:

  1. In [346]: pd.PeriodIndex(['2011-1', '2011-2', '2011-3'], freq='M')
  2. Out[346]: PeriodIndex(['2011-01', '2011-02', '2011-03'], dtype='period[M]', freq='M')

Passing multiplied frequency outputs a sequence of Period which has multiplied span.

  1. In [347]: pd.period_range(start='2014-01', freq='3M', periods=4)
  2. Out[347]: 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.

  1. In [348]: pd.period_range(start=pd.Period('2017Q1', freq='Q'),
  2.    .....:                 end=pd.Period('2017Q2', freq='Q'), freq='M')
  3.    .....: 
  4. Out[348]: 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:

  1. In [349]: ps = pd.Series(np.random.randn(len(prng)), prng)
  3. In [350]: ps
  4. Out[350]: 
  5. 2011-01   -2.916901
  6. 2011-02    0.514474
  7. 2011-03    1.346470
  8. 2011-04    0.816397
  9. 2011-05    2.258648
  10. 2011-06    0.494789
  11. 2011-07    0.301239
  12. 2011-08    0.464776
  13. 2011-09   -1.393581
  14. 2011-10    0.056780
  15. 2011-11    0.197035
  16. 2011-12    2.261385
  17. 2012-01   -0.329583
  18. Freq: M, dtype: float64

PeriodIndex supports addition and subtraction with the same rule as Period.

  1. In [351]: idx = pd.period_range('2014-07-01 09:00', periods=5, freq='H')
  3. In [352]: idx
  4. Out[352]: 
  5. PeriodIndex(['2014-07-01 09:00', '2014-07-01 10:00', '2014-07-01 11:00',
  6.              '2014-07-01 12:00', '2014-07-01 13:00'],
  7.             dtype='period[H]', freq='H')
  9. In [353]: idx + pd.offsets.Hour(2)
  10. Out[353]: 
  11. PeriodIndex(['2014-07-01 11:00', '2014-07-01 12:00', '2014-07-01 13:00',
  12.              '2014-07-01 14:00', '2014-07-01 15:00'],
  13.             dtype='period[H]', freq='H')
  15. In [354]: idx = pd.period_range('2014-07', periods=5, freq='M')
  17. In [355]: idx
  18. Out[355]: PeriodIndex(['2014-07', '2014-08', '2014-09', '2014-10', '2014-11'], dtype='period[M]', freq='M')
  20. In [356]: idx + pd.offsets.MonthEnd(3)
  21. Out[356]: 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 dtypes

New in version 0.19.0.

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.

  1. In [357]: pi = pd.period_range('2016-01-01', periods=3, freq='M')
  3. In [358]: pi
  4. Out[358]: PeriodIndex(['2016-01', '2016-02', '2016-03'], dtype='period[M]', freq='M')
  6. In [359]: pi.dtype
  7. Out[359]: 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():

  1. # change monthly freq to daily freq
  2. In [360]: pi.astype('period[D]')
  3. Out[360]: PeriodIndex(['2016-01-31', '2016-02-29', '2016-03-31'], dtype='period[D]', freq='D')
  5. # convert to DatetimeIndex
  6. In [361]: pi.astype('datetime64[ns]')
  7. Out[361]: DatetimeIndex(['2016-01-01', '2016-02-01', '2016-03-01'], dtype='datetime64[ns]', freq='MS')
  9. # convert to PeriodIndex
  10. In [362]: dti = pd.date_range('2011-01-01', freq='M', periods=3)
  12. In [363]: dti
  13. Out[363]: DatetimeIndex(['2011-01-31', '2011-02-28', '2011-03-31'], dtype='datetime64[ns]', freq='M')
  15. In [364]: dti.astype('period[M]')
  16. Out[364]: PeriodIndex(['2011-01', '2011-02', '2011-03'], dtype='period[M]', freq='M')

PeriodIndex partial string indexing

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.

