[BUG] Timeseries.from_dataframe() does not handle localized datetime64 index correctly anymore

Thank you for keeping up the great work on darts!

Describe the bug In 0.9 with xarray backend of Timeseries, localized datetimeindex is not handled correctly (anymore). Using the toy example from another issue, you can see that plain datetimeindex is converted datetime64 in Timeseries. Once localized to UTC, the index is converted to object in Timeseries leading to downstream problems in plot(), etc.

To reproduce

df = pd.DataFrame(data={'First': [0,1,2,3,4,4.5,4,3,2,1]})
inds = [f'2021-01-{day}' for day in range(1, 11)]
df.index = pd.to_datetime(inds)

df_localized = df.copy()
df_localized.index = df.index.tz_localize('UTC')

ts = TimeSeries.from_dataframe(df)
ts_localized = TimeSeries.from_dataframe(df_localized)

Expected behaviour Localized datetimeindex is converted to datetime64 in Timeseries.

System Python version: Python 3.9.5 darts install: pip install u8darts

[dennisbader]

@milxss you are actually right. When I ran it with darts==0.11.0 from command line, it didn’t raise the Error (in my session from above, it probably used the darts scripts from my local v==0.12.0 git repo).

We actually implemented some changes/improvements from 0.11.0 to 0.12.0 in the way how TimeSeries are created. It is very likely that this is why the output is different between the versions.

And as you said, it’s good that the new version now raises this Error.

[hrzn]

Thanks for the investigation @dennisbader - closing the issue now.