Time Series / Date functionality — pandas 0.18.1 documentation (original) (raw)

pandas has proven very successful as a tool for working with time series data, especially in the financial data analysis space. Using the NumPy datetime64 and timedelta64 dtypes, we have 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.

In working with time series data, we will frequently seek to:

• generate sequences of fixed-frequency dates and time spans
• conform or convert time series to a particular frequency
• compute “relative” dates based on various non-standard time increments (e.g. 5 business days before the last business day of the year), or “roll” dates forward or backward

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

Create a range of dates:

72 hours starting with midnight Jan 1st, 2011

In [1]: rng = pd.date_range('1/1/2011', periods=72, freq='H')

In [2]: rng[:5] Out[2]: DatetimeIndex(['2011-01-01 00:00:00', '2011-01-01 01:00:00', '2011-01-01 02:00:00', '2011-01-01 03:00:00', '2011-01-01 04:00:00'], dtype='datetime64[ns]', freq='H')

Index pandas objects with dates:

In [3]: ts = pd.Series(np.random.randn(len(rng)), index=rng)

In [4]: ts.head() Out[4]: 2011-01-01 00:00:00 0.469112 2011-01-01 01:00:00 -0.282863 2011-01-01 02:00:00 -1.509059 2011-01-01 03:00:00 -1.135632 2011-01-01 04:00:00 1.212112 Freq: H, dtype: float64

Change frequency and fill gaps:

to 45 minute frequency and forward fill

In [5]: converted = ts.asfreq('45Min', method='pad')

In [6]: converted.head() Out[6]: 2011-01-01 00:00:00 0.469112 2011-01-01 00:45:00 0.469112 2011-01-01 01:30:00 -0.282863 2011-01-01 02:15:00 -1.509059 2011-01-01 03:00:00 -1.135632 Freq: 45T, dtype: float64

Resample:

Daily means

In [7]: ts.resample('D').mean() Out[7]: 2011-01-01 -0.319569 2011-01-02 -0.337703 2011-01-03 0.117258 Freq: D, dtype: float64

Overview¶

Following table shows the type of time-related classes pandas can handle and how to create them.

Time Stamps vs. Time Spans¶

Time-stamped data is the most basic type of timeseries data that associates values with points in time. For pandas objects it means using the points in time.

In [8]: pd.Timestamp(datetime(2012, 5, 1)) Out[8]: Timestamp('2012-05-01 00:00:00')

In [9]: pd.Timestamp('2012-05-01') Out[9]: Timestamp('2012-05-01 00:00:00')

However, in many cases it is more natural to associate things like change variables with a time span instead. The span represented by Period can be specified explicitly, or inferred from datetime string format.

For example:

In [10]: pd.Period('2011-01') Out[10]: Period('2011-01', 'M')

In [11]: pd.Period('2012-05', freq='D') Out[11]: Period('2012-05-01', 'D')

Timestamp and Period can be the index. Lists of Timestamp andPeriod are automatically coerce to DatetimeIndex and PeriodIndexrespectively.

In [12]: dates = [pd.Timestamp('2012-05-01'), pd.Timestamp('2012-05-02'), pd.Timestamp('2012-05-03')]

In [13]: ts = pd.Series(np.random.randn(3), dates)

In [14]: type(ts.index) Out[14]: pandas.tseries.index.DatetimeIndex

In [15]: ts.index Out[15]: DatetimeIndex(['2012-05-01', '2012-05-02', '2012-05-03'], dtype='datetime64[ns]', freq=None)

In [16]: ts Out[16]: 2012-05-01 -0.410001 2012-05-02 -0.078638 2012-05-03 0.545952 dtype: float64

In [17]: periods = [pd.Period('2012-01'), pd.Period('2012-02'), pd.Period('2012-03')]

In [18]: ts = pd.Series(np.random.randn(3), periods)

In [19]: type(ts.index) Out[19]: pandas.tseries.period.PeriodIndex

In [20]: ts.index Out[20]: PeriodIndex(['2012-01', '2012-02', '2012-03'], dtype='int64', freq='M')

In [21]: ts Out[21]: 2012-01 -1.219217 2012-02 -1.226825 2012-03 0.769804 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 ofDatetimeIndex. 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:

In [22]: pd.to_datetime(pd.Series(['Jul 31, 2009', '2010-01-10', None])) Out[22]: 0 2009-07-31 1 2010-01-10 2 NaT dtype: datetime64[ns]

In [23]: pd.to_datetime(['2005/11/23', '2010.12.31']) Out[23]: 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:

In [24]: pd.to_datetime(['04-01-2012 10:00'], dayfirst=True) Out[24]: DatetimeIndex(['2012-01-04 10:00:00'], dtype='datetime64[ns]', freq=None)

In [25]: pd.to_datetime(['14-01-2012', '01-14-2012'], dayfirst=True) Out[25]: DatetimeIndex(['2012-01-14', '2012-01-14'], dtype='datetime64[ns]', freq=None)

Warning

You see in the above example that dayfirst isn’t strict, so if a date can’t be parsed with the day being first it will be parsed as ifdayfirst were False.

Note

Specifying a format argument will potentially speed up the conversion considerably and on versions later then 0.13.0 explicitly specifying a format string of ‘%Y%m%d’ takes a faster path still.

If you pass a single string to to_datetime, it returns single Timestamp. Also, Timestamp can accept the string input. Note that Timestamp doesn’t accept string parsing option like dayfirstor format, use to_datetime if these are required.

In [26]: pd.to_datetime('2010/11/12') Out[26]: Timestamp('2010-11-12 00:00:00')

In [27]: pd.Timestamp('2010/11/12') Out[27]: Timestamp('2010-11-12 00:00:00')

New in version 0.18.1.

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

In [28]: df = pd.DataFrame({'year': [2015, 2016], ....: 'month': [2, 3], ....: 'day': [4, 5], ....: 'hour': [2, 3]}) ....:

In [29]: pd.to_datetime(df) Out[29]: 0 2015-02-04 02:00:00 1 2016-03-05 03:00:00 dtype: datetime64[ns]

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

In [30]: pd.to_datetime(df[['year', 'month', 'day']]) Out[30]: 0 2015-02-04 1 2016-03-05 dtype: datetime64[ns]

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

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

Invalid Data¶

Note

In version 0.17.0, the default for to_datetime is now errors='raise', rather than errors='ignore'. This means that invalid parsing will raise rather that return the original input as in previous versions.

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

Raise when unparseable, this is the default

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

Return the original input when unparseable

In [4]: pd.to_datetime(['2009/07/31', 'asd'], errors='ignore') Out[4]: array(['2009/07/31', 'asd'], dtype=object)

Return NaT for input when unparseable

In [6]: pd.to_datetime(['2009/07/31', 'asd'], errors='coerce') Out[6]: DatetimeIndex(['2009-07-31', 'NaT'], dtype='datetime64[ns]', freq=None)

Epoch Timestamps¶

It’s also possible to convert integer or float epoch times. The default unit for these is nanoseconds (since these are how Timestamp s are stored). However, often epochs are stored in another unit which can be specified:

Typical epoch stored units

In [31]: pd.to_datetime([1349720105, 1349806505, 1349892905, ....: 1349979305, 1350065705], unit='s') ....: Out[31]: DatetimeIndex(['2012-10-08 18:15:05', '2012-10-09 18:15:05', '2012-10-10 18:15:05', '2012-10-11 18:15:05', '2012-10-12 18:15:05'], dtype='datetime64[ns]', freq=None)

In [32]: pd.to_datetime([1349720105100, 1349720105200, 1349720105300, ....: 1349720105400, 1349720105500 ], unit='ms') ....: Out[32]: DatetimeIndex(['2012-10-08 18:15:05.100000', '2012-10-08 18:15:05.200000', '2012-10-08 18:15:05.300000', '2012-10-08 18:15:05.400000', '2012-10-08 18:15:05.500000'], dtype='datetime64[ns]', freq=None)

These work, but the results may be unexpected.

In [33]: pd.to_datetime([1]) Out[33]: DatetimeIndex(['1970-01-01 00:00:00.000000001'], dtype='datetime64[ns]', freq=None)

In [34]: pd.to_datetime([1, 3.14], unit='s') Out[34]: DatetimeIndex(['1970-01-01 00:00:01', '1970-01-01 00:00:03'], dtype='datetime64[ns]', freq=None)

Note

Epoch times will be rounded to the nearest nanosecond.

Generating Ranges of Timestamps¶

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

In [35]: dates = [datetime(2012, 5, 1), datetime(2012, 5, 2), datetime(2012, 5, 3)]

Note the frequency information

In [36]: index = pd.DatetimeIndex(dates)

In [37]: index Out[37]: DatetimeIndex(['2012-05-01', '2012-05-02', '2012-05-03'], dtype='datetime64[ns]', freq=None)

Automatically converted to DatetimeIndex

In [38]: index = pd.Index(dates)

In [39]: index Out[39]: DatetimeIndex(['2012-05-01', '2012-05-02', '2012-05-03'], dtype='datetime64[ns]', freq=None)

Practically, 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 pandas functions date_range and bdate_rangeto create timestamp indexes.

In [40]: index = pd.date_range('2000-1-1', periods=1000, freq='M')

In [41]: index Out[41]: DatetimeIndex(['2000-01-31', '2000-02-29', '2000-03-31', '2000-04-30', '2000-05-31', '2000-06-30', '2000-07-31', '2000-08-31', '2000-09-30', '2000-10-31', ... '2082-07-31', '2082-08-31', '2082-09-30', '2082-10-31', '2082-11-30', '2082-12-31', '2083-01-31', '2083-02-28', '2083-03-31', '2083-04-30'], dtype='datetime64[ns]', length=1000, freq='M')

In [42]: index = pd.bdate_range('2012-1-1', periods=250)

In [43]: index Out[43]: DatetimeIndex(['2012-01-02', '2012-01-03', '2012-01-04', '2012-01-05', '2012-01-06', '2012-01-09', '2012-01-10', '2012-01-11', '2012-01-12', '2012-01-13', ... '2012-12-03', '2012-12-04', '2012-12-05', '2012-12-06', '2012-12-07', '2012-12-10', '2012-12-11', '2012-12-12', '2012-12-13', '2012-12-14'], dtype='datetime64[ns]', length=250, freq='B')

Convenience functions like date_range and bdate_range utilize a variety of frequency aliases. The default frequency for date_range is acalendar day while the default for bdate_range is a business day

In [44]: start = datetime(2011, 1, 1)

In [45]: end = datetime(2012, 1, 1)

In [46]: rng = pd.date_range(start, end)

In [47]: rng Out[47]: DatetimeIndex(['2011-01-01', '2011-01-02', '2011-01-03', '2011-01-04', '2011-01-05', '2011-01-06', '2011-01-07', '2011-01-08', '2011-01-09', '2011-01-10', ... '2011-12-23', '2011-12-24', '2011-12-25', '2011-12-26', '2011-12-27', '2011-12-28', '2011-12-29', '2011-12-30', '2011-12-31', '2012-01-01'], dtype='datetime64[ns]', length=366, freq='D')

In [48]: rng = pd.bdate_range(start, end)

