MobileClientExtractor error during extract() on days with daylight savings on non-UTC systems

[mdragaschnig]

Using MobileClientExtractor extract() with data on days when daylight saving starts (e.g. 2018-03-25) leads to an error

>           raise exception
E           pymeos_cffi.errors.MeosInvalidArgValueError: MeosInvalidArgValueError (12): Timestamps for temporal value must be increasing: 2018-03-25 03:05:00+02, 2018-03-25 03:00:00+02

../.venv/lib/python3.12/site-packages/pymeos_cffi/errors.py:178: MeosInvalidArgValueError

[mdragaschnig]

On systems with non-UTC timezone.

Unittest to document the error:

mobiml/tests/test_mobile_client_extractor.py

Line 240 in 1456bff

def test_pymeos(self):

[mdragaschnig]

The bug occurs with AIS data on a day where daylight savings start (e.g. 2018-03-25):

• movingpandas drops timezones for trajectories.
• mobiml MobileClientExtractor creates a wkt string from those movingpandas-based trajectory dataframes.
• pymeos TGeogPointSeq consumes this wkt. It faints on non-UTC systems if the wkt string has tz-unaware timestamps with "Timestamps must be increasing", as a daylight-saving tz is applied which causes non-monotonous timestamps (02:00:00 -> 03:00:00)

2018-03-25 02:05:00 -> 2018-03-25 03:05:00+02,
2018-03-25 03:00:00 -> 2018-03-25 03:00:00+02

[mdragaschnig]

Could maybe use df_orig_tz of the trajectory? This is set if the original dataframe had a timezone set.

So e.g.:
traj.df.geometry.index.tz_localize(traj.df_orig_tz).astype(str) when creating the wkt string

[anitagraser]

Alternatively, the mobiml.Dataset class could also store everything in UTC internally