Groupby.groups doesn't work by a groups convert from DateTimeIndex

[grgsolymosi]

i = pd.DatetimeIndex(pd.date_range('2015/01/01', periods=5), name='date')
d = pd.DataFrame({'A':[5,6,7,8,9], 'B':[1,2,3,4,5]}, index=i)

print i
print d

dg = d.groupby(level='date')

print dg.get_group('2015-01-01')
print dg.groups

Output:

DatetimeIndex(['2015-01-01', '2015-01-02', '2015-01-03', '2015-01-04',
               '2015-01-05'],
              dtype='datetime64[ns]', name=u'date', freq='D', tz=None)
            A  B
date            
2015-01-01  5  1
2015-01-02  6  2
2015-01-03  7  3
2015-01-04  8  4
2015-01-05  9  5
            A  B
date            
2015-01-01  5  1
Traceback (most recent call last):

  File "<ipython-input-2-f2b4d1146750>", line 12, in <module>
    print dg.groups

  File "/home/george/anaconda/lib/python2.7/site-packages/pandas/core/groupby.py", line 405, in groups
    return self.grouper.groups

  File "pandas/src/properties.pyx", line 34, in pandas.lib.cache_readonly.__get__ (pandas/lib.c:41917)

  File "/home/george/anaconda/lib/python2.7/site-packages/pandas/core/groupby.py", line 1352, in groups
    return self.groupings[0].groups

  File "pandas/src/properties.pyx", line 34, in pandas.lib.cache_readonly.__get__ (pandas/lib.c:41917)

  File "/home/george/anaconda/lib/python2.7/site-packages/pandas/core/groupby.py", line 2041, in groups
    return self.index.groupby(self.grouper)

  File "/home/george/anaconda/lib/python2.7/site-packages/pandas/tseries/base.py", line 60, in groupby
    return _algos.groupby_object(objs, f)

TypeError: Argument 'labels' has incorrect type (expected numpy.ndarray, got DatetimeIndex)

[jreback]

hmm, does seem a bit buggy

[jreback]

@temmplar reason you closed this?

[grgsolymosi]

@jreback sorry I guess, I'm not familiar with the system and I thought if you answered I have to close it.

[jreback]

nope
would be closed by a pull request to fix which are welcome!

[jzwinck]

I tried this in tseries/base.py:

-        return _algos.groupby_object(objs, f)
+        return _algos.groupby_object(objs, np.asarray(f))

And it sort of works, but the results are ugly, especially if a timezone is involved. An index like this:

DatetimeIndex(['2016-06-28 05:30:00-05:00', '2016-06-28 05:31:00-05:00'], dtype='datetime64[ns, America/Chicago]')

Produces naive UTC results:

{numpy.datetime64('2016-06-28T10:30:00.000000000'): [Timestamp('2016-06-28 05:30:00-0500', ...

@jreback Do you have any idea how to fix this simply and properly?

[akapocsi]

I am getting this error too.