  1. In [365]: ps['2011-01']
  2. Out[365]: -2.9169013294054507
  4. In [366]: ps[datetime.datetime(2011, 12, 25):]
  5. Out[366]: 
  6. 2011-12    2.261385
  7. 2012-01   -0.329583
  8. Freq: M, dtype: float64
  10. In [367]: ps['10/31/2011':'12/31/2011']
  11. Out[367]: 
  12. 2011-10    0.056780
  13. 2011-11    0.197035
  14. 2011-12    2.261385
  15. Freq: M, dtype: float64

Passing a string representing a lower frequency than PeriodIndex returns partial sliced data.

  1. In [368]: ps['2011']
  2. Out[368]: 
  3. 2011-01   -2.916901
  4. 2011-02    0.514474
  5. 2011-03    1.346470
  6. 2011-04    0.816397
  7. 2011-05    2.258648
  8. 2011-06    0.494789
  9. 2011-07    0.301239
  10. 2011-08    0.464776
  11. 2011-09   -1.393581
  12. 2011-10    0.056780
  13. 2011-11    0.197035
  14. 2011-12    2.261385
  15. Freq: M, dtype: float64
  17. In [369]: dfp = pd.DataFrame(np.random.randn(600, 1),
  18.    .....:                    columns=['A'],
  19.    .....:                    index=pd.period_range('2013-01-01 9:00',
  20.    .....:                                          periods=600,
  21.    .....:                                          freq='T'))
  22.    .....: 
  24. In [370]: dfp
  25. Out[370]: 
  26.                          A
  27. 2013-01-01 09:00 -0.538468
  28. 2013-01-01 09:01 -1.365819
  29. 2013-01-01 09:02 -0.969051
  30. 2013-01-01 09:03 -0.331152
  31. 2013-01-01 09:04 -0.245334
  32. ...                    ...
  33. 2013-01-01 18:55  0.522460
  34. 2013-01-01 18:56  0.118710
  35. 2013-01-01 18:57  0.167517
  36. 2013-01-01 18:58  0.922883
  37. 2013-01-01 18:59  1.721104
  39. [600 rows x 1 columns]
  41. In [371]: dfp['2013-01-01 10H']
  42. Out[371]: 
  43.                          A
  44. 2013-01-01 10:00 -0.308975
  45. 2013-01-01 10:01  0.542520
  46. 2013-01-01 10:02  1.061068
  47. 2013-01-01 10:03  0.754005
  48. 2013-01-01 10:04  0.352933
  49. ...                    ...
  50. 2013-01-01 10:55 -0.865621
  51. 2013-01-01 10:56 -1.167818
  52. 2013-01-01 10:57 -2.081748
  53. 2013-01-01 10:58 -0.527146
  54. 2013-01-01 10:59  0.802298
  56. [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.

  1. In [372]: dfp['2013-01-01 10H':'2013-01-01 11H']
  2. Out[372]: 
  3.                          A
  4. 2013-01-01 10:00 -0.308975
  5. 2013-01-01 10:01  0.542520
  6. 2013-01-01 10:02  1.061068
  7. 2013-01-01 10:03  0.754005
  8. 2013-01-01 10:04  0.352933
  9. ...                    ...
  10. 2013-01-01 11:55 -0.590204
  11. 2013-01-01 11:56  1.539990
  12. 2013-01-01 11:57 -1.224826
  13. 2013-01-01 11:58  0.578798
  14. 2013-01-01 11:59 -0.685496
  16. [120 rows x 1 columns]

Frequency conversion and resampling with PeriodIndex

The frequency of Period and PeriodIndex can be converted via the asfreq method. Let’s start with the fiscal year 2011, ending in December:

  1. In [373]: p = pd.Period('2011', freq='A-DEC')
  3. In [374]: p
  4. Out[374]: 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:

  1. In [375]: p.asfreq('M', how='start')
  2. Out[375]: Period('2011-01', 'M')
  4. In [376]: p.asfreq('M', how='end')
  5. Out[376]: Period('2011-12', 'M')

The shorthands ‘s’ and ‘e’ are provided for convenience:

