API/BUG: SparseSeries.shape ignores fill_value

[sinhrks]

SparseSeries.shape seems to return incorrect result ignoring shapes filled by fill_value.

s = pd.SparseSeries([0, 0, 1, 0], fill_value=0)
s
#0    0
#1    0
#2    1
#3    0
# dtype: int64
# BlockIndex
# Block locations: array([2], dtype=int32)
# Block lengths: array([1], dtype=int32)

len(s)
#4

# must be (4, )?
s.shape
# (1,)

# OK
s._data.shape
# (4,)

[jreback]

This refers to the shape of the sparse values, not the shape of the series itself.

In [19]: s._data.blocks[0].sp_values
Out[19]: array([1])

In [20]: s._data.blocks[0].sp_values.shape
Out[20]: (1,)

[sinhrks]

I see. This is a spec and not be fixed?

May be a option to fix docstring to say underlying means sparse values.

help(pd.SparseSeries.shape)
# return a tuple of the shape of the underlying data

[jreback]

I don't remember where exactly this is used and whether its just user-facing. If its user facing only, then it should be consistent and return the shape of a densified frame, and we should have a sparse_shape or something. I am not sure what this really should do. If you want to experiment and report back would be gr8.