localize timestamp in pandas

Question

There are some questions about how to localiza a timestamp in pandas, but all of them require the column to be localized to be the index. I don't want that, I want to localize a column, which in not the index, for example:

df = pd.DataFrame({'start': pd.to_datetime([1439687730.439, 1439687731.439], unit='s')})
df['start'].tz_localize('utc')

this return me an error:

TypeError: index is not a valid DatetimeIndex or PeriodIndex

My goal is to convert the timestamp in my localtime (CEST) using pandas (datetime does it automatically).

Answer 1

You need to use .dt accessor as you're operating on a Series not a DatetimeIndex:

In [191]:
df = pd.DataFrame({'start': pd.to_datetime([1439687730.439, 1439687731.439], unit='s')})
df['start'].dt.tz_localize('utc')

Out[191]:
0   2015-08-16 01:15:30.439000+00:00
1   2015-08-16 01:15:31.439000+00:00
Name: start, dtype: datetime64[ns, UTC]

the tz_localize method is only available on datetimeIndex objects hence the error but it's available through the dt accessor