  1. In [377]: p.asfreq('M', 's')
  2. Out[377]: Period('2011-01', 'M')
  4. In [378]: p.asfreq('M', 'e')
  5. Out[378]: 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:

  1. In [379]: p = pd.Period('2011-12', freq='M')
  3. In [380]: p.asfreq('A-NOV')
  4. Out[380]: 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-DEC define regular calendar quarters:

  1. In [381]: p = pd.Period('2012Q1', freq='Q-DEC')
  3. In [382]: p.asfreq('D', 's')
  4. Out[382]: Period('2012-01-01', 'D')
  6. In [383]: p.asfreq('D', 'e')
  7. Out[383]: Period('2012-03-31', 'D')

Q-MAR defines fiscal year end in March:

  1. In [384]: p = pd.Period('2011Q4', freq='Q-MAR')
  3. In [385]: p.asfreq('D', 's')
  4. Out[385]: Period('2011-01-01', 'D')
  6. In [386]: p.asfreq('D', 'e')
  7. Out[386]: Period('2011-03-31', 'D')

Converting between representations

Timestamped data can be converted to PeriodIndex-ed data using to_period and vice-versa using to_timestamp:

  1. In [387]: rng = pd.date_range('1/1/2012', periods=5, freq='M')
  3. In [388]: ts = pd.Series(np.random.randn(len(rng)), index=rng)
  5. In [389]: ts
  6. Out[389]: 
  7. 2012-01-31    1.931253
  8. 2012-02-29   -0.184594
  9. 2012-03-31    0.249656
  10. 2012-04-30   -0.978151
  11. 2012-05-31   -0.873389
  12. Freq: M, dtype: float64
  14. In [390]: ps = ts.to_period()
  16. In [391]: ps
  17. Out[391]: 
  18. 2012-01    1.931253
  19. 2012-02   -0.184594
  20. 2012-03    0.249656
  21. 2012-04   -0.978151
  22. 2012-05   -0.873389
  23. Freq: M, dtype: float64
  25. In [392]: ps.to_timestamp()
  26. Out[392]: 
  27. 2012-01-01    1.931253
  28. 2012-02-01   -0.184594
  29. 2012-03-01    0.249656
  30. 2012-04-01   -0.978151
  31. 2012-05-01   -0.873389
  32. Freq: MS, dtype: float64

Remember that ‘s’ and ‘e’ can be used to return the timestamps at the start or end of the period:

  1. In [393]: ps.to_timestamp('D', how='s')
  2. Out[393]: 
  3. 2012-01-01    1.931253
  4. 2012-02-01   -0.184594
  5. 2012-03-01    0.249656
  6. 2012-04-01   -0.978151
  7. 2012-05-01   -0.873389
  8. 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:

  1. In [394]: prng = pd.period_range('1990Q1', '2000Q4', freq='Q-NOV')
  3. In [395]: ts = pd.Series(np.random.randn(len(prng)), prng)
  5. In [396]: ts.index = (prng.asfreq('M', 'e') + 1).asfreq('H', 's') + 9
  7. In [397]: ts.head()
  8. Out[397]: 
  9. 1990-03-01 09:00   -0.109291
  10. 1990-06-01 09:00   -0.637235
  11. 1990-09-01 09:00   -1.735925
  12. 1990-12-01 09:00    2.096946
  13. 1991-03-01 09:00   -1.039926
  14. Freq: H, dtype: float64

Representing out-of-bounds spans

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.

  1. In [398]: span = pd.period_range('1215-01-01', '1381-01-01', freq='D')
  3. In [399]: span
  4. Out[399]: 
  5. PeriodIndex(['1215-01-01', '1215-01-02', '1215-01-03', '1215-01-04',
  6.              '1215-01-05', '1215-01-06', '1215-01-07', '1215-01-08',
  7.              '1215-01-09', '1215-01-10',
  8.              ...
  9.              '1380-12-23', '1380-12-24', '1380-12-25', '1380-12-26',
  10.              '1380-12-27', '1380-12-28', '1380-12-29', '1380-12-30',
  11.              '1380-12-31', '1381-01-01'],
  12.             dtype='period[D]', length=60632, freq='D')

To convert from an int64 based YYYYMMDD representation.

