set locale to data frame with ambiguous local time leads to

[phaabe]

I read CSV-files with ambiguous local time -- ambiguous due to daylight saving time.

When I apply tz_localize() I run in the following error:

AmbiguousTimeError: There are %i dst switches when there should only be 1.

Here is my code to reproduce the error. Note that from 2am to 3am every datetime is a duplicate due to daylight saving time.

index = pd.DatetimeIndex(['2018-10-28 01:00:00', '2018-10-28 01:15:00',
                          '2018-10-28 01:30:00', '2018-10-28 01:45:00',
                          '2018-10-28 02:00:00', '2018-10-28 02:00:00',
                          '2018-10-28 02:15:00', '2018-10-28 02:15:00',
                          '2018-10-28 02:30:00', '2018-10-28 02:30:00',
                          '2018-10-28 02:45:00', '2018-10-28 02:45:00',
                          '2018-10-28 03:00:00', '2018-10-28 03:15:00',
                          '2018-10-28 03:30:00', '2018-10-28 03:45:00',
                          '2018-10-28 04:00:00'], freq='infer')

data = list(range(len(index)))

df = pd.DataFrame(data=data,index=index)
df.index = df.index.tz_localize('Europe/Berlin', ambiguous='infer')

I don't quite understand why the error occurs even with parameter ambiguous='infer'.

Does my dataframe have to be "sorted" in any way? I think data frames are ordered by default since there is DataFrame.iloc

Or is there a specific problem with pd.read_csv()?

I wish to apply ambiguous='infer' with respect to the order of the CSV file. Precisly I mean: the order of the ambiguous time format helps to differ between summer and winter time.

I would be glad about help for a solution and about a broader explanation. I couldn't find out more about how order can play a role in data frames--does it sometimes?

If order is not the problem here, than I believe this is an issue.

[mroeschke]

The infer logic essentially looks for subsequent dates that "fall behind" a previous date (see the example in https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.Series.tz_localize.html). Your index is monotonically increasing.

Nonetheless, it looks like the error message is malformed and could be fixed. PR's to fix that are welcome!

[phaabe]

Thanks for the clear answer.

But then I see a feature request here. Would you agree?

[mroeschke]

Indeed. The infer logic could be improved to handle cases like this!