In [49]: rng Out[49]: DatetimeIndex(['2011-01-03', '2011-01-04', '2011-01-05', '2011-01-06', '2011-01-07', '2011-01-10', '2011-01-11', '2011-01-12', '2011-01-13', '2011-01-14', ... '2011-12-19', '2011-12-20', '2011-12-21', '2011-12-22', '2011-12-23', '2011-12-26', '2011-12-27', '2011-12-28', '2011-12-29', '2011-12-30'], dtype='datetime64[ns]', length=260, freq='B')

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:

In [50]: pd.date_range(start, end, freq='BM') Out[50]: DatetimeIndex(['2011-01-31', '2011-02-28', '2011-03-31', '2011-04-29', '2011-05-31', '2011-06-30', '2011-07-29', '2011-08-31', '2011-09-30', '2011-10-31', '2011-11-30', '2011-12-30'], dtype='datetime64[ns]', freq='BM')

In [51]: pd.date_range(start, end, freq='W') Out[51]: DatetimeIndex(['2011-01-02', '2011-01-09', '2011-01-16', '2011-01-23', '2011-01-30', '2011-02-06', '2011-02-13', '2011-02-20', '2011-02-27', '2011-03-06', '2011-03-13', '2011-03-20', '2011-03-27', '2011-04-03', '2011-04-10', '2011-04-17', '2011-04-24', '2011-05-01', '2011-05-08', '2011-05-15', '2011-05-22', '2011-05-29', '2011-06-05', '2011-06-12', '2011-06-19', '2011-06-26', '2011-07-03', '2011-07-10', '2011-07-17', '2011-07-24', '2011-07-31', '2011-08-07', '2011-08-14', '2011-08-21', '2011-08-28', '2011-09-04', '2011-09-11', '2011-09-18', '2011-09-25', '2011-10-02', '2011-10-09', '2011-10-16', '2011-10-23', '2011-10-30', '2011-11-06', '2011-11-13', '2011-11-20', '2011-11-27', '2011-12-04', '2011-12-11', '2011-12-18', '2011-12-25', '2012-01-01'], dtype='datetime64[ns]', freq='W-SUN')

In [52]: pd.bdate_range(end=end, periods=20) Out[52]: DatetimeIndex(['2011-12-05', '2011-12-06', '2011-12-07', '2011-12-08', '2011-12-09', '2011-12-12', '2011-12-13', '2011-12-14', '2011-12-15', '2011-12-16', '2011-12-19', '2011-12-20', '2011-12-21', '2011-12-22', '2011-12-23', '2011-12-26', '2011-12-27', '2011-12-28', '2011-12-29', '2011-12-30'], dtype='datetime64[ns]', freq='B')

In [53]: pd.bdate_range(start=start, periods=20) Out[53]: DatetimeIndex(['2011-01-03', '2011-01-04', '2011-01-05', '2011-01-06', '2011-01-07', '2011-01-10', '2011-01-11', '2011-01-12', '2011-01-13', '2011-01-14', '2011-01-17', '2011-01-18', '2011-01-19', '2011-01-20', '2011-01-21', '2011-01-24', '2011-01-25', '2011-01-26', '2011-01-27', '2011-01-28'], dtype='datetime64[ns]', freq='B')

The start and end dates are strictly inclusive. So it will not generate any dates outside of those dates if specified.

Timestamp limitations¶

Since pandas represents timestamps in nanosecond resolution, the timespan that can be represented using a 64-bit integer is limited to approximately 584 years:

In [54]: pd.Timestamp.min Out[54]: Timestamp('1677-09-22 00:12:43.145225')

In [55]: pd.Timestamp.max Out[55]: Timestamp('2262-04-11 23:47:16.854775807')

See here for ways to represent data outside these bound.

DatetimeIndex¶

One of the main uses for DatetimeIndex is as an index for pandas objects. The DatetimeIndex class contains many timeseries 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 has all the basic functionality of regular Index objects and a smorgasbord of advanced timeseries-specific methods for easy frequency processing.

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. So please be careful.

DatetimeIndex can be used like a regular index and offers all of its intelligent functionality like selection, slicing, etc.

In [56]: rng = pd.date_range(start, end, freq='BM')

In [57]: ts = pd.Series(np.random.randn(len(rng)), index=rng)

In [58]: ts.index Out[58]: DatetimeIndex(['2011-01-31', '2011-02-28', '2011-03-31', '2011-04-29', '2011-05-31', '2011-06-30', '2011-07-29', '2011-08-31', '2011-09-30', '2011-10-31', '2011-11-30', '2011-12-30'], dtype='datetime64[ns]', freq='BM')

In [59]: ts[:5].index Out[59]: DatetimeIndex(['2011-01-31', '2011-02-28', '2011-03-31', '2011-04-29', '2011-05-31'], dtype='datetime64[ns]', freq='BM')

In [60]: ts[::2].index Out[60]: DatetimeIndex(['2011-01-31', '2011-03-31', '2011-05-31', '2011-07-29', '2011-09-30', '2011-11-30'], dtype='datetime64[ns]', freq='2BM')

DatetimeIndex Partial String Indexing¶

You can pass in dates and strings that parse to dates as indexing parameters:

In [61]: ts['1/31/2011'] Out[61]: -1.2812473076599531

In [62]: ts[datetime(2011, 12, 25):] Out[62]: 2011-12-30 0.687738 Freq: BM, dtype: float64

In [63]: ts['10/31/2011':'12/31/2011'] Out[63]: 2011-10-31 0.149748 2011-11-30 -0.732339 2011-12-30 0.687738 Freq: BM, dtype: float64

To provide convenience for accessing longer time series, you can also pass in the year or year and month as strings:

In [64]: ts['2011'] Out[64]: 2011-01-31 -1.281247 2011-02-28 -0.727707 2011-03-31 -0.121306 2011-04-29 -0.097883 2011-05-31 0.695775 2011-06-30 0.341734 2011-07-29 0.959726 2011-08-31 -1.110336 2011-09-30 -0.619976 2011-10-31 0.149748 2011-11-30 -0.732339 2011-12-30 0.687738 Freq: BM, dtype: float64

In [65]: ts['2011-6'] Out[65]: 2011-06-30 0.341734 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. Here’s an example:

In [66]: dft = pd.DataFrame(randn(100000,1), ....: columns=['A'], ....: index=pd.date_range('20130101',periods=100000,freq='T')) ....:

In [67]: dft Out[67]: A 2013-01-01 00:00:00 0.176444 2013-01-01 00:01:00 0.403310 2013-01-01 00:02:00 -0.154951 2013-01-01 00:03:00 0.301624 2013-01-01 00:04:00 -2.179861 2013-01-01 00:05:00 -1.369849 2013-01-01 00:06:00 -0.954208 ... ... 2013-03-11 10:33:00 -0.293083 2013-03-11 10:34:00 -0.059881 2013-03-11 10:35:00 1.252450 2013-03-11 10:36:00 0.046611 2013-03-11 10:37:00 0.059478 2013-03-11 10:38:00 -0.286539 2013-03-11 10:39:00 0.841669

[100000 rows x 1 columns]

In [68]: dft['2013'] Out[68]: A 2013-01-01 00:00:00 0.176444 2013-01-01 00:01:00 0.403310 2013-01-01 00:02:00 -0.154951 2013-01-01 00:03:00 0.301624 2013-01-01 00:04:00 -2.179861 2013-01-01 00:05:00 -1.369849 2013-01-01 00:06:00 -0.954208 ... ... 2013-03-11 10:33:00 -0.293083 2013-03-11 10:34:00 -0.059881 2013-03-11 10:35:00 1.252450 2013-03-11 10:36:00 0.046611 2013-03-11 10:37:00 0.059478 2013-03-11 10:38:00 -0.286539 2013-03-11 10:39:00 0.841669

[100000 rows x 1 columns]

This starts on the very first time in the month, and includes the last date & time for the month

In [69]: dft['2013-1':'2013-2'] Out[69]: A 2013-01-01 00:00:00 0.176444 2013-01-01 00:01:00 0.403310 2013-01-01 00:02:00 -0.154951 2013-01-01 00:03:00 0.301624 2013-01-01 00:04:00 -2.179861 2013-01-01 00:05:00 -1.369849 2013-01-01 00:06:00 -0.954208 ... ... 2013-02-28 23:53:00 0.103114 2013-02-28 23:54:00 -1.303422 2013-02-28 23:55:00 0.451943 2013-02-28 23:56:00 0.220534 2013-02-28 23:57:00 -1.624220 2013-02-28 23:58:00 0.093915 2013-02-28 23:59:00 -1.087454

[84960 rows x 1 columns]

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

In [70]: dft['2013-1':'2013-2-28'] Out[70]: A 2013-01-01 00:00:00 0.176444 2013-01-01 00:01:00 0.403310 2013-01-01 00:02:00 -0.154951 2013-01-01 00:03:00 0.301624 2013-01-01 00:04:00 -2.179861 2013-01-01 00:05:00 -1.369849 2013-01-01 00:06:00 -0.954208 ... ... 2013-02-28 23:53:00 0.103114 2013-02-28 23:54:00 -1.303422 2013-02-28 23:55:00 0.451943 2013-02-28 23:56:00 0.220534 2013-02-28 23:57:00 -1.624220 2013-02-28 23:58:00 0.093915 2013-02-28 23:59:00 -1.087454

[84960 rows x 1 columns]

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

In [71]: dft['2013-1':'2013-2-28 00:00:00'] Out[71]: A 2013-01-01 00:00:00 0.176444 2013-01-01 00:01:00 0.403310 2013-01-01 00:02:00 -0.154951 2013-01-01 00:03:00 0.301624 2013-01-01 00:04:00 -2.179861 2013-01-01 00:05:00 -1.369849 2013-01-01 00:06:00 -0.954208 ... ... 2013-02-27 23:54:00 0.897051 2013-02-27 23:55:00 -0.309230 2013-02-27 23:56:00 1.944713 2013-02-27 23:57:00 0.369265 2013-02-27 23:58:00 0.053071 2013-02-27 23:59:00 -0.019734 2013-02-28 00:00:00 1.388189

[83521 rows x 1 columns]

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

In [72]: dft['2013-1-15':'2013-1-15 12:30:00'] Out[72]: A 2013-01-15 00:00:00 0.501288 2013-01-15 00:01:00 -0.605198 2013-01-15 00:02:00 0.215146 2013-01-15 00:03:00 0.924732 2013-01-15 00:04:00 -2.228519 2013-01-15 00:05:00 1.517331 2013-01-15 00:06:00 -1.188774 ... ... 2013-01-15 12:24:00 1.358314 2013-01-15 12:25:00 -0.737727 2013-01-15 12:26:00 1.838323 2013-01-15 12:27:00 -0.774090 2013-01-15 12:28:00 0.622261 2013-01-15 12:29:00 -0.631649 2013-01-15 12:30:00 0.193284

[751 rows x 1 columns]

Warning

The following selection will raise a KeyError; otherwise this selection methodology would be inconsistent with other selection methods in pandas (as this is not a slice, nor does it resolve to one)

dft['2013-1-15 12:30:00']

To select a single row, use .loc

In [73]: dft.loc['2013-1-15 12:30:00'] Out[73]: A 0.193284 Name: 2013-01-15 12:30:00, dtype: float64

New in version 0.18.0.

