Groupby aggregation of date/datetime columns returns datetime64 rather than numeric type

[michaelbilow]

import pandas as pd
import datetime
u = [datetime.datetime(2015, x, 1) for x in range(12)]
v = list('aaabbbbbbccd')
df = pd.DataFrame('X':v, 'Y':u)
df.groupby('X')['Y'].agg(len)
## Returns the following:
X
a   1970-01-01 00:00:00.000000003
b   1970-01-01 00:00:00.000000006
c   1970-01-01 00:00:00.000000002
d   1970-01-01 00:00:00.000000001

You can fix the problem by casting the dates to strings before groupby/agg, but if you try to cast the returned datetimes to ints, errors go off in some versions of pandas. Either way, aggregating by length should always return an int. Also, this may be similar to #11442, which was just posted.

[jreback]

this is a dupe of #10972

thanks for the report