BUG: Select an all-zero sparse series by indexer results in NaN

[HYChou0515]

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• (optional) I have confirmed this bug exists on the master branch of pandas.

---

Note: Please read this guide detailing how to provide the necessary information for us to reproduce your bug.

Code Sample, a copy-pastable example

import pandas as pd
import scipy as sp
import scipy.sparse

spmatrix = sp.sparse.csr_matrix((2, 2))
spmatrix[0, 0] = 1

df = pd.DataFrame.sparse.from_spmatrix(spmatrix)

# work as expected, cool
df.loc[1]
# 0    0.0
# 1    0.0
# Name: 0, dtype: Sparse[float64, 0.0]

# Indexing by index list. I expect 1,0, 0.0, not cool
df.loc[[1]]
#      0   1
# 1  0.0 NaN

# whenever indexing an all-zero series, it returns Nan.
df.loc[[1]].loc[[1]] 
#     0   1
# 1 NaN NaN

# dtype is good
df.loc[[1]].dtypes
# 0    Sparse[float64, 0.0]
# 1    Sparse[float64, 0.0]
# dtype: object

# indexing by bool list also not cool
df.loc[[True, True]]
#      0   1
# 0  1.0 NaN
# 1  0.0 NaN

# Slicing is cool
df.loc[1:2]
#      0    1
# 1  0.0  0.0

Problem description

Index slice an all-zero sparse series by indexer (whether by bool array or index array) results in NaN. It should be the fill_value of the series.

Expected Output

Output of pd.show_versions()

INSTALLED VERSIONS
------------------
commit           : None
python           : 3.8.1.final.0
python-bits      : 64
OS               : Linux
OS-release       : 4.15.0-101-generic
machine          : x86_64
processor        : x86_64
byteorder        : little
LC_ALL           : en_US.UTF-8
LANG             : en_US.UTF-8
LOCALE           : en_US.UTF-8

pandas           : 1.0.3
numpy            : 1.18.1
pytz             : 2019.3
dateutil         : 2.8.1
pip              : 20.0.2
setuptools       : 45.2.0
Cython           : None
pytest           : 5.3.5
hypothesis       : None
sphinx           : None
blosc            : None
feather          : None
xlsxwriter       : None
lxml.etree       : None
html5lib         : None
pymysql          : None
psycopg2         : None
jinja2           : 2.11.1
IPython          : 7.12.0
pandas_datareader: None
bs4              : None
bottleneck       : None
fastparquet      : None
gcsfs            : None
lxml.etree       : None
matplotlib       : 3.2.0
numexpr          : None
odfpy            : None
openpyxl         : None
pandas_gbq       : None
pyarrow          : None
pytables         : None
pytest           : 5.3.5
pyxlsb           : None
s3fs             : None
scipy            : 1.4.1
sqlalchemy       : None
tables           : None
tabulate         : None
xarray           : None
xlrd             : None
xlwt             : None
xlsxwriter       : None
numba            : None

[HYChou0515]

This issue is fixed in the master branch of pandas. But it indexing sparse series still has some issue ( #34540 )

[jreback]

@HYChou0515 can u check if we have all of the cases in tests? if not would love to have a PR to add missing ones

[jorisvandenbossche]

Also on master, the dtypes are still wrong:

In [79]: df.dtypes  
Out[79]: 
0    Sparse[float64, 0]
1    Sparse[float64, 0]
dtype: object

In [80]: df.loc[[1]] 
Out[80]: 
     0  1
1  0.0  0

In [81]: df.loc[[1]].dtypes 
Out[81]: 
0    Sparse[float64, 0]
1      Sparse[int64, 0]
dtype: object

The Sparse[float] column gets turned into a Sparse[int] during the indexing operation.