pd.concat reset the tz-aware index to UTC

[alessioarena]

Hi,

Having two DataFrames with index being tz-aware Timestamp objects

print(concat_df[0].index.tz, concat_df[1].index.tz)
fxt = pd.concat(concat_df, axis=1)
print(fxt.index.tz)

This leads to the following output

Australia/Melbourne Australia/Melbourne
UTC

I believe this may be in some way related to #7562

pandas version = 0.21.0

[jreback]

pls show a copy pastable example

[sinhrks]

Thanks for the report. Unfortunatelly, I couldn't reproduce the issue using 0.21.0. Can you attach reproducible code and pd.show_versions()?

import pandas as pd

idx = pd.date_range('2011-01-01', periods=3, freq='H', 
                    tz='Australia/Melbourne')
df = pd.DataFrame({'a': [1, 2, 3]}, index=idx)

res = pd.concat([df, df], axis=1)
res.index.tz
# <DstTzInfo 'Australia/Melbourne' LMT+9:40:00 STD>

[alessioarena]

Hi there,

Unfortunately those files are quite large, and I cannot upload them.
I tried to make some synthetic data and also could not reproduce it.

However I have some additional information about this issue:

The first dataset (df_concat[0]) has a Timestamp object converted from UTC as following:

dft['local_time'] = pd.to_datetime(dft.timestamp_utc, utc=True, format="%d/%m/%Y %H:%M:%S")
dft.index = dft.index.tz_convert('Australia/Melbourne')

This lead to
<DstTzInfo 'Australia/Melbourne' LMT+9:40:00 STD>

The second dataset (df_concat[1]) has the Timestamp object being localized as following
df.index = df.index.tz_localize('Australia/Melbourne', ambiguous=np.ones(len(df.index), dtype=bool))
This lead to
<DstTzInfo 'Australia/Melbourne' AEDT+11:00:00 DST>

As you can see the resulting timezones are both Australia/Melbourne, but they are expressed differently.
I believe that this is the cause of the problem describe, and I also suspect that the datasets are misaligned during this operation. I will do more testing on this and update you.

[antvig]

Hi,
I get the same issue, one of my DataFrame index is <DstTzInfo 'Europe/Paris' CET+1:00:00 STD>, and the other one is <DstTzInfo 'Europe/Paris' LMT+0:09:00 STD>.
When I concatenate the two dataframe, the resulting dataframe index is in UTC.

I work with Pandas 18.0 but I guess this is the same "issue" (maybe it's not really an issue)

[jreback]

closing as no repro.

If you have a copy-pastable example pls open a new issue.