API: Should Index.min and max use nanmin and nanmax?

[sinhrks]

Index and Series min and max handles nan and NaT differently. Even though min and max are defined in IndexOpsMixin, Series doesn't use them and use NDFrame definitions.

pd.Index([np.nan, 1.0]).min()
# nan

pd.Index([np.nan, 1.0]).max()
# nan

pd.DatetimeIndex([pd.NaT, '2011-01-01']).min()
# NaT

pd.DatetimeIndex([pd.NaT, '2011-01-01']).max()
#2011-01-01 00:00:00

# Series excludes nan and NaT
pd.Series([np.nan, 1.0]).min()
#1.0

pd.Series([np.nan, 1.0]).max()
#1.0

pd.Series([pd.NaT, pd.Timestamp('2011-01-01')]).min()
#2011-01-01 00:00:00

pd.Series([pd.NaT, pd.Timestamp('2011-01-01')]).max()
#2011-01-01 00:00:00

[jreback]

yah that is right; these are actually overriden in core/generic.py for Series; so these can/should directly use nanmin/max