Pandas Inconsistent Behavior of min on Timestamps w/ and w/o timezones

related #4147

Observe w/o tz:

In [46]: pd.Series([pd.NaT, pd.Timestamp("2013-1-1")]).min()
Out[46]: Timestamp('2013-01-01 00:00:00', tz=None)

however w/ tz:

In [48]: pd.Series([pd.NaT, pd.Timestamp("2013-1-1", tz="US/Eastern")]).min()
Out[48]: NaT

strangely, max works:

In [60]: pd.Series([pd.NaT, pd.Timestamp("2013-1-1", tz="US/Eastern")]).max()
Out[60]: Timestamp('2013-01-01 00:00:00-0500', tz='US/Eastern')

Comment From: jreback

this is the same reason as #4147 as these series are object dtype

Comment From: jreback

related - the string rep of this type of mixed series is wrong....

In [7]: pd.Series([pd.NaT, pd.Timestamp("2013-1-1", tz="US/Eastern")])
Out[7]: 
0                          NaN
1    2013-01-01 00:00:00-05:00
dtype: object

In [8]: pd.Series([pd.NaT, pd.Timestamp("2013-1-1", tz="US/Eastern")]).iloc[0]
Out[8]: NaT

In [9]: pd.Series([pd.NaT, pd.Timestamp("2013-1-1", tz="US/Eastern")]).iloc[1]
Out[9]: Timestamp('2013-01-01 00:00:00-0500', tz='US/Eastern')

Comment From: cancan101

@jreback The issue you refer to in regards to formatting of NaT in an object array is caused by GenericArrayFormatter taking in only one value for na_rep which defaults to NaN. When formatting, that value is used as:

            if self.na_rep is not None and lib.checknull(x):
                if x is None:
                    return 'None'
                return self.na_rep

One option would be to add another argument nat_rep.

Comment From: jorisvandenbossche

In the meantime, this works as expected:

In [9]: pd.Series([pd.NaT, pd.Timestamp("2013-1-1", tz="US/Eastern")])
Out[9]: 
0                         NaT
1   2013-01-01 00:00:00-05:00
dtype: datetime64[ns, US/Eastern]

In [10]: pd.Series([pd.NaT, pd.Timestamp("2013-1-1", tz="US/Eastern")]).min()
Out[10]: Timestamp('2013-01-01 00:00:00-0500', tz='US/Eastern')

probably since timezone aware datetime is now a proper dtype and no longer object