`pl.Datetime` `time_zone` parameter has no type or value check

[mcrumiller]

Checks

• [X] I have checked that this issue has not already been reported.
• [X] I have confirmed this bug exists on the latest version of Polars.

Issue Description

The time zone parameter in pl.Datetime can be an invalid string, and can even be any type of Python object. We should probably ensure that the time zone is a valid time zone when constructing a pl.Datetime datatype.

import polars as pl

pl.Datetime(time_zone="invalid_time_zone")
# Datetime(time_unit='us', time_zone='invalid_time_zone')

pl.Datetime(time_zone=object())
# Datetime(time_unit='us', time_zone=<object object at 0x7f245331d0d0>)

Installed versions

main

[MarcoGorelli]

thanks for reporting - the validation currently happens later:

In [9]: pl.Series([datetime(2020, 1, 1)], dtype=pl.Datetime(time_zone='cabbage'))
---------------------------------------------------------------------------
ComputeError                              Traceback (most recent call last)
Cell In[9], line 1
----> 1 pl.Series([datetime(2020, 1, 1)], dtype=pl.Datetime(time_zone='cabbage'))

File ~/scratch/.venv/lib/python3.11/site-packages/polars/series/series.py:312, in Series.__init__(self, name, values, dtype, strict, nan_to_null, dtype_if_empty)
    309         raise TypeError(msg)
    311 if isinstance(values, Sequence):
--> 312     self._s = sequence_to_pyseries(
    313         name,
    314         values,
    315         dtype=dtype,
    316         strict=strict,
    317         nan_to_null=nan_to_null,
    318     )
    320 elif values is None:
    321     self._s = sequence_to_pyseries(name, [], dtype=dtype)

File ~/scratch/.venv/lib/python3.11/site-packages/polars/_utils/construction/series.py:235, in sequence_to_pyseries(name, values, dtype, strict, nan_to_null)
    225         if values_tz != "UTC" and dtype_tz is None:
    226             warnings.warn(
    227                 "Constructing a Series with time-zone-aware "
    228                 "datetimes results in a Series with UTC time zone. "
   (...)
    233                 stacklevel=find_stacklevel(),
    234             )
--> 235         return s.dt.replace_time_zone(dtype_tz or "UTC")._s
    236     return s._s
    238 elif (
    239     _check_for_numpy(value)
    240     and isinstance(value, np.ndarray)
    241     and len(value.shape) == 1
    242 ):

File ~/scratch/.venv/lib/python3.11/site-packages/polars/series/utils.py:107, in call_expr.<locals>.wrapper(self, *args, **kwargs)
    105     expr = getattr(expr, namespace)
    106 f = getattr(expr, func.__name__)
--> 107 return s.to_frame().select_seq(f(*args, **kwargs)).to_series()

File ~/scratch/.venv/lib/python3.11/site-packages/polars/dataframe/frame.py:7906, in DataFrame.select_seq(self, *exprs, **named_exprs)
   7883 def select_seq(
   7884     self, *exprs: IntoExpr | Iterable[IntoExpr], **named_exprs: IntoExpr
   7885 ) -> DataFrame:
   7886     """
   7887     Select columns from this DataFrame.
   7888
   (...)
   7904     select
   7905     """
-> 7906     return self.lazy().select_seq(*exprs, **named_exprs).collect(_eager=True)

File ~/scratch/.venv/lib/python3.11/site-packages/polars/lazyframe/frame.py:1810, in LazyFrame.collect(self, type_coercion, predicate_pushdown, projection_pushdown, simplify_expression, slice_pushdown, comm_subplan_elim, comm_subexpr_elim, no_optimization, streaming, background, _eager)
   1807 if background:
   1808     return InProcessQuery(ldf.collect_concurrently())
-> 1810 return wrap_df(ldf.collect())

ComputeError: unable to parse time zone: 'cabbage'. Please check the Time Zone Database for a list of available time zones

[luke396]

Hi @MarcoGorelli, should we add some checks here

https://github.com/pola-rs/polars/blob/2970c5706a5bbc39840ca5b0ffc9d4b3aab502cb/py-polars/polars/datatypes/classes.py#L461-L463

Similar to what's done here

https://github.com/pola-rs/polars/blob/2970c5706a5bbc39840ca5b0ffc9d4b3aab502cb/py-polars/polars/_utils/convert.py#L163-L194

Or should we leave the error as it is? Do you have any thoughts on this? If possible, I'd like to improve it.

[max-muoto]

Using Pyright/MyPy, you'll get a static type-checking warning if you try and pass in object() to time_zone here. The only way this could perhaps be built on (from a type-checking perspective), is maintaining a literal of all possible timezones, and using that instead of str.

[stinodego]

Related to this: it's a bit concerning that the invalid time zone can end up in Rust without it being validated:

import polars as pl

dtype = pl.Datetime(time_zone="invalid_time_zone")

s = pl.Series(dtype=dtype)
print(s)

shape: (0,)
Series: '' [datetime[μs, invalid_time_zone]]
[
]

[MarcoGorelli]

true, validation only kicks in if there's an actual date in the column. maybe it can happen earlier

[bionicles]

hey for testing time zones in polars, can we add

df[col_name].dt.tz to get a series of time zone strings?

could be more performant than

def tz_from_df(df: pl.DataFrame, date_col_name: str = "date") -> str:
    tz = df.schema[date_col_name].time_zone
    return tz