Operations on NaT returning float instead of datetime64[ns]

[rrbarbosa]

Code Sample, a copy-pastable example if possible

df = pd.DataFrame({'t1': pd.NaT, 't2': pd.NaT}, index=[1])
df.min(axis=1)
df.max(axis=1)
df.sum(axis=1)

Expected Output

All results above return a float64 series instead of a datetime64[ns] series.

output of pd.show_versions()

INSTALLED VERSIONS
------------------
commit: None
python: 3.5.1.final.0
python-bits: 64
OS: Darwin
OS-release: 15.4.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: en_US.UTF-8
LANG: en_US.UTF-8

pandas: 0.18.0
nose: None
pip: 8.1.1
setuptools: 20.9.0
Cython: None
numpy: 1.11.0
scipy: 0.17.0
statsmodels: 0.6.1
xarray: None
IPython: 4.2.0
sphinx: None
patsy: 0.4.1
dateutil: 2.5.3
pytz: 2016.3
blosc: None
bottleneck: None
tables: None
numexpr: 2.5.2
matplotlib: 1.5.1
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml: None
bs4: 4.4.1
html5lib: None
httplib2: None
apiclient: None
sqlalchemy: 1.0.12
pymysql: 0.7.2.None
psycopg2: 2.6.1 (dt dec pq3 ext lo64)
jinja2: 2.8
boto: None

[jorisvandenbossche]

SO: http://stackoverflow.com/questions/36620256/why-is-pandas-converting-datetimes-to-float-in-aggregate-function#

[jorisvandenbossche]

The min/max should return NaT, but the sum should fail I think for datetime64

[jreback]

so .sum(..) is correct, in the the underlying operation raises a TypeError, which is then caught and numeric_only=True is passed so you end up with .sum() of an empty frame and get 0.0.

The other 2 are not being masked correctly inside nanops._maybe_null_out which needs to handle M8,m8 dtypes (rather than converting to float).

• same cause is here: agg() function on groupby dataframe changes dtype of datetime64[ns] column to float64 if all items in a single group are NaT #12821
• xref BUG: Series.median() treats NaT as INT_MIN #8617 (might be a duplicate)
• xref BUG: min on non-numeric data with nans #4147