DatetimeIndex Partial String Indexing also works on DataFrames with a MultiIndex. For example:

In [74]: dft2 = pd.DataFrame(np.random.randn(20, 1), ....: columns=['A'], ....: index=pd.MultiIndex.from_product([pd.date_range('20130101', ....: periods=10, ....: freq='12H'), ....: ['a', 'b']])) ....:

In [75]: dft2 Out[75]: A 2013-01-01 00:00:00 a -0.659574 b 1.494522 2013-01-01 12:00:00 a -0.778425 b -0.253355 2013-01-02 00:00:00 a -2.816159 b -1.210929 2013-01-02 12:00:00 a 0.144669 ... ... 2013-01-04 00:00:00 b -1.624463 2013-01-04 12:00:00 a 0.056912 b 0.149867 2013-01-05 00:00:00 a -1.256173 b 2.324544 2013-01-05 12:00:00 a -1.067396 b -0.660996

[20 rows x 1 columns]

In [76]: dft2.loc['2013-01-05'] Out[76]: A 2013-01-05 00:00:00 a -1.256173 b 2.324544 2013-01-05 12:00:00 a -1.067396 b -0.660996

In [77]: idx = pd.IndexSlice

In [78]: dft2 = dft2.swaplevel(0, 1).sort_index()

In [79]: dft2.loc[idx[:, '2013-01-05'], :] Out[79]: A a 2013-01-05 00:00:00 -1.256173 2013-01-05 12:00:00 -1.067396 b 2013-01-05 00:00:00 2.324544 2013-01-05 12:00:00 -0.660996

Datetime Indexing¶

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 frequency of the index. In contrast, indexing with datetime objects is exact, because the objects have exact meaning. These also follow the semantics of including both endpoints.

These datetime objects are specific hours, minutes, and seconds even though they were not explicitly specified (they are 0).

In [80]: dft[datetime(2013, 1, 1):datetime(2013,2,28)] Out[80]: A 2013-01-01 00:00:00 0.176444 2013-01-01 00:01:00 0.403310 2013-01-01 00:02:00 -0.154951 2013-01-01 00:03:00 0.301624 2013-01-01 00:04:00 -2.179861 2013-01-01 00:05:00 -1.369849 2013-01-01 00:06:00 -0.954208 ... ... 2013-02-27 23:54:00 0.897051 2013-02-27 23:55:00 -0.309230 2013-02-27 23:56:00 1.944713 2013-02-27 23:57:00 0.369265 2013-02-27 23:58:00 0.053071 2013-02-27 23:59:00 -0.019734 2013-02-28 00:00:00 1.388189

[83521 rows x 1 columns]

With no defaults.

In [81]: dft[datetime(2013, 1, 1, 10, 12, 0):datetime(2013, 2, 28, 10, 12, 0)] Out[81]: A 2013-01-01 10:12:00 -0.246733 2013-01-01 10:13:00 -1.429225 2013-01-01 10:14:00 -1.265339 2013-01-01 10:15:00 0.710986 2013-01-01 10:16:00 -0.818200 2013-01-01 10:17:00 0.543542 2013-01-01 10🔞00 1.577713 ... ... 2013-02-28 10:06:00 0.311249 2013-02-28 10:07:00 2.366080 2013-02-28 10:08:00 -0.490372 2013-02-28 10:09:00 0.373340 2013-02-28 10:10:00 0.638442 2013-02-28 10:11:00 1.330135 2013-02-28 10:12:00 -0.945450

[83521 rows x 1 columns]

Truncating & Fancy Indexing¶

A truncate convenience function is provided that is equivalent to slicing:

In [82]: ts.truncate(before='10/31/2011', after='12/31/2011') Out[82]: 2011-10-31 0.149748 2011-11-30 -0.732339 2011-12-30 0.687738 Freq: BM, dtype: float64

Even complicated fancy indexing that breaks the DatetimeIndex’s frequency regularity will result in a DatetimeIndex (but frequency is lost):

In [83]: ts[[0, 2, 6]].index Out[83]: DatetimeIndex(['2011-01-31', '2011-03-31', '2011-07-29'], 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.

Furthermore, if you have a Series with datetimelike values, then you can access these properties via the .dt accessor, see the docs

DateOffset objects¶

In the preceding examples, we created DatetimeIndex objects at various frequencies by passing in frequency strings like ‘M’, ‘W’, and ‘BM to thefreq keyword. Under the hood, these frequency strings are being translated into an instance of pandas DateOffset, which represents a regular frequency increment. Specific offset logic like “month”, “business day”, or “one hour” is represented in its various subclasses.

The basic DateOffset takes the same arguments asdateutil.relativedelta, which works like:

In [84]: d = datetime(2008, 8, 18, 9, 0)

In [85]: d + relativedelta(months=4, days=5) Out[85]: datetime.datetime(2008, 12, 23, 9, 0)

We could have done the same thing with DateOffset:

In [86]: from pandas.tseries.offsets import *

In [87]: d + DateOffset(months=4, days=5) Out[87]: Timestamp('2008-12-23 09:00:00')

The key features of a DateOffset object are:

• it can be added / subtracted to/from a datetime object to obtain a shifted date
• it can be multiplied by an integer (positive or negative) so that the increment will be applied multiple times
• it has rollforward and rollback methods for moving a date forward or backward to the next or previous “offset date”

Subclasses of DateOffset define the apply function which dictates custom date increment logic, such as adding business days:

class BDay(DateOffset): """DateOffset increments between business days""" def apply(self, other): ...

In [88]: d - 5 * BDay() Out[88]: Timestamp('2008-08-11 09:00:00')

In [89]: d + BMonthEnd() Out[89]: Timestamp('2008-08-29 09:00:00')

The rollforward and rollback methods do exactly what you would expect:

In [90]: d Out[90]: datetime.datetime(2008, 8, 18, 9, 0)

In [91]: offset = BMonthEnd()

In [92]: offset.rollforward(d) Out[92]: Timestamp('2008-08-29 09:00:00')

In [93]: offset.rollback(d) Out[93]: Timestamp('2008-07-31 09:00:00')

It’s definitely worth exploring the pandas.tseries.offsets module and the various docstrings for the classes.

These operations (apply, rollforward and rollback) preserves time (hour, minute, etc) information by default. To reset time, use normalize=True keyword when creating the offset instance. If normalize=True, result is normalized after the function is applied.

In [94]: day = Day()

In [95]: day.apply(pd.Timestamp('2014-01-01 09:00')) Out[95]: Timestamp('2014-01-02 09:00:00')

In [96]: day = Day(normalize=True)

In [97]: day.apply(pd.Timestamp('2014-01-01 09:00')) Out[97]: Timestamp('2014-01-02 00:00:00')

In [98]: hour = Hour()

In [99]: hour.apply(pd.Timestamp('2014-01-01 22:00')) Out[99]: Timestamp('2014-01-01 23:00:00')

In [100]: hour = Hour(normalize=True)

In [101]: hour.apply(pd.Timestamp('2014-01-01 22:00')) Out[101]: Timestamp('2014-01-01 00:00:00')

In [102]: hour.apply(pd.Timestamp('2014-01-01 23:00')) Out[102]: 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 aweekday parameter which results in the generated dates always lying on a particular day of the week:

In [103]: d Out[103]: datetime.datetime(2008, 8, 18, 9, 0)

In [104]: d + Week() Out[104]: Timestamp('2008-08-25 09:00:00')

In [105]: d + Week(weekday=4) Out[105]: Timestamp('2008-08-22 09:00:00')

In [106]: (d + Week(weekday=4)).weekday() Out[106]: 4

In [107]: d - Week() Out[107]: Timestamp('2008-08-11 09:00:00')

normalize option will be effective for addition and subtraction.

In [108]: d + Week(normalize=True) Out[108]: Timestamp('2008-08-25 00:00:00')

In [109]: d - Week(normalize=True) Out[109]: Timestamp('2008-08-11 00:00:00')

Another example is parameterizing YearEnd with the specific ending month:

In [110]: d + YearEnd() Out[110]: Timestamp('2008-12-31 09:00:00')

In [111]: d + YearEnd(month=6) Out[111]: 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.

In [112]: rng = pd.date_range('2012-01-01', '2012-01-03')

In [113]: s = pd.Series(rng)

In [114]: rng Out[114]: DatetimeIndex(['2012-01-01', '2012-01-02', '2012-01-03'], dtype='datetime64[ns]', freq='D')

In [115]: rng + DateOffset(months=2) Out[115]: DatetimeIndex(['2012-03-01', '2012-03-02', '2012-03-03'], dtype='datetime64[ns]', freq='D')

In [116]: s + DateOffset(months=2) Out[116]: 0 2012-03-01 1 2012-03-02 2 2012-03-03 dtype: datetime64[ns]

In [117]: s - DateOffset(months=2) Out[117]: 0 2011-11-01 1 2011-11-02 2 2011-11-03 dtype: datetime64[ns]

If the offset class maps directly to a Timedelta (Day, Hour,Minute, Second, Micro, Milli, Nano) it can be used exactly like a Timedelta - see theTimedelta section for more examples.

In [118]: s - Day(2) Out[118]: 0 2011-12-30 1 2011-12-31 2 2012-01-01 dtype: datetime64[ns]

In [119]: td = s - pd.Series(pd.date_range('2011-12-29', '2011-12-31'))

In [120]: td Out[120]: 0 3 days 1 3 days 2 3 days dtype: timedelta64[ns]

In [121]: td + Minute(15) Out[121]: 0 3 days 00:15:00 1 3 days 00:15:00 2 3 days 00:15:00 dtype: timedelta64[ns]

Note that some offsets (such as BQuarterEnd) do not have a vectorized implementation. They can still be used but may calculate significantly slower and will raise a PerformanceWarning

In [122]: rng + BQuarterEnd() Out[122]: DatetimeIndex(['2012-03-30', '2012-03-30', '2012-03-30'], dtype='datetime64[ns]', freq=None)

Custom Business Days (Experimental)¶

The CDay or CustomBusinessDay class provides a parametricBusinessDay class which can be used to create customized business day calendars which account for local holidays and local weekend conventions.

As an interesting example, let’s look at Egypt where a Friday-Saturday weekend is observed.

In [123]: from pandas.tseries.offsets import CustomBusinessDay

In [124]: weekmask_egypt = 'Sun Mon Tue Wed Thu'

They also observe International Workers' Day so let's

add that for a couple of years

In [125]: holidays = ['2012-05-01', datetime(2013, 5, 1), np.datetime64('2014-05-01')]

In [126]: bday_egypt = CustomBusinessDay(holidays=holidays, weekmask=weekmask_egypt)

In [127]: dt = datetime(2013, 4, 30)

In [128]: dt + 2 * bday_egypt Out[128]: Timestamp('2013-05-05 00:00:00')

Let’s map to the weekday names

In [129]: dts = pd.date_range(dt, periods=5, freq=bday_egypt)

In [130]: pd.Series(dts.weekday, dts).map(pd.Series('Mon Tue Wed Thu Fri Sat Sun'.split())) Out[130]: 2013-04-30 Tue 2013-05-02 Thu 2013-05-05 Sun 2013-05-06 Mon 2013-05-07 Tue Freq: C, dtype: object

As of v0.14 holiday calendars can be used to provide the list of holidays. See theholiday calendar section for more information.

