DataFrame aggregation methods inconsistent when object columns have missing values

[tdpetrou]

Code Sample, a copy-pastable example if possible

>>> df = pd.DataFrame({'A':['a', None], 'B':['a', 'b']})
>>> df.min()
Series([], dtype: float64)

>>> df[['B', 'A']].min()
B    a
dtype: object

Problem description

Both DataFrames should return the same thing. Order should not matter here. I suppose this is happening based on whether or not the first column has any missing values?

This happens for the other aggregation methods max and sum

Expected Output

For consistency, both should output the same thing.

Output of pd.show_versions()

INSTALLED VERSIONS ------------------ commit: None python: 3.6.1.final.0 python-bits: 64 OS: Darwin OS-release: 15.6.0 machine: x86_64 processor: i386 byteorder: little LC_ALL: None LANG: en_US.UTF-8 LOCALE: en_US.UTF-8

pandas: 0.20.2
pytest: 3.0.7
pip: 9.0.1
setuptools: 35.0.2
Cython: 0.25.2
numpy: 1.13.0
scipy: 0.19.0
xarray: None
IPython: 6.0.0
sphinx: 1.5.5
patsy: 0.4.1
dateutil: 2.6.0
pytz: 2017.2
blosc: None
bottleneck: 1.2.0
tables: 3.4.2
numexpr: 2.6.2
feather: None
matplotlib: 2.0.2
openpyxl: 2.4.7
xlrd: 1.0.0
xlwt: 1.2.0
xlsxwriter: 0.9.6
lxml: 3.7.3
bs4: 4.6.0
html5lib: 0.999999999
sqlalchemy: 1.1.9
pymysql: None
psycopg2: None
jinja2: 2.9.6
s3fs: None
pandas_gbq: None
pandas_datareader: 0.3.0.post

[jreback]

generally doing numeric operations on object dtypes is pretty suspect, though I suppose this should work. If you'd have a look would be great.

[jondo]

My workaround:

In: [df[col].dropna().min() for col in df.columns]
Out: ['a', 'a']

(For some reason, this is also much faster on my current data.)

[mroeschke]

This looks to work on master. Could use a test.

In [63]: >>> df = pd.DataFrame({'A':['a', None], 'B':['a', 'b']})
    ...: >>> df.min()
Out[63]:
B    a
dtype: object

In [64]: df[['B', 'A']].min()
Out[64]:
B    a
dtype: object

In [65]: pd.__version__
Out[65]: '0.26.0.dev0+593.g9d45934af'