ENH: sparse astype now supports int64 and bool

[sinhrks]

• split from #13849
• tests added / passed
• passes git diff upstream/master | flake8 --diff
• whatsnew entry

SparseXXX now support astype other than float64 dtype. It raises ValueError when its sp_values or fill_value cannot be converted to specified dtype.

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

pd.SparseSeries([1, np.nan, 2, np.nan]).astype(np.int64)
# ValueError: unable to coerce current fill_value nan to int64 dtype

Also, fill_value setter now checks input value dtype and raises ValueError if value cannot be stored to current dtype.

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

s.fill_value = np.nan
# ValueError: unable to set fill_value nan to int64 dtype

[codecov-io]

Current coverage is 85.29% (diff: 100%)

Merging #13900 into master will increase coverage by <.01%

@@             master     #13900   diff @@
==========================================
  Files           139        139          
  Lines         50057      50077    +20   
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
+ Hits          42692      42712    +20   
  Misses         7365       7365          
  Partials          0          0          

Powered by Codecov. Last update 45d54d0...1c669ad

[jreback]

lgtm, is this dependent on anything or standalone?

[sinhrks]

thx for the review. it's independent.

[jreback]

thanks!