  1. In [400]: s = pd.Series([20121231, 20141130, 99991231])
  3. In [401]: s
  4. Out[401]: 
  5. 0    20121231
  6. 1    20141130
  7. 2    99991231
  8. dtype: int64
  10. In [402]: def conv(x):
  11.    .....:     return pd.Period(year=x // 10000, month=x // 100 % 100,
  12.    .....:                      day=x % 100, freq='D')
  13.    .....: 
  15. In [403]: s.apply(conv)
  16. Out[403]: 
  17. 0    2012-12-31
  18. 1    2014-11-30
  19. 2    9999-12-31
  20. dtype: period[D]
  22. In [404]: s.apply(conv)[2]
  23. Out[404]: Period('9999-12-31', 'D')

These can easily be converted to a PeriodIndex:

  1. In [405]: span = pd.PeriodIndex(s.apply(conv))
  3. In [406]: span
  4. Out[406]: PeriodIndex(['2012-12-31', '2014-11-30', '9999-12-31'], dtype='period[D]', freq='D')

Time zone handling

pandas provides rich support for working with timestamps in different time zones using the pytz and dateutil libraries or class:datetime.timezone objects from the standard library.

Working with time zones

By default, pandas objects are time zone unaware:

  1. In [407]: rng = pd.date_range('3/6/2012 00:00', periods=15, freq='D')
  3. In [408]: rng.tz is None
  4. Out[408]: 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.

  • In pytz you can find a list of common (and less common) time zones using 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.
  1. In [409]: import dateutil
  3. # pytz
  4. In [410]: rng_pytz = pd.date_range('3/6/2012 00:00', periods=3, freq='D',
  5.    .....:                          tz='Europe/London')
  6.    .....: 
  8. In [411]: rng_pytz.tz
  9. Out[411]: <DstTzInfo 'Europe/London' LMT-1 day, 23:59:00 STD>
  11. # dateutil
  12. In [412]: rng_dateutil = pd.date_range('3/6/2012 00:00', periods=3, freq='D')
  14. In [413]: rng_dateutil = rng_dateutil.tz_localize('dateutil/Europe/London')
  16. In [414]: rng_dateutil.tz
  17. Out[414]: tzfile('/usr/share/zoneinfo/Europe/London')
  19. # dateutil - utc special case
  20. In [415]: rng_utc = pd.date_range('3/6/2012 00:00', periods=3, freq='D',
  21.    .....:                         tz=dateutil.tz.tzutc())
  22.    .....: 
  24. In [416]: rng_utc.tz
  25. Out[416]: tzutc()

New in version 0.25.0. 

  1. # datetime.timezone
  2. In [417]: rng_utc = pd.date_range('3/6/2012 00:00', periods=3, freq='D',
  3.    .....:                         tz=datetime.timezone.utc)
  4.    .....: 
  6. In [418]: rng_utc.tz
  7. Out[418]: 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.

  1. In [419]: import pytz
  3. # pytz
  4. In [420]: tz_pytz = pytz.timezone('Europe/London')
  6. In [421]: rng_pytz = pd.date_range('3/6/2012 00:00', periods=3, freq='D')
  8. In [422]: rng_pytz = rng_pytz.tz_localize(tz_pytz)
  10. In [423]: rng_pytz.tz == tz_pytz
  11. Out[423]: True
  13. # dateutil
  14. In [424]: tz_dateutil = dateutil.tz.gettz('Europe/London')
  16. In [425]: rng_dateutil = pd.date_range('3/6/2012 00:00', periods=3, freq='D',
  17.    .....:                              tz=tz_dateutil)
  18.    .....: 
  20. In [426]: rng_dateutil.tz == tz_dateutil
  21. Out[426]: True

To convert a time zone aware pandas object from one time zone to another, you can use the tz_convert method.

