MonthOffset() gives increase of 1 day instead of 1 month

[jorisvandenbossche]

From #7375 (comment):

In [67]: from pandas.tseries import offsets

In [69]: d = pd.Timestamp('2014-01-01 09:00:00')

In [70]: d + offsets.MonthOffset()
Out[70]: Timestamp('2014-01-02 09:00:00')

In [71]: offsets.MonthOffset().apply(d)
Out[71]: Timestamp('2014-01-02 09:00:00')

In [72]: d + offsets.Day()
Out[72]: Timestamp('2014-01-02 09:00:00')

[jreback]

as noted in #7375. This (and YearOffset), need an apply implementation as they are using the DateOffset one (which is day-based). The constructor is ok.

doc-string should be expanded as well.

[hayd]

Also noticed this with #7599, confusing.

I need to fix up #6318 in 2.6 (grumble grumble), I don't recall if that addressed this... ? (maybe not...)

[jreback]

you need to specify month=1 and it will work
(it's interpreting the first arg as the day keyword)

[hayd]

Hmmm, + seems to set the month:

In [11]: pd.Timestamp.now()
Out[11]: Timestamp('2014-08-09 13:59:37.263338')

In [12]: pd.Timestamp.now() + pd.offsets.MonthOffset(month=1)
Out[12]: Timestamp('2014-01-09 13:59:39.467444')

In [13]: pd.Timestamp.now() + pd.offsets.MonthOffset(month=2)
Out[13]: Timestamp('2014-02-09 13:59:41.586575')