SeriesGroupBy.count() errors when original series is MultiIndex with Datetime as level 0

[dalejung]

from pandas import *

dr = date_range(start='1/1/2012', freq='5min', periods=10)

# BAD Example, datetimes first
labels = zip(dr, range(10))
s = Series(np.arange(10), index=MultiIndex.from_tuples(labels, names=['a','b']))
grouped = s.groupby(lambda x: x[1] % 2 == 0)

grouped.count()
# DateParseError: Could not parse COUNT

# Works fine
labels = zip(range(10), dr)
s = Series(np.arange(10), index=MultiIndex.from_tuples(labels, names=['a','b']))
grouped = s.groupby(lambda x: x[0] % 2 == 0)
grouped.count()
# False    5
# True     5

Seems to have happened within the last few days, maybe even just today.

[wesm]

Thanks. Fixed the underlying issue