MultiIndex groupby bugs in 0.7.3, 0.8.0b1 and 0.8.0dev

[ruidc]

import pandas
l = [['count', 'values'], ['to filter', '']]
midx = pandas.MultiIndex.from_tuples(l)
df = pandas.DataFrame([[1L, 'A']], columns=midx) #one line
print(df.groupby('to filter').groups)
#Out: {'to filter': [0L]} #was expecting 'A': [0L]
print(df.groupby([('to filter', '')]).groups)
#Out: {'to filter': [0L]} #was expecting same as above
df = pandas.DataFrame([[1L, 'A'], [2L, 'B']], columns=midx) #two lines, different group
print(df.groupby('to filter').groups)
#Out: {'A': [0L], 'B': [1L]} #fine
print(df.groupby([('to filter', '')]).groups)
#Out: {'': [1L], 'to filter': [0L]} #was expecting same as above
df = pandas.DataFrame([[1L, 'A'], [2L, 'A']], columns=midx) #two lines, same group
print(df.groupby('to filter').groups)
#Out: {'A': [0L, 1L]} #fine
print(df.groupby([('to filter', '')]).groups)
#Out: {'': [1L], 'to filter': [0L]} #was expecting same as above

[ruidc]

see also ML post: https://groups.google.com/forum/?fromgroups#!topic/pystatsmodels/MwvSs9rU_i4

[wesm]

Fixed these bugs, a few little hacks but stuff I can live with

[ruidc]

Thanks!
perhaps I'm just nitpicking, but in the tests, wouldn't it be better to test for the explicit results rather than just a comparison match between a one-level and two-level groupby in case the the results were the same but incorrect?