Masked arrays with hard masks don't read as missing · Issue #24574 · pandas-dev/pandas (original) (raw)

Code Sample, a copy-pastable example if possible

arr = np.ma.array([1.0, 2.0], mask=[True, False], dtype='float64') arr_hard = np.ma.array([1.0, 2.0], mask=[True, False], dtype='float64', hard_mask=True) print(pd.DataFrame(arr)) print(pd.DataFrame(arr_hard))

Normal masked array:

0
0	NaN
1	2.0

Hard masked array:

0
0	1.0
1	2.0

Problem description

I expected the output for a "hard" masked array to be the same as a regular masked array – I thought both would give NaN for the masked value. If the two cases behaved the same, it would be less surprising.

The issue is the same for Series as well as DataFrame, as well as other types of float.

Expected Output

Normal masked array: no change

Hard masked array:

0
0	NaN
1	2.0

Output of `pd.show_versions()`

INSTALLED VERSIONS

commit: None
python: 3.7.1.final.0
python-bits: 64
OS: Linux
OS-release: 4.18.0-13-generic
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: en_US.UTF-8