  1. In [427]: rng_pytz.tz_convert('US/Eastern')
  2. Out[427]: 
  3. DatetimeIndex(['2012-03-05 19:00:00-05:00', '2012-03-06 19:00:00-05:00',
  4.                '2012-03-07 19:00:00-05:00'],
  5.               dtype='datetime64[ns, US/Eastern]', freq='D')

::: tip Note

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.

  1. In [428]: dti = pd.date_range('2019-01-01', periods=3, freq='D', tz='US/Pacific')
  3. In [429]: dti.tz
  4. Out[429]: <DstTzInfo 'US/Pacific' LMT-1 day, 16:07:00 STD>
  6. In [430]: ts = pd.Timestamp('2019-01-01', tz='US/Pacific')
  8. In [431]: ts.tz
  9. Out[431]: <DstTzInfo 'US/Pacific' PST-1 day, 16:00:00 STD>

:::

::: danger Warning

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.

:::

::: danger Warning

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.

:::

::: danger Warning

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.

:::

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:

  1. In [432]: rng_eastern = rng_utc.tz_convert('US/Eastern')
  3. In [433]: rng_berlin = rng_utc.tz_convert('Europe/Berlin')
  5. In [434]: rng_eastern[2]
  6. Out[434]: Timestamp('2012-03-07 19:00:00-0500', tz='US/Eastern', freq='D')
  8. In [435]: rng_berlin[2]
  9. Out[435]: Timestamp('2012-03-08 01:00:00+0100', tz='Europe/Berlin', freq='D')
  11. In [436]: rng_eastern[2] == rng_berlin[2]
  12. Out[436]: True

Operations between Series in different time zones will yield UTC Series, aligning the data on the UTC timestamps:

  1. In [437]: ts_utc = pd.Series(range(3), pd.date_range('20130101', periods=3, tz='UTC'))
  3. In [438]: eastern = ts_utc.tz_convert('US/Eastern')
  5. In [439]: berlin = ts_utc.tz_convert('Europe/Berlin')
  7. In [440]: result = eastern + berlin
  9. In [441]: result
  10. Out[441]: 
  11. 2013-01-01 00:00:00+00:00    0
  12. 2013-01-02 00:00:00+00:00    2
  13. 2013-01-03 00:00:00+00:00    4
  14. Freq: D, dtype: int64
  16. In [442]: result.index
  17. Out[442]: 
  18. DatetimeIndex(['2013-01-01 00:00:00+00:00', '2013-01-02 00:00:00+00:00',
  19.                '2013-01-03 00:00:00+00:00'],
  20.               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.

  1. In [443]: didx = pd.date_range(start='2014-08-01 09:00', freq='H',
  2.    .....:                      periods=3, tz='US/Eastern')
  3.    .....: 
  5. In [444]: didx
  6. Out[444]: 
  7. DatetimeIndex(['2014-08-01 09:00:00-04:00', '2014-08-01 10:00:00-04:00',
  8.                '2014-08-01 11:00:00-04:00'],
  9.               dtype='datetime64[ns, US/Eastern]', freq='H')
  11. In [445]: didx.tz_localize(None)
  12. Out[445]: 
  13. DatetimeIndex(['2014-08-01 09:00:00', '2014-08-01 10:00:00',
  14.                '2014-08-01 11:00:00'],
  15.               dtype='datetime64[ns]', freq='H')
  17. In [446]: didx.tz_convert(None)
  18. Out[446]: 
  19. DatetimeIndex(['2014-08-01 13:00:00', '2014-08-01 14:00:00',
  20.                '2014-08-01 15:00:00'],
  21.               dtype='datetime64[ns]', freq='H')
  23. # tz_convert(None) is identical to tz_convert('UTC').tz_localize(None)
  24. In [447]: didx.tz_convert('UTC').tz_localize(None)
  25. Out[447]: 
  26. DatetimeIndex(['2014-08-01 13:00:00', '2014-08-01 14:00:00',
  27.                '2014-08-01 15:00:00'],
  28.               dtype='datetime64[ns]', freq='H')

