DataFrame.min/max with axis=1, datetime64[ns, tz] types and NaT values returns NaNs

[jkukul]

Code Sample, a copy-pastable example if possible

s1 = pd.Series([pd.Timestamp('2017-01-01T10', tz='UTC'), pd.Timestamp('2017-01-01T11', tz='UTC')])
s2 = pd.Series([pd.Timestamp('2017-01-01T12', tz='UTC'), pd.NaT])

df = pd.DataFrame(dict(s1=s1, s2=s2))
df.min(axis=1)

produces:

0  NaN
1  NaN
dtype: float64

Problem description

For tz-aware timestamp columns, as soon as pd.NaT value is present in any row in any of the columns, than the column-wise min()/max() operations returns all NaNs.

If columns are non-tz-aware, then min()/max() work as expected, i.e. the result is similar to the one pasted below (except for the timezone part).

Expected Output

0   2017-01-01 10:00:00+00:00
1   2017-01-01 11:00:00+00:00
dtype: datetime64[ns, UTC]

Output of pd.show_versions()

INSTALLED VERSIONS

commit : None
python : 3.6.9.final.0
python-bits : 64
OS : Darwin
OS-release : 18.7.0
machine : x86_64
processor : i386
byteorder : little
LC_ALL : en_US.UTF-8
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 0.25.1
numpy : 1.17.2
pytz : 2019.2
dateutil : 2.8.0
pip : 18.0
setuptools : 41.2.0
Cython : 0.27.3
pytest : 4.4.0
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : 4.2.6
html5lib : None
pymysql : None
psycopg2 : 2.7.3.2 (dt dec pq3 ext lo64)
jinja2 : 2.10.1
IPython : 7.7.0
pandas_datareader: None
bs4 : 4.6.3
bottleneck : None
fastparquet : None
gcsfs : None
lxml.etree : 4.2.6
matplotlib : 3.1.1
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : 0.11.0
pyarrow : 0.13.0
pytables : None
s3fs : None
scipy : 1.0.0
sqlalchemy : 1.3.8
tables : None
xarray : None
xlrd : None
xlwt : None
xlsxwriter : None

[mroeschke]

Thanks for the report. I think this is a duplicate of #27794.