Group together arbitrary date objects that are within a time range of each other

Question

I want to split the calendar into two-week intervals starting at 2008-May-5, or any arbitrary starting point.

So I start with several date objects:

Answer 1

Using a pandas DataFrame with resample works too. Given OP's data, but change "some data here" to 'abcd'.

>>> import datetime as DT
>>> raw = ("2010-08-01",
...        "2010-06-25",
...        "2010-07-01",
...        "2010-07-08")
>>> transactions = [(DT.datetime.strptime(datestring, "%Y-%m-%d"), data) for
...                 datestring, data in zip(raw,'abcd')]
[(datetime.datetime(2010, 8, 1, 0, 0), 'a'),
 (datetime.datetime(2010, 6, 25, 0, 0), 'b'),
 (datetime.datetime(2010, 7, 1, 0, 0), 'c'),
 (datetime.datetime(2010, 7, 8, 0, 0), 'd')]

Now try using pandas. First create a DataFrame, naming the columns and setting the indices to the dates.

>>> import pandas as pd
>>> df = pd.DataFrame(transactions,
...                   columns=['date','data']).set_index('date')
           data
date
2010-08-01    a
2010-06-25    b
2010-07-01    c
2010-07-08    d

Now use the Series Offset Aliases to every 2 weeks starting on Sundays and concatenate the results.

>>> fortnight = df.resample('2W-SUN').sum()
           data
date
2010-06-27    b
2010-07-11   cd
2010-07-25    0
2010-08-08    a

Now drill into the data as needed by weekstart

>>> fortnight.loc['2010-06-27']['data']
b

or index

>>> fortnight.iloc[0]['data']
b

or indices

>>> data = fortnight.iloc[:2]['data']
b
date
2010-06-27     b
2010-07-11    cd
Freq: 2W-SUN, Name: data, dtype: object
>>> data[0]
b
>>> data[1]
cd