Ambiguous times when localizing

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)
  • 'infer': Attempt to determine the correct offset base on the monotonicity of the timestamps
  • 'NaT': Replaces ambiguous times with 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.
  1. In [448]: rng_hourly = pd.DatetimeIndex(['11/06/2011 00:00', '11/06/2011 01:00',
  2.    .....:                                '11/06/2011 01:00', '11/06/2011 02:00'])
  3.    .....:

This will fail as there are ambiguous times ('11/06/2011 01:00')

  1. In [2]: rng_hourly.tz_localize('US/Eastern')
  2. 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.

  1. In [449]: rng_hourly.tz_localize('US/Eastern', ambiguous='infer')
  2. Out[449]: 
  3. DatetimeIndex(['2011-11-06 00:00:00-04:00', '2011-11-06 01:00:00-04:00',
  4.                '2011-11-06 01:00:00-05:00', '2011-11-06 02:00:00-05:00'],
  5.               dtype='datetime64[ns, US/Eastern]', freq=None)
  7. In [450]: rng_hourly.tz_localize('US/Eastern', ambiguous='NaT')
  8. Out[450]: 
  9. DatetimeIndex(['2011-11-06 00:00:00-04:00', 'NaT', 'NaT',
  10.                '2011-11-06 02:00:00-05:00'],
  11.               dtype='datetime64[ns, US/Eastern]', freq=None)
  13. In [451]: rng_hourly.tz_localize('US/Eastern', ambiguous=[True, True, False, False])
  14. Out[451]: 
  15. DatetimeIndex(['2011-11-06 00:00:00-04:00', '2011-11-06 01:00:00-04:00',
  16.                '2011-11-06 01:00:00-05:00', '2011-11-06 02:00:00-05:00'],
  17.               dtype='datetime64[ns, US/Eastern]', freq=None)

Nonexistent times when localizing

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:

  • 'raise': Raises a pytz.NonExistentTimeError (the default behavior)
  • 'NaT': Replaces nonexistent times with NaT
  • 'shift_forward': Shifts nonexistent times forward to the closest real time
  • 'shift_backward': Shifts nonexistent times backward to the closest real time
  • timedelta object: Shifts nonexistent times by the timedelta duration
  1. In [452]: dti = pd.date_range(start='2015-03-29 02:30:00', periods=3, freq='H')
  3. # 2:30 is a nonexistent time

Localization of nonexistent times will raise an error by default.

  1. In [2]: dti.tz_localize('Europe/Warsaw')
  2. NonExistentTimeError: 2015-03-29 02:30:00

Transform nonexistent times to NaT or shift the times.

  1. In [453]: dti
  2. Out[453]: 
  3. DatetimeIndex(['2015-03-29 02:30:00', '2015-03-29 03:30:00',
  4.                '2015-03-29 04:30:00'],
  5.               dtype='datetime64[ns]', freq='H')
  7. In [454]: dti.tz_localize('Europe/Warsaw', nonexistent='shift_forward')
  8. Out[454]: 
  9. DatetimeIndex(['2015-03-29 03:00:00+02:00', '2015-03-29 03:30:00+02:00',
  10.                '2015-03-29 04:30:00+02:00'],
  11.               dtype='datetime64[ns, Europe/Warsaw]', freq='H')
  13. In [455]: dti.tz_localize('Europe/Warsaw', nonexistent='shift_backward')
  14. Out[455]: 
  15. DatetimeIndex(['2015-03-29 01:59:59.999999999+01:00',
  16.                          '2015-03-29 03:30:00+02:00',
  17.                          '2015-03-29 04:30:00+02:00'],
  18.               dtype='datetime64[ns, Europe/Warsaw]', freq='H')
  20. In [456]: dti.tz_localize('Europe/Warsaw', nonexistent=pd.Timedelta(1, unit='H'))
  21. Out[456]: 
  22. DatetimeIndex(['2015-03-29 03:30:00+02:00', '2015-03-29 03:30:00+02:00',
  23.                '2015-03-29 04:30:00+02:00'],
  24.               dtype='datetime64[ns, Europe/Warsaw]', freq='H')
  26. In [457]: dti.tz_localize('Europe/Warsaw', nonexistent='NaT')
  27. Out[457]: 
  28. DatetimeIndex(['NaT', '2015-03-29 03:30:00+02:00',
  29.                '2015-03-29 04:30:00+02:00'],
  30.               dtype='datetime64[ns, Europe/Warsaw]', freq='H')

Time zone series operations

A Series with time zone naive values is represented with a dtype of datetime64[ns].

