[Python] Converting python array to TimestampArray with naive datetime and datetime with various timezones
- Langage dominant
- C++
- Étoiles
- 17.1k
- Forks
- 4.3k
- Merge moyen
- 3 j 18 h
- PR mergées (30 j)
- 91
Description
### Describe the bug, including details regarding any error messages, version, and platform.
When converting a python array with datetime elements and mixed timezones into a pyarrow array there are two points that seem to be incorrect/could be improved:
- the values seem to be calculated to the UTC timezone but the `tz` attribute of the `timestamp` is defaulted to the timezone of the first element in an array (seems wrong to me),
- together with datetime elements with timezones a naive element can also be present and it is presumed that the naive element is in UTC timezone, which is not necessarily true.
```python
>>> import zoneinfo
>>> import datetime
>>> import pyarrow as pa
# Mixed timezones without naive datetime
>>> data_mixed = [
... datetime.datetime(2006, 1, 13, 12, 34, 56, 432539, tzinfo=zoneinfo.ZoneInfo(key='US/Eastern')),
... datetime.datetime(2008, 1, 5, 5, 0, 0, 1000, tzinfo=datetime.timezone.utc),
... datetime.datetime(2010, 8, 13, 5, 0, 0, 437699, tzinfo=zoneinfo.ZoneInfo(key='Europe/Moscow')),
... ]
>>> pa.array(data_mixed)
[
2006-01-13 17:34:56.432539,
2008-01-05 05:00:00.001000,
2010-08-13 01:00:00.437699
]
>>> pa.array(data_mixed).type
TimestampType(timestamp[us, tz=US/Eastern])
# Mixed timezones with naive datetime as the first element
>>> data_mixed_with_naive_first = [
... datetime.datetime(2007, 7, 13, 8, 23, 34, 123456), # naive
... datetime.datetime(2008, 1, 5, 5, 0, 0, 1000, tzinfo=datetime.timezone.utc),
... None,
... datetime.datetime(2006, 1, 13, 12, 34, 56, 432539, tzinfo=zoneinfo.ZoneInfo(key='US/Eastern')),
... datetime.datetime(2010, 8, 13, 5, 0, 0, 437699, tzinfo=zoneinfo.ZoneInfo(key='Europe/Moscow')),
... ]
>>> pa.array(data_mixed_with_naive_first)
[
2007-07-13 08:23:34.123456,
2008-01-05 05:00:00.001000,
null,
2006-01-13 17:34:56.432539,
2010-08-13 01:00:00.437699
]
>>> pa.array(data_mixed_with_naive_first).type
TimestampType(timestamp[us])
# Mixed timezones with naive datetime not as first element
>>> data_mixed_with_naive = [
... datetime.datetime(2006, 1, 13, 12, 34, 56, 432539, tzinfo=zoneinfo.ZoneInfo(key='US/Eastern')),
... datetime.datetime(2010, 8, 13, 5, 0, 0, 437699, tzinfo=zoneinfo.ZoneInfo(key='Europe/Moscow')),
... datetime.datetime(2008, 1, 5, 5, 0, 0, 1000, tzinfo=datetime.timezone.utc),
... datetime.datetime(2007, 7, 13, 8, 23, 34, 123456), # naive
... None,
... ]
>>> pa.array(data_mixed_with_naive)
[
2006-01-13 17:34:56.432539,
2010-08-13 01:00:00.437699,
2008-01-05 05:00:00.001000,
2007-07-13 08:23:34.123456,
null
]
>>> pa.array(data_mixed_with_naive).type
TimestampType(timestamp[us, tz=US/Eastern])
```
I think that if the datetime elements with various timezones are defaulted to UTC then we should also do the same with the `tz` attribute.
As for the case where a naive element is present the conversion could turn out an error and advise the user to add a timezone or have all elements naive.
### Component(s)
Python
Guide de contribution
Ouvrir le guide de contribution
Piste de recherche
Commencez par reproduire les exemples avec pa.array et examinez le chemin de conversion de datetime Python vers TimestampArray. Comparez les valeurs d’horodatage obtenues et les métadonnées de fuseau horaire pour des datetimes aware mélangés et pour des listes contenant des datetimes naive. La tâche doit établir des métadonnées de fuseau horaire cohérentes ou une erreur claire pour les valeurs naive et avec fuseau horaire mélangées, avec une couverture de régression pour les cas signalés.
Rédigé par le modèle d'indexation à partir du texte de l'issue.
Évaluation
- Stack technique
- python
- Domaine
- data
- Type d'issue
- Bug
- Difficulté
- 4/5
- Temps estimé
- 3-5 jours
- Activité
- À l'abandon
- Clarté
- Plutôt claire
- Accessibilité débutants
- 35/100