Localize a column of Timestamps

[hayd]

Migrated from StackOverflow:

 s = pd.Series(pd.date_range('2012-1-1 1:30', periods=3, freq='min', tz='EST'))

You can convert each item individually:

In [86]: s[0]
Out[86]: <Timestamp: 2012-01-01 06:30:00>

In [87]: s[0].tz_localize('UTC')
Out[87]: <Timestamp: 2012-01-01 06:30:00+0000 UTC, tz=UTC>

But when you try to apply across all columns (same with convert_dtype=True):

In [81]: s.apply(lambda x: x.tz_localize('UTC'), convert_dtype=False)
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-81-0b4fa5c03650> in <module>()
----> 1 s.apply(lambda x: x.tz_localize('UTC'), convert_dtype=False)

/usr/lib/pymodules/python2.7/pandas/core/series.pyc in apply(self, func, convert_dtype, args, **kwds)
   2244             return f(self)
   2245 
-> 2246         mapped = lib.map_infer(self.values, f, convert=convert_dtype)
   2247         if isinstance(mapped[0], Series):
   2248             from pandas.core.frame import DataFrame

/usr/lib/pymodules/python2.7/pandas/lib.so in pandas.lib.map_infer (pandas/lib.c:39084)()

<ipython-input-81-0b4fa5c03650> in <lambda>(x)
----> 1 s.apply(lambda x: x.tz_localize('UTC'), convert_dtype=False)

AttributeError: 'numpy.datetime64' object has no attribute 'tz_localize'

One workaround is forcing it to be a Timestamp:

In [91]: from pandas.lib import Timestamp

In [92]: s.apply(lambda x: Timestamp(x).tz_localize('UTC')) 
Out[92]: 
0    2012-01-01 06:30:00+00:00
1    2012-01-01 06:31:00+00:00
2    2012-01-01 06:32:00+00:00

[jreback]

Seems to be working now.....

In [8]: s = pd.Series(pd.date_range('2012-1-1 1:30', periods=3, freq='min', tz='EST'))

In [9]: s
Out[9]: 
0   2012-01-01 06:30:00
1   2012-01-01 06:31:00
2   2012-01-01 06:32:00
dtype: datetime64[ns]

In [10]: s.apply(lambda x: x.tz_localize('UTC'), convert_dtype=False)
Out[10]: 
0    2012-01-01 06:30:00+00:00
1    2012-01-01 06:31:00+00:00
2    2012-01-01 06:32:00+00:00
dtype: object

In [11]: s.apply(lambda x: x.tz_localize('UTC'))
Out[11]: 
0    2012-01-01 06:30:00+00:00
1    2012-01-01 06:31:00+00:00
2    2012-01-01 06:32:00+00:00
dtype: object

In [12]: s.apply(lambda x: x.tz_localize('UTC'))[0]
Out[12]: Timestamp('2012-01-01 06:30:00+0000', tz='UTC')