DataFrame.to_sparse fill_value keyword doesn't seem to fill NA

[minggli]

>>> from pandas import *
>>> from numpy.random import randn
>>>
>>> df = DataFrame(randn(20, 4))
>>> df = df.applymap(lambda x: np.nan if x > 0 else x)
>>> sparse_df = df.to_sparse(fill_value=0.35)
>>> sparse_df
           0         1         2         3
0  -1.101095       NaN -1.512058 -0.283055
1  -0.460331       NaN -0.995369       NaN
2        NaN -0.835043 -1.367358 -0.950400
3  -0.704320 -0.746556 -0.389821       NaN
4        NaN -0.573624       NaN -1.033302
5        NaN       NaN -1.684434 -1.805903
6        NaN       NaN       NaN       NaN
7  -0.129019 -2.279948 -0.229312 -0.725433
8  -1.264768 -0.050307       NaN -0.453808
9  -0.976076 -0.127872 -0.794526       NaN
10 -0.126079 -0.774120 -0.424461 -0.061588
11 -0.474522       NaN -0.437307 -0.727405
12       NaN -0.883299       NaN -1.435336
13 -1.588188       NaN -1.740614 -0.645563
14 -1.218019       NaN       NaN -0.341365
15 -0.522648 -1.377164       NaN       NaN
16       NaN -1.193746 -1.295383       NaN
17 -1.445435 -0.034316 -0.879550 -1.570954
18       NaN -0.060612       NaN       NaN
19       NaN       NaN -0.009722       NaN

Problem description

I've looked at issues log and don't seem to find an issue regarding this. It appears that DataFrame.to_sparse keyword fill_value which subsequently injected into SparseDataFrame doesn't seem to do anything.

Is it actually a bug or the documentation isn't accurate enough?

Expected Output

           0         1         2         3
0  -1.101095  0.350000 -1.512058 -0.283055
1  -0.460331  0.350000 -0.995369  0.350000
2   0.350000 -0.835043 -1.367358 -0.950400
3  -0.704320 -0.746556 -0.389821  0.350000
4   0.350000 -0.573624  0.350000 -1.033302
5   0.350000  0.350000 -1.684434 -1.805903
6   0.350000  0.350000  0.350000  0.350000
7  -0.129019 -2.279948 -0.229312 -0.725433
8  -1.264768 -0.050307  0.350000 -0.453808
9  -0.976076 -0.127872 -0.794526  0.350000
10 -0.126079 -0.774120 -0.424461 -0.061588
11 -0.474522  0.350000 -0.437307 -0.727405
12  0.350000 -0.883299  0.350000 -1.435336
13 -1.588188  0.350000 -1.740614 -0.645563
14 -1.218019  0.350000  0.350000 -0.341365
15 -0.522648 -1.377164  0.350000  0.350000
16  0.350000 -1.193746 -1.295383  0.350000
17 -1.445435 -0.034316 -0.879550 -1.570954
18  0.350000 -0.060612  0.350000  0.350000
19  0.350000  0.350000 -0.009722  0.350000

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.6.1.final.0
python-bits: 64
OS: Darwin
OS-release: 17.4.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: None
LANG: en_GB.UTF-8
LOCALE: en_GB.UTF-8

pandas: 0.22.0
pytest: None
pip: 9.0.1
setuptools: 36.3.0
Cython: None
numpy: 1.14.1
scipy: 0.19.1
pyarrow: None
xarray: None
IPython: 5.3.0
sphinx: None
patsy: 0.4.1
dateutil: 2.6.1
pytz: 2018.3
blosc: None
bottleneck: None
tables: None
numexpr: None
feather: None
matplotlib: 2.0.0
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml: None
bs4: 4.5.3
html5lib: 0.9999999
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: 2.9.6
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: None

[TomAugspurger]

to_sparse is being removed, but fill_value=NaN means that NaN is the "sparse value" not stored in memory. It doesn't mean that it'll be filled.