BUG: groupby regression

[jreback]

in #2893

we changed the groupby behavior to return a series if a simple index is passed in and only have 1 group (rather than return a mi)

I hit this regression as I doing several groupby's on different input, all of which yield multiple groups, except for 1, so I result with a bunch of frames and 1 series

in 0.10.1 this would yield all frames.

so the question is should the automatic magic of 2893 always happen?
e.g. I guess that should create an option to not always reduce the input dims

so my example, shoudl result1 == DataFrame ?

In [6]: df1 = DataFrame(dict(A = range(4), B = 0))

In [7]: def func(dataf):
   ...:     return Series({ dataf.name : 1})
   ...: 

In [8]: result1 = df1.groupby("B").apply(func)

In [9]: result2 = df1.groupby("A").apply(func)

In [10]: result1
Out[10]: 
B   
0  0    1
Name: 0, dtype: int64

In [11]: result2
Out[11]: 
A   
0  0    1
1  1    1
2  2    1
3  3    1
dtype: int64

[jreback]

closed by #3599