DOC: hour/hours (and other plurals) based on dateutil.relativedelta (original) (raw)

discovered in #5292

This seems odd to me, documented anywhere?

In [2]: t = pd.Timestamp('2014-06-10 8am')

In [3]: t + pd.DateOffset(hour=9)
Out[3]: Timestamp('2014-06-10 09:00:00', tz=None)

In [4]: t + pd.DateOffset(hours=9)
Out[4]: Timestamp('2014-06-10 17:00:00', tz=None)

note that

t + pd.offsets.Hour(9) == t + timedelta(hours=9)

so this definitly seems suspect

(also problematic for second/seconds and minute/minutes)

this is only an issue in DateOffset constructions.