In [131]: from pandas.tseries.holiday import USFederalHolidayCalendar

In [132]: bday_us = CustomBusinessDay(calendar=USFederalHolidayCalendar())

Friday before MLK Day

In [133]: dt = datetime(2014, 1, 17)

Tuesday after MLK Day (Monday is skipped because it's a holiday)

In [134]: dt + bday_us Out[134]: Timestamp('2014-01-21 00:00:00')

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

In [135]: from pandas.tseries.offsets import CustomBusinessMonthBegin

In [136]: bmth_us = CustomBusinessMonthBegin(calendar=USFederalHolidayCalendar())

Skip new years

In [137]: dt = datetime(2013, 12, 17)

In [138]: dt + bmth_us Out[138]: Timestamp('2014-01-02 00:00:00')

Define date index with custom offset

In [139]: pd.DatetimeIndex(start='20100101',end='20120101',freq=bmth_us) Out[139]: DatetimeIndex(['2010-01-04', '2010-02-01', '2010-03-01', '2010-04-01', '2010-05-03', '2010-06-01', '2010-07-01', '2010-08-02', '2010-09-01', '2010-10-01', '2010-11-01', '2010-12-01', '2011-01-03', '2011-02-01', '2011-03-01', '2011-04-01', '2011-05-02', '2011-06-01', '2011-07-01', '2011-08-01', '2011-09-01', '2011-10-03', '2011-11-01', '2011-12-01'], dtype='datetime64[ns]', freq='CBMS')

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. 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, remaining is added to the next business day.

In [140]: bh = BusinessHour()

In [141]: bh Out[141]: <BusinessHour: BH=09:00-17:00>

2014-08-01 is Friday

In [142]: pd.Timestamp('2014-08-01 10:00').weekday() Out[142]: 4

In [143]: pd.Timestamp('2014-08-01 10:00') + bh Out[143]: Timestamp('2014-08-01 11:00:00')

Below example is the same as: pd.Timestamp('2014-08-01 09:00') + bh

In [144]: pd.Timestamp('2014-08-01 08:00') + bh Out[144]: Timestamp('2014-08-01 10:00:00')

If the results is on the end time, move to the next business day

In [145]: pd.Timestamp('2014-08-01 16:00') + bh Out[145]: Timestamp('2014-08-04 09:00:00')

Remainings are added to the next day

In [146]: pd.Timestamp('2014-08-01 16:30') + bh Out[146]: Timestamp('2014-08-04 09:30:00')

Adding 2 business hours

In [147]: pd.Timestamp('2014-08-01 10:00') + BusinessHour(2) Out[147]: Timestamp('2014-08-01 12:00:00')

Subtracting 3 business hours

In [148]: pd.Timestamp('2014-08-01 10:00') + BusinessHour(-3) Out[148]: Timestamp('2014-07-31 15:00:00')

Also, you can specify start and end time by keywords. Argument must be str which has hour:minute representation or datetime.time instance. Specifying seconds, microseconds and nanoseconds as business hour results in ValueError.

In [149]: bh = BusinessHour(start='11:00', end=time(20, 0))

In [150]: bh Out[150]: <BusinessHour: BH=11:00-20:00>

In [151]: pd.Timestamp('2014-08-01 13:00') + bh Out[151]: Timestamp('2014-08-01 14:00:00')

In [152]: pd.Timestamp('2014-08-01 09:00') + bh Out[152]: Timestamp('2014-08-01 12:00:00')

In [153]: pd.Timestamp('2014-08-01 18:00') + bh Out[153]: Timestamp('2014-08-01 19:00:00')

Passing start time later than end represents midnight business hour. In this case, business hour exceeds midnight and overlap to the next day. Valid business hours are distinguished by whether it started from valid BusinessDay.

In [154]: bh = BusinessHour(start='17:00', end='09:00')

In [155]: bh Out[155]: <BusinessHour: BH=17:00-09:00>

In [156]: pd.Timestamp('2014-08-01 17:00') + bh Out[156]: Timestamp('2014-08-01 18:00:00')

In [157]: pd.Timestamp('2014-08-01 23:00') + bh Out[157]: Timestamp('2014-08-02 00:00:00')

Although 2014-08-02 is Satuaday,

it is valid because it starts from 08-01 (Friday).

In [158]: pd.Timestamp('2014-08-02 04:00') + bh Out[158]: Timestamp('2014-08-02 05:00:00')

Although 2014-08-04 is Monday,

it is out of business hours because it starts from 08-03 (Sunday).

In [159]: pd.Timestamp('2014-08-04 04:00') + bh Out[159]: 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.rollforwardmay 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 and2014-08-04 09:00.

This adjusts a Timestamp to business hour edge

In [160]: BusinessHour().rollback(pd.Timestamp('2014-08-02 15:00')) Out[160]: Timestamp('2014-08-01 17:00:00')

In [161]: BusinessHour().rollforward(pd.Timestamp('2014-08-02 15:00')) Out[161]: Timestamp('2014-08-04 09:00:00')

It is the same as BusinessHour().apply(pd.Timestamp('2014-08-01 17:00')).

And it is the same as BusinessHour().apply(pd.Timestamp('2014-08-04 09:00'))

In [162]: BusinessHour().apply(pd.Timestamp('2014-08-02 15:00')) Out[162]: Timestamp('2014-08-04 10:00:00')

BusinessDay results (for reference)

In [163]: BusinessHour().rollforward(pd.Timestamp('2014-08-02')) Out[163]: Timestamp('2014-08-04 09:00:00')

It is the same as BusinessDay().apply(pd.Timestamp('2014-08-01'))

The result is the same as rollworward because BusinessDay never overlap.

In [164]: BusinessHour().apply(pd.Timestamp('2014-08-02')) Out[164]: Timestamp('2014-08-04 10:00:00')

BusinessHour regards Saturday and Sunday as holidays. To use arbitrary holidays, you can use CustomBusinessHour offset, see Custom Business Hour:

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.

In [165]: from pandas.tseries.holiday import USFederalHolidayCalendar

In [166]: bhour_us = CustomBusinessHour(calendar=USFederalHolidayCalendar())

Friday before MLK Day

In [167]: dt = datetime(2014, 1, 17, 15)

In [168]: dt + bhour_us Out[168]: Timestamp('2014-01-17 16:00:00')

Tuesday after MLK Day (Monday is skipped because it's a holiday)

In [169]: dt + bhour_us * 2 Out[169]: Timestamp('2014-01-21 09:00:00')

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

In [170]: bhour_mon = CustomBusinessHour(start='10:00', weekmask='Tue Wed Thu Fri')

Monday is skipped because it's a holiday, business hour starts from 10:00

In [171]: dt + bhour_mon * 2 Out[171]: 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(referred to as time rules prior to v0.8.0).

Combining Aliases¶

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

In [172]: pd.date_range(start, periods=5, freq='B') Out[172]: DatetimeIndex(['2011-01-03', '2011-01-04', '2011-01-05', '2011-01-06', '2011-01-07'], dtype='datetime64[ns]', freq='B')

In [173]: pd.date_range(start, periods=5, freq=BDay()) Out[173]: DatetimeIndex(['2011-01-03', '2011-01-04', '2011-01-05', '2011-01-06', '2011-01-07'], dtype='datetime64[ns]', freq='B')

You can combine together day and intraday offsets:

In [174]: pd.date_range(start, periods=10, freq='2h20min') Out[174]: DatetimeIndex(['2011-01-01 00:00:00', '2011-01-01 02:20:00', '2011-01-01 04:40:00', '2011-01-01 07:00:00', '2011-01-01 09:20:00', '2011-01-01 11:40:00', '2011-01-01 14:00:00', '2011-01-01 16:20:00', '2011-01-01 18:40:00', '2011-01-01 21:00:00'], dtype='datetime64[ns]', freq='140T')

In [175]: pd.date_range(start, periods=10, freq='1D10U') Out[175]: DatetimeIndex([ '2011-01-01 00:00:00', '2011-01-02 00:00:00.000010', '2011-01-03 00:00:00.000020', '2011-01-04 00:00:00.000030', '2011-01-05 00:00:00.000040', '2011-01-06 00:00:00.000050', '2011-01-07 00:00:00.000060', '2011-01-08 00:00:00.000070', '2011-01-09 00:00:00.000080', '2011-01-10 00:00:00.000090'], dtype='datetime64[ns]', freq='86400000010U')

Anchored Offsets¶

For some frequencies you can specify an anchoring suffix:

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.

In [176]: pd.Timestamp('2014-01-02') + MonthBegin(n=1) Out[176]: Timestamp('2014-02-01 00:00:00')

In [177]: pd.Timestamp('2014-01-02') + MonthEnd(n=1) Out[177]: Timestamp('2014-01-31 00:00:00')

In [178]: pd.Timestamp('2014-01-02') - MonthBegin(n=1) Out[178]: Timestamp('2014-01-01 00:00:00')

In [179]: pd.Timestamp('2014-01-02') - MonthEnd(n=1) Out[179]: Timestamp('2013-12-31 00:00:00')

In [180]: pd.Timestamp('2014-01-02') + MonthBegin(n=4) Out[180]: Timestamp('2014-05-01 00:00:00')

In [181]: pd.Timestamp('2014-01-02') - MonthBegin(n=4) Out[181]: Timestamp('2013-10-01 00:00:00')

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

In [182]: pd.Timestamp('2014-01-01') + MonthBegin(n=1) Out[182]: Timestamp('2014-02-01 00:00:00')

In [183]: pd.Timestamp('2014-01-31') + MonthEnd(n=1) Out[183]: Timestamp('2014-02-28 00:00:00')

In [184]: pd.Timestamp('2014-01-01') - MonthBegin(n=1) Out[184]: Timestamp('2013-12-01 00:00:00')

In [185]: pd.Timestamp('2014-01-31') - MonthEnd(n=1) Out[185]: Timestamp('2013-12-31 00:00:00')

In [186]: pd.Timestamp('2014-01-01') + MonthBegin(n=4) Out[186]: Timestamp('2014-05-01 00:00:00')

In [187]: pd.Timestamp('2014-01-31') - MonthBegin(n=4) Out[187]: 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.

In [188]: pd.Timestamp('2014-01-02') + MonthBegin(n=0) Out[188]: Timestamp('2014-02-01 00:00:00')

In [189]: pd.Timestamp('2014-01-02') + MonthEnd(n=0) Out[189]: Timestamp('2014-01-31 00:00:00')

In [190]: pd.Timestamp('2014-01-01') + MonthBegin(n=0) Out[190]: Timestamp('2014-01-01 00:00:00')

In [191]: pd.Timestamp('2014-01-31') + MonthEnd(n=0) Out[191]: Timestamp('2014-01-31 00:00:00')

Legacy Aliases¶

Note that prior to v0.8.0, time rules had a slightly different look. These are deprecated in v0.17.0, and removed in future version.

As you can see, legacy quarterly and annual frequencies are business quarters and business year ends. Please also note the legacy time rule for millisecondsms versus the new offset alias for month start MS. This means that offset alias parsing is case sensitive.

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. Further, start_date and end_dateclass 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:

An example of how holidays and holiday calendars are defined:

In [192]: from pandas.tseries.holiday import Holiday, USMemorialDay,
.....: AbstractHolidayCalendar, nearest_workday, MO .....:

In [193]: class ExampleCalendar(AbstractHolidayCalendar): .....: rules = [ .....: USMemorialDay, .....: Holiday('July 4th', month=7, day=4, observance=nearest_workday), .....: Holiday('Columbus Day', month=10, day=1, .....: offset=DateOffset(weekday=MO(2))), #same as 2*Week(weekday=2) .....: ] .....:

In [194]: cal = ExampleCalendar()

In [195]: cal.holidays(datetime(2012, 1, 1), datetime(2012, 12, 31)) Out[195]: DatetimeIndex(['2012-05-28', '2012-07-04', '2012-10-08'], dtype='datetime64[ns]', freq=None)

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 datetimeor Timestamp objects.

