BUG: tz lost on combine first/fillna

[jreback]

http://stackoverflow.com/questions/24502838/pandas-tzinfo-lost-by-combine-first

In [10]: df1 = pd.DataFrame(columns=['UTCdatetime','abc'],
   ....:                   data=pd.to_datetime('20100101 01:01').tz_localize('UTC'),
   ....:                   index=pd.date_range('20140627', periods=1))

In [11]: print(df1)
                          UTCdatetime                        abc
2014-06-27  2010-01-01 01:01:00+00:00  2010-01-01 01:01:00+00:00

In [12]: df2 = pd.DataFrame(columns=['UTCdatetime','xyz'],
   ....:                   data=pd.to_datetime('20121212 12:12').tz_localize('UTC'),
   ....:                   index=pd.date_range('20140628', periods=1))

In [13]: 

In [13]: df1
Out[13]: 
                          UTCdatetime                        abc
2014-06-27  2010-01-01 01:01:00+00:00  2010-01-01 01:01:00+00:00

In [14]: df2
Out[14]: 
                          UTCdatetime                        xyz
2014-06-28  2012-12-12 12:12:00+00:00  2012-12-12 12:12:00+00:00

In [15]: df1.dtypes
Out[15]: 
UTCdatetime    object
abc            object
dtype: object

In [16]: df2.dtypes
Out[16]: 
UTCdatetime    object
xyz            object
dtype: object

In [17]: df2[['UTCdatetime']].combine_first(df1)
Out[17]: 
                   UTCdatetime                        abc
2014-06-27 2010-01-01 01:01:00  2010-01-01 01:01:00+00:00
2014-06-28 2012-12-12 12:12:00                        NaN

In [18]: df2[['UTCdatetime']].combine_first(df1).dtypes
Out[18]: 
UTCdatetime    datetime64[ns]
abc                    object
dtype: object

[fantabolous]

This also happened with fillna, so perhaps it's a more wide-ranging problem?

    utcdate = pd.to_datetime('20000101', utc=True)
    df1 = pd.DataFrame(columns=['UTCdatetime'], data=utcdate, index=pd.date_range('20140627', periods=1))
    print(df1)
    df1 = df1.fillna(utcdate)
    print(df1)

                          UTCdatetime
2014-06-27  2000-01-01 00:00:00+00:00

           UTCdatetime
2014-06-27  2000-01-01

[jreback]

this is a related issue, its marked as a bug. feel free to do a pull-request to fix.

[fantabolous]

Thanks. As much as I'd love to fix it pandas is my first foray into really programming so I think I'd probably do more harm than good at this stage.. ;) Later hopefully!