Useful math stats functions for collections of datetime objects.
Ever had a list or set of datetimes and needed to perform some basic statistical functions to find the median, mean, minimum, or maximum? datetimestats provides some basic utilities to do this without the overhead of converting to a numeric data type (e.g., unix timestamps at millisecond-level) or pandas data frames (in order to employ pandas time series functionality).
datetimestats takes any iterable of datetime objects (including panda Series). It supports naive and non-naive datetimes, including iterables of objects with different Olson timezones. However, it does not support iterables that contain both naive and non-naive datetime objects.
Currently, datetimestats is supported for use with Python 2.7.
Install using pip with:
pip install datetimestatsOr download a wheel or source archive from PyPI.
datatimestats currently supports obtaining the the mean, median, min, or max of an Iterable of datetime objects. The functions are designed to minimize time complexity where possible.
Mean is calculated as the the arithmetic mean across datetime objects with micro-second precision.
When given a list of naive datetimes, mean returns a naive datetime object
>>> import datetime as dt
>>> naive_1 = dt.datetime(2015, 9, 10, 12, 30, 0)
>>> naive_2 = dt.datetime(2015, 9, 10, 12, 0, 0)
>>> from datetimestats import mean
>>> mean([naive_1, naive_2])
datetime.datetime(2015, 9, 10, 12, 15)When given a list of non-naive datetimes, it returns the mean as a
datetime object in UTC time:
>>> import datetime as dt
>>> import pytz
>>> nyc_noon = dt.datetime(2014, 1, 1, 12, 0, 0, tzinfo=pytz.timezone('America/New_York'))
>>> print nyc_noon # Curiosity of pytz, it does not convert to whole time zones
2014-01-01 12:00:00-04:56
>>> london_noon = dt.datetime(2014, 1, 1, 12, 0, 0, tzinfo=pytz.timezone('Europe/London'))
>>> singapore_noon = dt.datetime(2014, 1, 1, 12, 0, 0, tzinfo=pytz.timezone('Asia/Singapore'))
>>> from datetimestats import mean
>>> mean([nyc_noon, london_noon, singapore_noon])
datetime.datetime(2014, 1, 1, 11, 20, 40, tzinfo=<UTC>)If an odd number of objects are provided, median is calculated is the inner-most, sorted value. If an even number is provided, median is the arithmetic-mean of the two inner-most, sorted values.
Odd number of values:
>>> import datetime as dt
>>> import pytz
>>> london_noon = dt.datetime(2014, 1, 1, 12, 0, 0, tzinfo=pytz.timezone('Europe/London'))
>>> nyc_noon = dt.datetime(2014, 1, 1, 12, 0, 0, tzinfo=pytz.timezone('America/New_York'))
>>> singapore_noon = dt.datetime(2014, 1, 1, 12, 0, 0, tzinfo=pytz.timezone('Asia/Singapore'))
>>> from datetimestats import median
>>> median([nyc_noon, london_noon, singapore_noon])
datetime.datetime(2014, 1, 1, 12, 0, tzinfo=<DstTzInfo 'Europe/London' LMT-1 day, 23:59:00 STD>)
>>> median([nyc_noon, london_noon, singapore_noon]) == london_noon # Noon in London does fall between Singapore and NYC
TrueEven number of values:
>>> import datetime as dt
>>> naive_1 = dt.datetime(2015, 9, 10, 12, 0, 0)
>>> naive_2 = dt.datetime(2015, 9, 10, 14, 0, 0)
>>> naive_3 = dt.datetime(2015, 9, 10, 13, 0, 0)
>>> naive_4 = dt.datetime(2015, 9, 10, 15, 0, 0)
>>> from datetimestats import median
>>> median([naive_1, naive_2, naive_3, naive_4])
datetime.datetime(2015, 9, 10, 14, 30)The min datetime is the earliest datetime. Conversely, the max is the latest. This is most interesting when calculating across multiple timezones:
>>> import datetime as dt
>>> import pytz
>>> london_noon = dt.datetime(2014, 1, 1, 12, 0, 0, tzinfo=pytz.timezone('Europe/London'))
>>> nyc_noon = dt.datetime(2014, 1, 1, 12, 0, 0, tzinfo=pytz.timezone('America/New_York'))
>>> singapore_noon = dt.datetime(2014, 1, 1, 12, 0, 0, tzinfo=pytz.timezone('Asia/Singapore'))
>>> from datetimestats import min, max
>>> min([nyc_noon, london_noon, singapore_noon])
datetime.datetime(2014, 1, 1, 12, 0, tzinfo=<DstTzInfo 'Asia/Singapore' LMT+6:55:00 STD>)
>>> min([nyc_noon, london_noon, singapore_noon]) == singapore # It is noon in Singapore EARLIEST
True
>>> max([nyc_noon, london_noon, singapore_noon])
datetime.datetime(2014, 1, 1, 12, 0, tzinfo=<DstTzInfo 'America/New_York' LMT-1 day, 19:04:00 STD>)
>>> max([nyc_noon, london_noon, singapore_noon]) == nyc_noon # It is noon in NYC LATEST
TrueSets, tuples, numpy Arrays and pandas Series are also supported:
>>> import datetime as dt
>>> naive_1 = dt.datetime(2015, 9, 10, 12, 0, 0)
>>> naive_2 = dt.datetime(2015, 9, 10, 14, 0, 0)
>>> from datetimestats import median
>>> median([naive_1, naive_2])
datetime.datetime(2015, 9, 10, 13, 0)
>>> median((naive_1, naive_2))
datetime.datetime(2015, 9, 10, 13, 0)
>>> import numpy as np
>>> median(np.asarray([naive_1, naive_2]))
datetime.datetime(2015, 9, 10, 13, 0)
>>> import pandas as pd
>>> median(pd.Series([naive_1, naive_2]))
datetime.datetime(2015, 9, 10, 13, 0)