In [196]: from pandas.tseries.offsets import CDay

In [197]: pd.DatetimeIndex(start='7/1/2012', end='7/10/2012', .....: freq=CDay(calendar=cal)).to_pydatetime() .....: Out[197]: array([datetime.datetime(2012, 7, 2, 0, 0), datetime.datetime(2012, 7, 3, 0, 0), datetime.datetime(2012, 7, 5, 0, 0), datetime.datetime(2012, 7, 6, 0, 0), datetime.datetime(2012, 7, 9, 0, 0), datetime.datetime(2012, 7, 10, 0, 0)], dtype=object)

In [198]: offset = CustomBusinessDay(calendar=cal)

In [199]: datetime(2012, 5, 25) + offset Out[199]: Timestamp('2012-05-29 00:00:00')

In [200]: datetime(2012, 7, 3) + offset Out[200]: Timestamp('2012-07-05 00:00:00')

In [201]: datetime(2012, 7, 3) + 2 * offset Out[201]: Timestamp('2012-07-06 00:00:00')

In [202]: datetime(2012, 7, 6) + offset Out[202]: Timestamp('2012-07-09 00:00:00')

Ranges are defined by the start_date and end_date class attributes of AbstractHolidayCalendar. The defaults are below.

In [203]: AbstractHolidayCalendar.start_date Out[203]: Timestamp('1970-01-01 00:00:00')

In [204]: AbstractHolidayCalendar.end_date Out[204]: Timestamp('2030-12-31 00:00:00')

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

In [205]: AbstractHolidayCalendar.start_date = datetime(2012, 1, 1)

In [206]: AbstractHolidayCalendar.end_date = datetime(2012, 12, 31)

In [207]: cal.holidays() Out[207]: 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, HolidayCalendarFactoryprovides an easy interface to create calendars that are combinations of calendars or calendars with additional rules.

In [208]: from pandas.tseries.holiday import get_calendar, HolidayCalendarFactory,
.....: USLaborDay .....:

In [209]: cal = get_calendar('ExampleCalendar')

In [210]: cal.rules Out[210]: [Holiday: MemorialDay (month=5, day=31, offset=<DateOffset: kwds={'weekday': MO(-1)}>), Holiday: July 4th (month=7, day=4, observance=<function nearest_workday at 0x12f657a28>), Holiday: Columbus Day (month=10, day=1, offset=<DateOffset: kwds={'weekday': MO(+2)}>)]

In [211]: new_cal = HolidayCalendarFactory('NewExampleCalendar', cal, USLaborDay)

In [212]: new_cal.rules Out[212]: [Holiday: Labor Day (month=9, day=1, offset=<DateOffset: kwds={'weekday': MO(+1)}>), Holiday: Columbus Day (month=10, day=1, offset=<DateOffset: kwds={'weekday': MO(+2)}>), Holiday: July 4th (month=7, day=4, observance=<function nearest_workday at 0x12f657a28>), Holiday: MemorialDay (month=5, day=31, offset=<DateOffset: kwds={'weekday': MO(-1)}>)]

Resampling¶

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

In [223]: rng = pd.date_range('1/1/2012', periods=100, freq='S')

In [224]: ts = pd.Series(np.random.randint(0, 500, len(rng)), index=rng)

In [225]: ts.resample('5Min').sum() Out[225]: 2012-01-01 24390 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.

The how parameter can be a function name or numpy array function that takes an array and produces aggregated values:

In [226]: ts.resample('5Min').mean() Out[226]: 2012-01-01 243.9 Freq: 5T, dtype: float64

In [227]: ts.resample('5Min').ohlc() Out[227]: open high low close 2012-01-01 161 495 1 245

In [228]: ts.resample('5Min').max() Out[228]: 2012-01-01 495 Freq: 5T, dtype: int64

Any function available via dispatching can be given to the how parameter by name, including sum, mean, std, sem,max, min, median, first, last, ohlc.

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

In [229]: ts.resample('5Min', closed='right').mean() Out[229]: 2011-12-31 23:55:00 161.000000 2012-01-01 00:00:00 244.737374 Freq: 5T, dtype: float64

In [230]: ts.resample('5Min', closed='left').mean() Out[230]: 2012-01-01 243.9 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.

In [231]: ts.resample('5Min').mean() # by default label='right' Out[231]: 2012-01-01 243.9 Freq: 5T, dtype: float64

In [232]: ts.resample('5Min', label='left').mean() Out[232]: 2012-01-01 243.9 Freq: 5T, dtype: float64

In [233]: ts.resample('5Min', label='left', loffset='1s').mean() Out[233]: 2012-01-01 00:00:01 243.9 dtype: float64

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 time-stamp and time-span representations. By default resampleretains 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.

Up Sampling¶

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

from secondly to every 250 milliseconds

In [234]: ts[:2].resample('250L').asfreq() Out[234]: 2012-01-01 00:00:00.000 161.0 2012-01-01 00:00:00.250 NaN 2012-01-01 00:00:00.500 NaN 2012-01-01 00:00:00.750 NaN 2012-01-01 00:00:01.000 199.0 Freq: 250L, dtype: float64

In [235]: ts[:2].resample('250L').ffill() Out[235]: 2012-01-01 00:00:00.000 161 2012-01-01 00:00:00.250 161 2012-01-01 00:00:00.500 161 2012-01-01 00:00:00.750 161 2012-01-01 00:00:01.000 199 Freq: 250L, dtype: int64

In [236]: ts[:2].resample('250L').ffill(limit=2) Out[236]: 2012-01-01 00:00:00.000 161.0 2012-01-01 00:00:00.250 161.0 2012-01-01 00:00:00.500 161.0 2012-01-01 00:00:00.750 NaN 2012-01-01 00:00:01.000 199.0 Freq: 250L, dtype: float64

Sparse Resampling¶

Sparse timeseries are 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

In [237]: rng = pd.date_range('2014-1-1', periods=100, freq='D') + pd.Timedelta('1s')

In [238]: ts = pd.Series(range(100), index=rng)

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

In [239]: ts.resample('3T').sum() Out[239]: 2014-01-01 00:00:00 0.0 2014-01-01 00:03:00 NaN 2014-01-01 00:06:00 NaN 2014-01-01 00:09:00 NaN 2014-01-01 00:12:00 NaN 2014-01-01 00:15:00 NaN 2014-01-01 00🔞00 NaN ... 2014-04-09 23:42:00 NaN 2014-04-09 23:45:00 NaN 2014-04-09 23:48:00 NaN 2014-04-09 23:51:00 NaN 2014-04-09 23:54:00 NaN 2014-04-09 23:57:00 NaN 2014-04-10 00:00:00 99.0 Freq: 3T, dtype: float64

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

In [240]: from functools import partial

In [241]: from pandas.tseries.frequencies import to_offset

In [242]: def round(t, freq): .....: freq = to_offset(freq) .....: return pd.Timestamp((t.value // freq.delta.value) * freq.delta.value) .....:

In [243]: ts.groupby(partial(round, freq='3T')).sum() Out[243]: 2014-01-01 0 2014-01-02 1 2014-01-03 2 2014-01-04 3 2014-01-05 4 2014-01-06 5 2014-01-07 6 .. 2014-04-04 93 2014-04-05 94 2014-04-06 95 2014-04-07 96 2014-04-08 97 2014-04-09 98 2014-04-10 99 dtype: int64

Aggregation¶

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

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

In [244]: df = pd.DataFrame(np.random.randn(1000, 3), .....: index=pd.date_range('1/1/2012', freq='S', periods=1000), .....: columns=['A', 'B', 'C']) .....:

In [245]: r = df.resample('3T')

In [246]: r.mean() Out[246]: A B C 2012-01-01 00:00:00 -0.220339 0.034854 -0.073757 2012-01-01 00:03:00 0.037070 0.040013 0.053754 2012-01-01 00:06:00 -0.041597 -0.144562 -0.007614 2012-01-01 00:09:00 0.043127 -0.076432 -0.032570 2012-01-01 00:12:00 -0.027609 0.054618 0.056878 2012-01-01 00:15:00 -0.014181 0.043958 0.077734

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

In [247]: r['A'].mean() Out[247]: 2012-01-01 00:00:00 -0.220339 2012-01-01 00:03:00 0.037070 2012-01-01 00:06:00 -0.041597 2012-01-01 00:09:00 0.043127 2012-01-01 00:12:00 -0.027609 2012-01-01 00:15:00 -0.014181 Freq: 3T, Name: A, dtype: float64

In [248]: r[['A','B']].mean() Out[248]: A B 2012-01-01 00:00:00 -0.220339 0.034854 2012-01-01 00:03:00 0.037070 0.040013 2012-01-01 00:06:00 -0.041597 -0.144562 2012-01-01 00:09:00 0.043127 -0.076432 2012-01-01 00:12:00 -0.027609 0.054618 2012-01-01 00:15:00 -0.014181 0.043958

You can pass a list or dict of functions to do aggregation with, outputting a DataFrame:

In [249]: r['A'].agg([np.sum, np.mean, np.std]) Out[249]: sum mean std 2012-01-01 00:00:00 -39.660974 -0.220339 1.033912 2012-01-01 00:03:00 6.672559 0.037070 0.971503 2012-01-01 00:06:00 -7.487453 -0.041597 1.018418 2012-01-01 00:09:00 7.762901 0.043127 1.025842 2012-01-01 00:12:00 -4.969624 -0.027609 0.961649 2012-01-01 00:15:00 -1.418119 -0.014181 0.978847

If a dict is passed, the keys will be used to name the columns. Otherwise the function’s name (stored in the function object) will be used.

