Sparse data frame doesn't groupby.mean() correctly

[langmore]

>>>import pandas as pd
>>> pd.__version__
 '0.12.0-660-gec77315'

>>> import numpy as np
>>> np.version.version
'1.7.1'

>>> df = pd.DataFrame({'a': [0, 1, 0, 0], 'b': [0, 1, 0, 0]})
>>> sdf = df.to_sparse(fill_value=0)
>>> df.groupby('a').mean() 
    b
a   
0  0
1  1

>>> sdf.groupby('a').mean() 
    b
a   
1  0

I'm not surpised that the mean for group a == 0 was not returned. It is surprising that the result for group a == 1 was incorrect.

[OmerJog]

Are there any planes to fix this any time soon?

[TomAugspurger]

@OmerJog this works with a DataFrame with sparse values

In [90]: df = pd.DataFrame({'a': [0, 1, 0, 0], 'b': [0, 1, 0, 0]})

In [91]: df.apply(pd.SparseArray).groupby('a').mean()
Out[91]:
   b
a
0  0
1  1