`fillna` using columns of dtype `category` also fills non-NaN values

[lcvriend]

Code Sample, a copy-pastable example if possible

import pandas as pd
import numpy as np

dct = {
    'A': ['a', 'b', 'c', 'b'], 
    'B': ['d', 'e', np.nan, np.nan]
}
df = pd.DataFrame.from_dict(dct).astype('category')
df['C'] = df['B']
df['C'].cat.add_categories(df['A'].cat.categories, inplace=True)
df['C'] = df['C'].fillna(df['A'])

output

	A	B	C
0	a	d	a
1	b	e	b
2	c	NaN	c
3	b	NaN	b

Problem description

I have two columns, A and B, of dtype category. Column B contains NaN values.
Applying fillna to B using A (after adding categories in A to categories in B), results in ALL values of B being overwritten with values of A. The issue is that fillna also fills non-NaN values.

Expected Output

Non-NaN values should not be overwritten:

	A	B	C
0	a	d	d
1	b	e	e
2	c	NaN	c
3	b	NaN	b

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.7.3.final.0
python-bits: 64
OS: Windows
OS-release: 10
machine: AMD64
processor: Intel64 Family 6 Model 94 Stepping 3, GenuineIntel
byteorder: little
LC_ALL: None
LANG: None
LOCALE: None.None

pandas: 0.24.2
pytest: 4.3.1
pip: 19.0.3
setuptools: 40.8.0
Cython: 0.29.6
numpy: 1.16.2
scipy: 1.2.1
pyarrow: None
xarray: None
IPython: 7.4.0
sphinx: 1.8.5
patsy: 0.5.1
dateutil: 2.8.0
pytz: 2018.9
blosc: None
bottleneck: 1.2.1
tables: 3.5.1
numexpr: 2.6.9
feather: None
matplotlib: 3.0.3
openpyxl: 2.6.1
xlrd: 1.2.0
xlwt: 1.3.0
xlsxwriter: 1.1.5
lxml.etree: 4.3.2
bs4: 4.7.1
html5lib: 1.0.1
sqlalchemy: 1.3.1
pymysql: None
psycopg2: None
jinja2: 2.10
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: None
gcsfs: None

[WillAyd]

Makes sense - investigation and PRs would certainly be welcome

[arpit1997]

@WillAyd Can I try this?

[gamis]

Confirmed this happening for me too.

[TomAugspurger]

Slightly simpler example

In [36]: s = pd.Series(pd.Categorical(['d', 'e', None, None], categories=['d', 'e', 'a', 'b', 'c']))

In [37]: s.fillna(pd.Series(['a', 'b', 'c', 'd']))
Out[37]:
0    a
1    b
2    c
3    d
dtype: category
Categories (5, object): [d, e, a, b, c]

@gamis I don't think anyone has opened a PR for this, if you're interested in working on it. Even diagnosing the issue would be valuable.

[MarcoGorelli]

Even simpler :)

>>> import pandas as pd
>>> pd.Categorical(['d', 'e', None, None], categories=['d', 'e', 'a', 'b', 'c']).fillna(pd.Series(['a', 'b', 'c', 'd']))
[a, b, c, d]
Categories (5, object): [d, e, a, b, c]