In [250]: r['A'].agg({'result1' : np.sum, .....: 'result2' : np.mean}) .....: Out[250]: result2 result1 2012-01-01 00:00:00 -0.220339 -39.660974 2012-01-01 00:03:00 0.037070 6.672559 2012-01-01 00:06:00 -0.041597 -7.487453 2012-01-01 00:09:00 0.043127 7.762901 2012-01-01 00:12:00 -0.027609 -4.969624 2012-01-01 00:15:00 -0.014181 -1.418119

On a resampled DataFrame, you can pass a list of functions to apply to each column, which produces an aggregated result with a hierarchical index:

In [251]: r.agg([np.sum, np.mean]) Out[251]: A B C
sum mean sum mean sum
2012-01-01 00:00:00 -39.660974 -0.220339 6.273786 0.034854 -13.276324
2012-01-01 00:03:00 6.672559 0.037070 7.202361 0.040013 9.675632
2012-01-01 00:06:00 -7.487453 -0.041597 -26.021155 -0.144562 -1.370600
2012-01-01 00:09:00 7.762901 0.043127 -13.757837 -0.076432 -5.862640
2012-01-01 00:12:00 -4.969624 -0.027609 9.831208 0.054618 10.237970
2012-01-01 00:15:00 -1.418119 -0.014181 4.395766 0.043958 7.773442

                     mean  

2012-01-01 00:00:00 -0.073757
2012-01-01 00:03:00 0.053754
2012-01-01 00:06:00 -0.007614
2012-01-01 00:09:00 -0.032570
2012-01-01 00:12:00 0.056878
2012-01-01 00:15:00 0.077734

By passing a dict to aggregate you can apply a different aggregation to the columns of a DataFrame:

In [252]: r.agg({'A' : np.sum, .....: 'B' : lambda x: np.std(x, ddof=1)}) .....: Out[252]: A B 2012-01-01 00:00:00 -39.660974 1.004756 2012-01-01 00:03:00 6.672559 0.963559 2012-01-01 00:06:00 -7.487453 0.950766 2012-01-01 00:09:00 7.762901 0.949182 2012-01-01 00:12:00 -4.969624 1.093736 2012-01-01 00:15:00 -1.418119 1.028869

The function names can also be strings. In order for a string to be valid it must be implemented on the Resampled object

In [253]: r.agg({'A' : 'sum', 'B' : 'std'}) Out[253]: A B 2012-01-01 00:00:00 -39.660974 1.004756 2012-01-01 00:03:00 6.672559 0.963559 2012-01-01 00:06:00 -7.487453 0.950766 2012-01-01 00:09:00 7.762901 0.949182 2012-01-01 00:12:00 -4.969624 1.093736 2012-01-01 00:15:00 -1.418119 1.028869

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

In [254]: r.agg({'A' : ['sum','std'], 'B' : ['mean','std'] }) Out[254]: A B
sum std mean std 2012-01-01 00:00:00 -39.660974 1.033912 0.034854 1.004756 2012-01-01 00:03:00 6.672559 0.971503 0.040013 0.963559 2012-01-01 00:06:00 -7.487453 1.018418 -0.144562 0.950766 2012-01-01 00:09:00 7.762901 1.025842 -0.076432 0.949182 2012-01-01 00:12:00 -4.969624 0.961649 0.054618 1.093736 2012-01-01 00:15:00 -1.418119 0.978847 0.043958 1.028869

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”.

In [255]: pd.Period('2012', freq='A-DEC') Out[255]: Period('2012', 'A-DEC')

In [256]: pd.Period('2012-1-1', freq='D') Out[256]: Period('2012-01-01', 'D')

In [257]: pd.Period('2012-1-1 19:00', freq='H') Out[257]: Period('2012-01-01 19:00', 'H')

In [258]: pd.Period('2012-1-1 19:00', freq='5H') Out[258]: Period('2012-01-01 19:00', '5H')

Adding and subtracting integers from periods shifts the period by its own frequency. Arithmetic is not allowed between Period with different freq (span).

In [259]: p = pd.Period('2012', freq='A-DEC')

In [260]: p + 1 Out[260]: Period('2013', 'A-DEC')

In [261]: p - 3 Out[261]: Period('2009', 'A-DEC')

In [262]: p = pd.Period('2012-01', freq='2M')

In [263]: p + 2 Out[263]: Period('2012-05', '2M')

In [264]: p - 1 Out[264]: Period('2011-11', '2M')

In [265]: p == pd.Period('2012-01', freq='3M')

IncompatibleFrequency Traceback (most recent call last) in () ----> 1 p == pd.Period('2012-01', freq='3M')

/Users/tom.augspurger/miniconda3/envs/docs/lib/python2.7/site-packages/pandas/pandas/src/period.pyx in pandas._period.Period.richcmp (pandas/src/period.c:12486)() 769 if other.freq != self.freq: 770 msg = _DIFFERENT_FREQ.format(self.freqstr, other.freqstr) --> 771 raise IncompatibleFrequency(msg) 772 if self.ordinal == tslib.iNaT or other.ordinal == tslib.iNaT: 773 return _nat_scalar_rules[op]

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.

In [266]: p = pd.Period('2014-07-01 09:00', freq='H')

In [267]: p + Hour(2) Out[267]: Period('2014-07-01 11:00', 'H')

In [268]: p + timedelta(minutes=120) Out[268]: Period('2014-07-01 11:00', 'H')

In [269]: p + np.timedelta64(7200, 's') Out[269]: Period('2014-07-01 11:00', 'H')

In [1]: p + Minute(5) Traceback ... ValueError: Input has different freq from Period(freq=H)

If Period has other freqs, only the same offsets can be added. Otherwise, ValueError will be raised.

In [270]: p = pd.Period('2014-07', freq='M')

In [271]: p + MonthEnd(3) Out[271]: Period('2014-10', 'M')

In [1]: p + MonthBegin(3) Traceback ... ValueError: Input has different freq from Period(freq=M)

Taking the difference of Period instances with the same frequency will return the number of frequency units between them:

In [272]: pd.Period('2012', freq='A-DEC') - pd.Period('2002', freq='A-DEC') Out[272]: 10

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:

In [273]: prng = pd.period_range('1/1/2011', '1/1/2012', freq='M')

In [274]: prng Out[274]: PeriodIndex(['2011-01', '2011-02', '2011-03', '2011-04', '2011-05', '2011-06', '2011-07', '2011-08', '2011-09', '2011-10', '2011-11', '2011-12', '2012-01'], dtype='int64', freq='M')

The PeriodIndex constructor can also be used directly:

In [275]: pd.PeriodIndex(['2011-1', '2011-2', '2011-3'], freq='M') Out[275]: PeriodIndex(['2011-01', '2011-02', '2011-03'], dtype='int64', freq='M')

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

In [276]: pd.PeriodIndex(start='2014-01', freq='3M', periods=4) Out[276]: PeriodIndex(['2014-01', '2014-04', '2014-07', '2014-10'], dtype='int64', freq='3M')

Just like DatetimeIndex, a PeriodIndex can also be used to index pandas objects:

In [277]: ps = pd.Series(np.random.randn(len(prng)), prng)

In [278]: ps Out[278]: 2011-01 -1.022670 2011-02 1.371155 2011-03 1.035277 2011-04 1.694400 2011-05 -1.659733 2011-06 0.511432 2011-07 0.433176 2011-08 -0.317955 2011-09 -0.517114 2011-10 -0.310466 2011-11 0.543957 2011-12 0.492003 2012-01 0.193420 Freq: M, dtype: float64

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

In [279]: idx = pd.period_range('2014-07-01 09:00', periods=5, freq='H')

In [280]: idx Out[280]: PeriodIndex(['2014-07-01 09:00', '2014-07-01 10:00', '2014-07-01 11:00', '2014-07-01 12:00', '2014-07-01 13:00'], dtype='int64', freq='H')

In [281]: idx + Hour(2) Out[281]: PeriodIndex(['2014-07-01 11:00', '2014-07-01 12:00', '2014-07-01 13:00', '2014-07-01 14:00', '2014-07-01 15:00'], dtype='int64', freq='H')

In [282]: idx = pd.period_range('2014-07', periods=5, freq='M')

In [283]: idx Out[283]: PeriodIndex(['2014-07', '2014-08', '2014-09', '2014-10', '2014-11'], dtype='int64', freq='M')

In [284]: idx + MonthEnd(3) Out[284]: PeriodIndex(['2014-10', '2014-11', '2014-12', '2015-01', '2015-02'], dtype='int64', 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.

In [285]: ps['2011-01'] Out[285]: -1.022669594890105

In [286]: ps[datetime(2011, 12, 25):] Out[286]: 2011-12 0.492003 2012-01 0.193420 Freq: M, dtype: float64

In [287]: ps['10/31/2011':'12/31/2011'] Out[287]: 2011-10 -0.310466 2011-11 0.543957 2011-12 0.492003 Freq: M, dtype: float64

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

In [288]: ps['2011'] Out[288]: 2011-01 -1.022670 2011-02 1.371155 2011-03 1.035277 2011-04 1.694400 2011-05 -1.659733 2011-06 0.511432 2011-07 0.433176 2011-08 -0.317955 2011-09 -0.517114 2011-10 -0.310466 2011-11 0.543957 2011-12 0.492003 Freq: M, dtype: float64

In [289]: dfp = pd.DataFrame(np.random.randn(600,1), .....: columns=['A'], .....: index=pd.period_range('2013-01-01 9:00', periods=600, freq='T')) .....:

In [290]: dfp Out[290]: A 2013-01-01 09:00 0.197720 2013-01-01 09:01 -0.284769 2013-01-01 09:02 0.061491 2013-01-01 09:03 1.630257 2013-01-01 09:04 2.042442 2013-01-01 09:05 -0.804392 2013-01-01 09:06 0.212760 ... ... 2013-01-01 18:53 0.150586 2013-01-01 18:54 -0.679569 2013-01-01 18:55 -0.910216 2013-01-01 18:56 -0.413168 2013-01-01 18:57 -0.247752 2013-01-01 18:58 1.590875 2013-01-01 18:59 -2.005294

[600 rows x 1 columns]

In [291]: dfp['2013-01-01 10H'] Out[291]: A 2013-01-01 10:00 -0.569936 2013-01-01 10:01 -1.179183 2013-01-01 10:02 -0.838602 2013-01-01 10:03 -1.727539 2013-01-01 10:04 1.334027 2013-01-01 10:05 0.417423 2013-01-01 10:06 -0.221189 ... ... 2013-01-01 10:53 -0.375925 2013-01-01 10:54 0.212750 2013-01-01 10:55 -0.592417 2013-01-01 10:56 -0.466064 2013-01-01 10:57 -1.715347 2013-01-01 10:58 -0.634913 2013-01-01 10:59 -0.809471

[60 rows x 1 columns]

As with DatetimeIndex, the endpoints will be included in the result. The example below slices data starting from 10:00 to 11:59.

In [292]: dfp['2013-01-01 10H':'2013-01-01 11H'] Out[292]: A 2013-01-01 10:00 -0.569936 2013-01-01 10:01 -1.179183 2013-01-01 10:02 -0.838602 2013-01-01 10:03 -1.727539 2013-01-01 10:04 1.334027 2013-01-01 10:05 0.417423 2013-01-01 10:06 -0.221189 ... ... 2013-01-01 11:53 0.616198 2013-01-01 11:54 2.843156 2013-01-01 11:55 0.572537 2013-01-01 11:56 1.709706 2013-01-01 11:57 -0.205490 2013-01-01 11:58 1.759719 2013-01-01 11:59 -1.181485

[120 rows x 1 columns]

Frequency Conversion and Resampling with PeriodIndex¶

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

In [293]: p = pd.Period('2011', freq='A-DEC')

In [294]: p Out[294]: 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:

In [295]: p.asfreq('M', how='start') Out[295]: Period('2011-01', 'M')

In [296]: p.asfreq('M', how='end') Out[296]: Period('2011-12', 'M')

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

In [297]: p.asfreq('M', 's') Out[297]: Period('2011-01', 'M')

In [298]: p.asfreq('M', 'e') Out[298]: Period('2011-12', 'M')

Converting to a “super-period” (e.g., annual frequency is a super-period of quarterly frequency) automatically returns the super-period that includes the input period:

In [299]: p = pd.Period('2011-12', freq='M')

In [300]: p.asfreq('A-NOV') Out[300]: 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:

In [301]: p = pd.Period('2012Q1', freq='Q-DEC')

In [302]: p.asfreq('D', 's') Out[302]: Period('2012-01-01', 'D')

In [303]: p.asfreq('D', 'e') Out[303]: Period('2012-03-31', 'D')

Q-MAR defines fiscal year end in March:

In [304]: p = pd.Period('2011Q4', freq='Q-MAR')

