Add "day of month" (e.g., 15th of month) frequencies, easier extension?

[wesm]

http://stackoverflow.com/questions/13445174/date-ranges-in-pandas

[poswald]

Do you know if there has been any movement on this at all since it was identified? The data I am looking at has bursts of activity monthly that aligns on the last day and 1st days of the month so month-over-month analysis would benefit considerably from being able to offset by X days to get mid-month reports. Thanks!

[jreback]

@poswald this is not a solution, but a straightforward way to create the day-of-month index (its still not the right frequency), but still may work for what you need

In [27]: from datetime import timedelta

In [28]: Index(Series(date_range('20130115',freq='MS',periods=10).tolist()) + timedelta(days=14))
Out[28]: 
<class 'pandas.tseries.index.DatetimeIndex'>
[2013-02-15 00:00:00, ..., 2013-11-15 00:00:00]
Length: 10, Freq: None, Timezone: None

In [29]: Index(Series(date_range('20130115',freq='MS',periods=10).tolist()) + timedelta(days=14)).tolist()
Out[29]: 
[Timestamp('2013-02-15 00:00:00', tz=None),
 Timestamp('2013-03-15 00:00:00', tz=None),
 Timestamp('2013-04-15 00:00:00', tz=None),
 Timestamp('2013-05-15 00:00:00', tz=None),
 Timestamp('2013-06-15 00:00:00', tz=None),
 Timestamp('2013-07-15 00:00:00', tz=None),
 Timestamp('2013-08-15 00:00:00', tz=None),
 Timestamp('2013-09-15 00:00:00', tz=None),
 Timestamp('2013-10-15 00:00:00', tz=None),
 Timestamp('2013-11-15 00:00:00', tz=None)]

[petrounias]

Also, it appears that you cannot even shift a PeriodIndex after constructing it, but what's puzzling is that the shift method's documentation states a 'freq' parameter which is missing from the method's signature. Specifically:

def shift(self, n):
    """
    Specialized shift which produces an PeriodIndex

    Parameters
    ----------
    n : int
        Periods to shift by
    freq : freq string

    Returns
    -------
    shifted : PeriodIndex
    """

[petehuang]

Related: #1573

Have there been any developments related to "day of month" frequencies?

[jreback]

We have an implementation of WeekOfMonth, see here, so this would be pretty easy. Not sure how useful it is, so we'll leave this open.

On another note, we are missing documentation of WOM and LWOM (last-week-of-month) in the main docs (at least a reference http://pandas.pydata.org/pandas-docs/stable/timeseries.html?highlight=weekofmonth#offset-aliases would be good (and a small example on usage) if you'd like to do a PR.