API: sparse.cumsum should default to stats axis

[jreback]

xref #12810

In [1]: sp = pd.SparseDataFrame([1, 2, 3])

In [2]: np.cumsum(sp)
ValueError: No axis named None for object type <class 'pandas.sparse.frame.SparseDataFrame'>

In [3]: np.cumsum(sp, axis=0)
Out[3]: 
     0
0  1.0
1  3.0
2  6.0

In [4]: np.cumsum(sp, axis=1)
Out[4]: 
     0
0  1.0
1  2.0
2  3.0

[jreback]

cc @gfyoung merging #12810

but I think this should default. Need to systematically check all of the cum functions.
(axis = 0). dense already does this. (yes the axis is passed as None but its set if its None.

cc @sinhrks

[jreback]

e.g. [6] works

In [4]: np.cumsum(sp, axis=0)
Out[4]: 
     0
0  1.0
1  3.0
2  6.0

In [5]: np.cumsum(sp.to_dense(), axis=0)
Out[5]: 
     0
0  1.0
1  3.0
2  6.0

In [6]: np.cumsum(sp.to_dense())
Out[6]: 
     0
0  1.0
1  3.0
2  6.0

[gfyoung]

@jreback : I'm not sure I fully understand. How do you want to handle the case when axis=None (when you say "stats" axis)? In numpy it's easy because you just flatten the array, but that doesn't make any sense with a DataFrame.

[jreback]

see the dense impl

it needs to be 0 (which is the stats axis)

[gfyoung]

Ah, I see. Just a simple:

if axis is None:
   axis = self._stat_axis_number

before making the self.apply function call

[jreback]

yep but merging soon
do after and need tests

[gfyoung]

Yes, indeed, certainly!