In [305]: p.asfreq('D', 's') Out[305]: Period('2011-01-01', 'D')

In [306]: p.asfreq('D', 'e') Out[306]: Period('2011-03-31', 'D')

Converting between Representations¶

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

In [307]: rng = pd.date_range('1/1/2012', periods=5, freq='M')

In [308]: ts = pd.Series(np.random.randn(len(rng)), index=rng)

In [309]: ts Out[309]: 2012-01-31 2.167674 2012-02-29 -1.505130 2012-03-31 1.005802 2012-04-30 0.481525 2012-05-31 -0.352151 Freq: M, dtype: float64

In [310]: ps = ts.to_period()

In [311]: ps Out[311]: 2012-01 2.167674 2012-02 -1.505130 2012-03 1.005802 2012-04 0.481525 2012-05 -0.352151 Freq: M, dtype: float64

In [312]: ps.to_timestamp() Out[312]: 2012-01-01 2.167674 2012-02-01 -1.505130 2012-03-01 1.005802 2012-04-01 0.481525 2012-05-01 -0.352151 Freq: MS, dtype: float64

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

In [313]: ps.to_timestamp('D', how='s') Out[313]: 2012-01-01 2.167674 2012-02-01 -1.505130 2012-03-01 1.005802 2012-04-01 0.481525 2012-05-01 -0.352151 Freq: MS, dtype: float64

Converting between period and timestamp enables some convenient arithmetic functions to be used. In the following example, we convert a quarterly frequency with year ending in November to 9am of the end of the month following the quarter end:

In [314]: prng = pd.period_range('1990Q1', '2000Q4', freq='Q-NOV')

In [315]: ts = pd.Series(np.random.randn(len(prng)), prng)

In [316]: ts.index = (prng.asfreq('M', 'e') + 1).asfreq('H', 's') + 9

In [317]: ts.head() Out[317]: 1990-03-01 09:00 -0.608988 1990-06-01 09:00 0.412294 1990-09-01 09:00 -0.715938 1990-12-01 09:00 1.297773 1991-03-01 09:00 -2.260765 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.

In [318]: span = pd.period_range('1215-01-01', '1381-01-01', freq='D')

In [319]: span Out[319]: PeriodIndex(['1215-01-01', '1215-01-02', '1215-01-03', '1215-01-04', '1215-01-05', '1215-01-06', '1215-01-07', '1215-01-08', '1215-01-09', '1215-01-10', ... '1380-12-23', '1380-12-24', '1380-12-25', '1380-12-26', '1380-12-27', '1380-12-28', '1380-12-29', '1380-12-30', '1380-12-31', '1381-01-01'], dtype='int64', length=60632, freq='D')

To convert from a int64 based YYYYMMDD representation.

In [320]: s = pd.Series([20121231, 20141130, 99991231])

In [321]: s Out[321]: 0 20121231 1 20141130 2 99991231 dtype: int64

In [322]: def conv(x): .....: return pd.Period(year = x // 10000, month = x//100 % 100, day = x%100, freq='D') .....:

In [323]: s.apply(conv) Out[323]: 0 2012-12-31 1 2014-11-30 2 9999-12-31 dtype: object

In [324]: s.apply(conv)[2] Out[324]: Period('9999-12-31', 'D')

These can easily be converted to a PeriodIndex

In [325]: span = pd.PeriodIndex(s.apply(conv))

In [326]: span Out[326]: PeriodIndex(['2012-12-31', '2014-11-30', '9999-12-31'], dtype='int64', freq='D')

Time Zone Handling¶

Pandas provides rich support for working with timestamps in different time zones using pytz and dateutil libraries.dateutil support is new in 0.14.1 and currently only supported for fixed offset and tzfile zones. The default library is pytz. Support for dateutil is provided for compatibility with other applications e.g. if you use dateutil in other python packages.

Working with Time Zones¶

By default, pandas objects are time zone unaware:

In [327]: rng = pd.date_range('3/6/2012 00:00', periods=15, freq='D')

In [328]: rng.tz is None Out[328]: True

To supply the time zone, you can use the tz keyword to date_range and other functions. Dateutil time zone strings are distinguished from pytztime zones by starting with dateutil/.

• In pytz you can find a list of common (and less common) time zones usingfrom pytz import common_timezones, all_timezones.
• dateutil uses the OS timezones so there isn’t a fixed list available. For common zones, the names are the same as pytz.

pytz

In [329]: rng_pytz = pd.date_range('3/6/2012 00:00', periods=10, freq='D', .....: tz='Europe/London') .....:

In [330]: rng_pytz.tz Out[330]: <DstTzInfo 'Europe/London' LMT-1 day, 23:59:00 STD>

dateutil

In [331]: rng_dateutil = pd.date_range('3/6/2012 00:00', periods=10, freq='D', .....: tz='dateutil/Europe/London') .....:

In [332]: rng_dateutil.tz Out[332]: tzfile('/usr/share/zoneinfo/Europe/London')

dateutil - utc special case

In [333]: rng_utc = pd.date_range('3/6/2012 00:00', periods=10, freq='D', .....: tz=dateutil.tz.tzutc()) .....:

In [334]: rng_utc.tz Out[334]: tzutc()

Note that the UTC timezone is a special case in dateutil and should be constructed explicitly as an instance of dateutil.tz.tzutc. You can also construct other timezones explicitly first, which gives you more control over which time zone is used:

pytz

In [335]: tz_pytz = pytz.timezone('Europe/London')

In [336]: rng_pytz = pd.date_range('3/6/2012 00:00', periods=10, freq='D', .....: tz=tz_pytz) .....:

In [337]: rng_pytz.tz == tz_pytz Out[337]: True

dateutil

In [338]: tz_dateutil = dateutil.tz.gettz('Europe/London')

In [339]: rng_dateutil = pd.date_range('3/6/2012 00:00', periods=10, freq='D', .....: tz=tz_dateutil) .....:

In [340]: rng_dateutil.tz == tz_dateutil Out[340]: True

Timestamps, like Python’s datetime.datetime object can be either time zone naive or time zone aware. Naive time series and DatetimeIndex objects can be_localized_ using tz_localize:

In [341]: ts = pd.Series(np.random.randn(len(rng)), rng)

In [342]: ts_utc = ts.tz_localize('UTC')

In [343]: ts_utc Out[343]: 2012-03-06 00:00:00+00:00 0.679135 2012-03-07 00:00:00+00:00 0.345668 2012-03-08 00:00:00+00:00 -1.143903 2012-03-09 00:00:00+00:00 0.487087 2012-03-10 00:00:00+00:00 -1.421073 2012-03-11 00:00:00+00:00 -0.327463 2012-03-12 00:00:00+00:00 0.169899 2012-03-13 00:00:00+00:00 0.867568 2012-03-14 00:00:00+00:00 -0.834122 2012-03-15 00:00:00+00:00 -1.698494 2012-03-16 00:00:00+00:00 0.974717 2012-03-17 00:00:00+00:00 0.966771 2012-03-18 00:00:00+00:00 -0.754168 2012-03-19 00:00:00+00:00 -1.434246 2012-03-20 00:00:00+00:00 0.848935 Freq: D, dtype: float64

Again, you can explicitly construct the timezone object first. You can use the tz_convert method to convert pandas objects to convert tz-aware data to another time zone:

In [344]: ts_utc.tz_convert('US/Eastern') Out[344]: 2012-03-05 19:00:00-05:00 0.679135 2012-03-06 19:00:00-05:00 0.345668 2012-03-07 19:00:00-05:00 -1.143903 2012-03-08 19:00:00-05:00 0.487087 2012-03-09 19:00:00-05:00 -1.421073 2012-03-10 19:00:00-05:00 -0.327463 2012-03-11 20:00:00-04:00 0.169899 2012-03-12 20:00:00-04:00 0.867568 2012-03-13 20:00:00-04:00 -0.834122 2012-03-14 20:00:00-04:00 -1.698494 2012-03-15 20:00:00-04:00 0.974717 2012-03-16 20:00:00-04:00 0.966771 2012-03-17 20:00:00-04:00 -0.754168 2012-03-18 20:00:00-04:00 -1.434246 2012-03-19 20:00:00-04:00 0.848935 Freq: D, dtype: float64

Warning

Be wary of conversions between libraries. For some zones pytz and dateutil have different definitions of the zone. This is more of a problem for unusual timezones than for ‘standard’ zones like US/Eastern.

Warning

Be aware that a timezone definition across versions of timezone 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.

Warning

It is incorrect to pass a timezone directly into the datetime.datetime constructor (e.g.,datetime.datetime(2011, 1, 1, tz=timezone('US/Eastern')). Instead, the datetime needs to be localized using the the localize method on the timezone.

Under the hood, all timestamps are stored in UTC. Scalar values from aDatetimeIndex with a time zone will have their fields (day, hour, minute) localized to the time zone. However, timestamps with the same UTC value are still considered to be equal even if they are in different time zones:

In [345]: rng_eastern = rng_utc.tz_convert('US/Eastern')

In [346]: rng_berlin = rng_utc.tz_convert('Europe/Berlin')

In [347]: rng_eastern[5] Out[347]: Timestamp('2012-03-10 19:00:00-0500', tz='US/Eastern', offset='D')

In [348]: rng_berlin[5] Out[348]: Timestamp('2012-03-11 01:00:00+0100', tz='Europe/Berlin', offset='D')

In [349]: rng_eastern[5] == rng_berlin[5] Out[349]: True

Like Series, DataFrame, and DatetimeIndex, Timestamp``s can be converted to other time zones using ``tz_convert:

In [350]: rng_eastern[5] Out[350]: Timestamp('2012-03-10 19:00:00-0500', tz='US/Eastern', offset='D')

In [351]: rng_berlin[5] Out[351]: Timestamp('2012-03-11 01:00:00+0100', tz='Europe/Berlin', offset='D')

In [352]: rng_eastern[5].tz_convert('Europe/Berlin') Out[352]: Timestamp('2012-03-11 01:00:00+0100', tz='Europe/Berlin')

Localization of Timestamp functions just like DatetimeIndex and Series:

In [353]: rng[5] Out[353]: Timestamp('2012-03-11 00:00:00', offset='D')

In [354]: rng[5].tz_localize('Asia/Shanghai') Out[354]: Timestamp('2012-03-11 00:00:00+0800', tz='Asia/Shanghai')

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

In [355]: eastern = ts_utc.tz_convert('US/Eastern')

In [356]: berlin = ts_utc.tz_convert('Europe/Berlin')

In [357]: result = eastern + berlin

In [358]: result Out[358]: 2012-03-06 00:00:00+00:00 1.358269 2012-03-07 00:00:00+00:00 0.691336 2012-03-08 00:00:00+00:00 -2.287805 2012-03-09 00:00:00+00:00 0.974174 2012-03-10 00:00:00+00:00 -2.842146 2012-03-11 00:00:00+00:00 -0.654926 2012-03-12 00:00:00+00:00 0.339798 2012-03-13 00:00:00+00:00 1.735136 2012-03-14 00:00:00+00:00 -1.668245 2012-03-15 00:00:00+00:00 -3.396988 2012-03-16 00:00:00+00:00 1.949435 2012-03-17 00:00:00+00:00 1.933541 2012-03-18 00:00:00+00:00 -1.508335 2012-03-19 00:00:00+00:00 -2.868493 2012-03-20 00:00:00+00:00 1.697870 Freq: D, dtype: float64

In [359]: result.index Out[359]: DatetimeIndex(['2012-03-06', '2012-03-07', '2012-03-08', '2012-03-09', '2012-03-10', '2012-03-11', '2012-03-12', '2012-03-13', '2012-03-14', '2012-03-15', '2012-03-16', '2012-03-17', '2012-03-18', '2012-03-19', '2012-03-20'], dtype='datetime64[ns, UTC]', freq='D')

To remove timezone from tz-aware DatetimeIndex, use tz_localize(None) or tz_convert(None).tz_localize(None) will remove timezone holding local time representations.tz_convert(None) will remove timezone after converting to UTC time.

