pandas how to override attibute error from timezome localization

I have a idaydf.index that I am trying to localize timezone

DatetimeIndex(['2022-10-24 16:00:00', '2022-10-24 16:15:00',
               ...
               '2023-06-16 21:58:00', '2023-06-16 22:00:00'],
              dtype='datetime64[ns]', name='DateTime', length=9012, freq=None)

with

idaydf.index = idaydf.index.tz_localize(LOCAL_TZ)

where LOCAL_TZ is

_PytzShimTimezone(zoneinfo.ZoneInfo(key='Europe/London'), 'Europe/London')

I get this error:

*** AttributeError: 'NoneType' object has no attribute 'total_seconds'

I have these versions:

python3-3.11.3
pandas-1.5.3
pytz-2023.3-1
tzlocal 4.2

How to fix?

Solution

You could use the string representation of the timezone object to decouple. That way,

pandas can chose whatever library it wants to handle the timezone (they're supposed to switch from pytz to zoneinfo still in 2023)
the library that generates LOCAL_TZ can switch to zoneinfo without crashing your code
you can avoid try/excepts

Ex:

from zoneinfo import ZoneInfo
import pytz_deprecation_shim as pds
import pandas as pd

LOCAL_TZ = pds.timezone('Europe/London')
print(repr(LOCAL_TZ))
# _PytzShimTimezone(zoneinfo.ZoneInfo(key='Europe/London'), 'Europe/London')

LOCAL_TZ_new = ZoneInfo('Europe/London')
print(repr(LOCAL_TZ_new))
# zoneinfo.ZoneInfo(key='Europe/London')

dti = pd.date_range('2022-10-24 16:00:00', '2023-06-16 22:00:00')

dti0 = dti.tz_localize(str(LOCAL_TZ))
dti1 = dti.tz_localize(str(LOCAL_TZ_new))

assert (dti0 == dti1).all()

Btw. tzlocal offers a way to directly obtain the IANA tz name as a string, which might make things even easier here -

from tzlocal import get_localzone_name
get_localzone_name()
# "Europe/Berlin" # on my machine...

This works already with tzlocal version 4.2