  1. In [458]: s_naive = pd.Series(pd.date_range('20130101', periods=3))
  3. In [459]: s_naive
  4. Out[459]: 
  5. 0   2013-01-01
  6. 1   2013-01-02
  7. 2   2013-01-03
  8. 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

  1. In [460]: s_aware = pd.Series(pd.date_range('20130101', periods=3, tz='US/Eastern'))
  3. In [461]: s_aware
  4. Out[461]: 
  5. 0   2013-01-01 00:00:00-05:00
  6. 1   2013-01-02 00:00:00-05:00
  7. 2   2013-01-03 00:00:00-05:00
  8. 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.

  1. In [462]: s_naive.dt.tz_localize('UTC').dt.tz_convert('US/Eastern')
  2. Out[462]: 
  3. 0   2012-12-31 19:00:00-05:00
  4. 1   2013-01-01 19:00:00-05:00
  5. 2   2013-01-02 19:00:00-05:00
  6. 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.

  1. # localize and convert a naive time zone
  2. In [463]: s_naive.astype('datetime64[ns, US/Eastern]')
  3. Out[463]: 
  4. 0   2012-12-31 19:00:00-05:00
  5. 1   2013-01-01 19:00:00-05:00
  6. 2   2013-01-02 19:00:00-05:00
  7. dtype: datetime64[ns, US/Eastern]
  9. # make an aware tz naive
  10. In [464]: s_aware.astype('datetime64[ns]')
  11. Out[464]: 
  12. 0   2013-01-01 05:00:00
  13. 1   2013-01-02 05:00:00
  14. 2   2013-01-03 05:00:00
  15. dtype: datetime64[ns]
  17. # convert to a new time zone
  18. In [465]: s_aware.astype('datetime64[ns, CET]')
  19. Out[465]: 
  20. 0   2013-01-01 06:00:00+01:00
  21. 1   2013-01-02 06:00:00+01:00
  22. 2   2013-01-03 06:00:00+01:00
  23. dtype: datetime64[ns, CET]

::: tip Note

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:

  1. In [466]: s_naive.to_numpy()
  2. Out[466]: 
  3. array(['2013-01-01T00:00:00.000000000', '2013-01-02T00:00:00.000000000',
  4.        '2013-01-03T00:00:00.000000000'], dtype='datetime64[ns]')
  6. In [467]: s_aware.to_numpy()
  7. Out[467]: 
  8. array([Timestamp('2013-01-01 00:00:00-0500', tz='US/Eastern', freq='D'),
  9.        Timestamp('2013-01-02 00:00:00-0500', tz='US/Eastern', freq='D'),
  10.        Timestamp('2013-01-03 00:00:00-0500', tz='US/Eastern', freq='D')],
  11.       dtype=object)

By converting to an object array of Timestamps, it preserves the time zone information. For example, when converting back to a Series:

  1. In [468]: pd.Series(s_aware.to_numpy())
  2. Out[468]: 
  3. 0   2013-01-01 00:00:00-05:00
  4. 1   2013-01-02 00:00:00-05:00
  5. 2   2013-01-03 00:00:00-05:00
  6. 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:

  1. In [469]: s_aware.to_numpy(dtype='datetime64[ns]')
  2. Out[469]: 
  3. array(['2013-01-01T05:00:00.000000000', '2013-01-02T05:00:00.000000000',
  4.        '2013-01-03T05:00:00.000000000'], dtype='datetime64[ns]')

:::