In [360]: didx = pd.DatetimeIndex(start='2014-08-01 09:00', freq='H', periods=10, tz='US/Eastern')

In [361]: didx Out[361]: DatetimeIndex(['2014-08-01 09:00:00-04:00', '2014-08-01 10:00:00-04:00', '2014-08-01 11:00:00-04:00', '2014-08-01 12:00:00-04:00', '2014-08-01 13:00:00-04:00', '2014-08-01 14:00:00-04:00', '2014-08-01 15:00:00-04:00', '2014-08-01 16:00:00-04:00', '2014-08-01 17:00:00-04:00', '2014-08-01 18:00:00-04:00'], dtype='datetime64[ns, US/Eastern]', freq='H')

In [362]: didx.tz_localize(None) Out[362]: DatetimeIndex(['2014-08-01 09:00:00', '2014-08-01 10:00:00', '2014-08-01 11:00:00', '2014-08-01 12:00:00', '2014-08-01 13:00:00', '2014-08-01 14:00:00', '2014-08-01 15:00:00', '2014-08-01 16:00:00', '2014-08-01 17:00:00', '2014-08-01 18:00:00'], dtype='datetime64[ns]', freq='H')

In [363]: didx.tz_convert(None) Out[363]: DatetimeIndex(['2014-08-01 13:00:00', '2014-08-01 14:00:00', '2014-08-01 15:00:00', '2014-08-01 16:00:00', '2014-08-01 17:00:00', '2014-08-01 18:00:00', '2014-08-01 19:00:00', '2014-08-01 20:00:00', '2014-08-01 21:00:00', '2014-08-01 22:00:00'], dtype='datetime64[ns]', freq='H')

tz_convert(None) is identical with tz_convert('UTC').tz_localize(None)

In [364]: didx.tz_convert('UCT').tz_localize(None) Out[364]: DatetimeIndex(['2014-08-01 13:00:00', '2014-08-01 14:00:00', '2014-08-01 15:00:00', '2014-08-01 16:00:00', '2014-08-01 17:00:00', '2014-08-01 18:00:00', '2014-08-01 19:00:00', '2014-08-01 20:00:00', '2014-08-01 21:00:00', '2014-08-01 22:00:00'], dtype='datetime64[ns]', freq='H')

Ambiguous Times when Localizing¶

In some cases, localize cannot determine the DST and non-DST hours when there are duplicates. This often happens when reading files or database records that simply duplicate the hours. Passing ambiguous='infer' (infer_dst argument in prior releases) into tz_localize will attempt to determine the right offset. Below the top example will fail as it contains ambiguous times and the bottom will infer the right offset.

In [365]: rng_hourly = pd.DatetimeIndex(['11/06/2011 00:00', '11/06/2011 01:00', .....: '11/06/2011 01:00', '11/06/2011 02:00', .....: '11/06/2011 03:00']) .....:

This will fail as there are ambiguous times

In [2]: rng_hourly.tz_localize('US/Eastern') AmbiguousTimeError: Cannot infer dst time from Timestamp('2011-11-06 01:00:00'), try using the 'ambiguous' argument

Infer the ambiguous times

In [366]: rng_hourly_eastern = rng_hourly.tz_localize('US/Eastern', ambiguous='infer')

In [367]: rng_hourly_eastern.tolist() Out[367]: [Timestamp('2011-11-06 00:00:00-0400', tz='US/Eastern'), Timestamp('2011-11-06 01:00:00-0400', tz='US/Eastern'), Timestamp('2011-11-06 01:00:00-0500', tz='US/Eastern'), Timestamp('2011-11-06 02:00:00-0500', tz='US/Eastern'), Timestamp('2011-11-06 03:00:00-0500', tz='US/Eastern')]

In addition to ‘infer’, there are several other arguments supported. Passing an array-like of bools or 0s/1s where True represents a DST hour and False a non-DST hour, allows for distinguishing more than one DST transition (e.g., if you have multiple records in a database each with their own DST transition). Or passing ‘NaT’ will fill in transition times with not-a-time values. These methods are available in the DatetimeIndexconstructor as well as tz_localize.

In [368]: rng_hourly_dst = np.array([1, 1, 0, 0, 0])

In [369]: rng_hourly.tz_localize('US/Eastern', ambiguous=rng_hourly_dst).tolist() Out[369]: [Timestamp('2011-11-06 00:00:00-0400', tz='US/Eastern'), Timestamp('2011-11-06 01:00:00-0400', tz='US/Eastern'), Timestamp('2011-11-06 01:00:00-0500', tz='US/Eastern'), Timestamp('2011-11-06 02:00:00-0500', tz='US/Eastern'), Timestamp('2011-11-06 03:00:00-0500', tz='US/Eastern')]

In [370]: rng_hourly.tz_localize('US/Eastern', ambiguous='NaT').tolist() Out[370]: [Timestamp('2011-11-06 00:00:00-0400', tz='US/Eastern'), NaT, NaT, Timestamp('2011-11-06 02:00:00-0500', tz='US/Eastern'), Timestamp('2011-11-06 03:00:00-0500', tz='US/Eastern')]

In [371]: didx = pd.DatetimeIndex(start='2014-08-01 09:00', freq='H', periods=10, tz='US/Eastern')

In [372]: didx Out[372]: DatetimeIndex(['2014-08-01 09:00:00-04:00', '2014-08-01 10:00:00-04:00', '2014-08-01 11:00:00-04:00', '2014-08-01 12:00:00-04:00', '2014-08-01 13:00:00-04:00', '2014-08-01 14:00:00-04:00', '2014-08-01 15:00:00-04:00', '2014-08-01 16:00:00-04:00', '2014-08-01 17:00:00-04:00', '2014-08-01 18:00:00-04:00'], dtype='datetime64[ns, US/Eastern]', freq='H')

In [373]: didx.tz_localize(None) Out[373]: DatetimeIndex(['2014-08-01 09:00:00', '2014-08-01 10:00:00', '2014-08-01 11:00:00', '2014-08-01 12:00:00', '2014-08-01 13:00:00', '2014-08-01 14:00:00', '2014-08-01 15:00:00', '2014-08-01 16:00:00', '2014-08-01 17:00:00', '2014-08-01 18:00:00'], dtype='datetime64[ns]', freq='H')

In [374]: didx.tz_convert(None) Out[374]: DatetimeIndex(['2014-08-01 13:00:00', '2014-08-01 14:00:00', '2014-08-01 15:00:00', '2014-08-01 16:00:00', '2014-08-01 17:00:00', '2014-08-01 18:00:00', '2014-08-01 19:00:00', '2014-08-01 20:00:00', '2014-08-01 21:00:00', '2014-08-01 22:00:00'], dtype='datetime64[ns]', freq='H')

tz_convert(None) is identical with tz_convert('UTC').tz_localize(None)

In [375]: didx.tz_convert('UCT').tz_localize(None) Out[375]: DatetimeIndex(['2014-08-01 13:00:00', '2014-08-01 14:00:00', '2014-08-01 15:00:00', '2014-08-01 16:00:00', '2014-08-01 17:00:00', '2014-08-01 18:00:00', '2014-08-01 19:00:00', '2014-08-01 20:00:00', '2014-08-01 21:00:00', '2014-08-01 22:00:00'], dtype='datetime64[ns]', freq='H')

TZ aware Dtypes¶

New in version 0.17.0.

Series/DatetimeIndex with a timezone naive value are represented with a dtype of datetime64[ns].

In [376]: s_naive = pd.Series(pd.date_range('20130101',periods=3))

In [377]: s_naive Out[377]: 0 2013-01-01 1 2013-01-02 2 2013-01-03 dtype: datetime64[ns]

Series/DatetimeIndex with a timezone aware value are represented with a dtype of datetime64[ns, tz].

In [378]: s_aware = pd.Series(pd.date_range('20130101',periods=3,tz='US/Eastern'))

In [379]: s_aware Out[379]: 0 2013-01-01 00:00:00-05:00 1 2013-01-02 00:00:00-05:00 2 2013-01-03 00:00:00-05:00 dtype: datetime64[ns, US/Eastern]

Both of these Series can be manipulated via the .dt accessor, see here.

For example, to localize and convert a naive stamp to timezone aware.

In [380]: s_naive.dt.tz_localize('UTC').dt.tz_convert('US/Eastern') Out[380]: 0 2012-12-31 19:00:00-05:00 1 2013-01-01 19:00:00-05:00 2 2013-01-02 19:00:00-05:00 dtype: datetime64[ns, US/Eastern]

Further more you can .astype(...) timezone aware (and naive). This operation is effectively a localize AND convert on a naive stamp, and a convert on an aware stamp.

localize and convert a naive timezone

In [381]: s_naive.astype('datetime64[ns, US/Eastern]') Out[381]: 0 2012-12-31 19:00:00-05:00 1 2013-01-01 19:00:00-05:00 2 2013-01-02 19:00:00-05:00 dtype: datetime64[ns, US/Eastern]

make an aware tz naive

In [382]: s_aware.astype('datetime64[ns]') Out[382]: 0 2013-01-01 05:00:00 1 2013-01-02 05:00:00 2 2013-01-03 05:00:00 dtype: datetime64[ns]

convert to a new timezone

In [383]: s_aware.astype('datetime64[ns, CET]') Out[383]: 0 2013-01-01 06:00:00+01:00 1 2013-01-02 06:00:00+01:00 2 2013-01-03 06:00:00+01:00 dtype: datetime64[ns, CET]

Note

Using the .values accessor on a Series, returns an numpy array of the data. These values are converted to UTC, as numpy does not currently support timezones (even though it is printing in the local timezone!).

In [384]: s_naive.values Out[384]: array(['2012-12-31T18:00:00.000000000-0600', '2013-01-01T18:00:00.000000000-0600', '2013-01-02T18:00:00.000000000-0600'], dtype='datetime64[ns]')

In [385]: s_aware.values Out[385]: array(['2012-12-31T23:00:00.000000000-0600', '2013-01-01T23:00:00.000000000-0600', '2013-01-02T23:00:00.000000000-0600'], dtype='datetime64[ns]')

Further note that once converted to a numpy array these would lose the tz tenor.

In [386]: pd.Series(s_aware.values) Out[386]: 0 2013-01-01 05:00:00 1 2013-01-02 05:00:00 2 2013-01-03 05:00:00 dtype: datetime64[ns]

However, these can be easily converted

In [387]: pd.Series(s_aware.values).dt.tz_localize('UTC').dt.tz_convert('US/Eastern') Out[387]: 0 2013-01-01 00:00:00-05:00 1 2013-01-02 00:00:00-05:00 2 2013-01-03 00:00:00-05:00 dtype: datetime64[ns, US/Eastern]