datatype with __slots__ breaks copy.deepcopy

[zechigan]

Checks

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

Issue description

The issue is twofold:

1. perf(python): Add __slots__ to most Polars classes #13236 added the __slots__ attributes to all the datatypes, which forced the removal of __dict__. However, on https://github.com/Object905/polars/blob/815732bc0f5bc06e07f3a995c164fa87e3c53ce8/py-polars/polars/datatypes/classes.py#L94 __dict__ is still being accessed
2. After fixing the first issue by replacing self.__dict__ with self.__slots__, one passes a datatype into copy.deepcopy (or specifically for my case, as a field on a dataclass, which gets passed into dataclasses.asdict) will quickly find that the function relies on the data structure having the __dict__ attribute and as a result throws: https://github.com/python/cpython/blob/6359141867595f84f9e803e1053ae51db3203c49/Lib/copy.py#L268

Reproducible example

>>> import polars as pl
>>> from copy import deepcopy
>>> deepcopy(pl.Int64())
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/zechigan/anaconda3/lib/python3.11/copy.py", line 161, in deepcopy
    rv = reductor(4)
         ^^^^^^^^^^^
  File "/Users/zechigan/anaconda3/lib/python3.11/site-packages/polars/datatypes/classes.py", line 96, in __reduce__
    return (_custom_reconstruct, (type(self), object, None), self.__dict__)
                                                             ^^^^^^^^^^^^^
AttributeError: 'Int64' object has no attribute '__dict__'. Did you mean: '__dir__'?

2. (After addressing 1 by replacing self.__dict__ with self.__slots__)

>>> from dataclasses import dataclass, asdict
>>> from typing import Any
>>> import polars as pl
>>> @dataclass
... class DummyDataClass:
...     schema: Any
... 
>>> dc = DummyDataClass(pl.DataFrame({'a':[1]}).schema)
>>> asdict(dc)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/zechigan/anaconda3/lib/python3.11/dataclasses.py", line 1277, in asdict
    return _asdict_inner(obj, dict_factory)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/zechigan/anaconda3/lib/python3.11/dataclasses.py", line 1284, in _asdict_inner
    value = _asdict_inner(getattr(obj, f.name), dict_factory)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/zechigan/anaconda3/lib/python3.11/dataclasses.py", line 1314, in _asdict_inner
    return type(obj)((_asdict_inner(k, dict_factory),
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/zechigan/anaconda3/lib/python3.11/dataclasses.py", line 1315, in <genexpr>
    _asdict_inner(v, dict_factory))
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/zechigan/anaconda3/lib/python3.11/dataclasses.py", line 1318, in _asdict_inner
    return copy.deepcopy(obj)
           ^^^^^^^^^^^^^^^^^^
  File "/Users/zechigan/anaconda3/lib/python3.11/copy.py", line 172, in deepcopy
    y = _reconstruct(x, memo, *rv)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/zechigan/anaconda3/lib/python3.11/copy.py", line 280, in _reconstruct
    y.__dict__.update(state)
    ^^^^^^^^^^
AttributeError: 'Int64' object has no attribute '__dict__'. Did you mean: '__dir__'?

Expected behavior

polars datatypes can be safely passed into copy.deepcopy

Installed versions

--------Version info---------
Polars:               0.20.13
Index type:           UInt32
Platform:             macOS-10.16-x86_64-i386-64bit
Python:               3.11.3 (main, Apr 19 2023, 18:51:09) [Clang 14.0.6 ]

----Optional dependencies----
adbc_driver_manager:  <not installed>
cloudpickle:          2.2.1
connectorx:           <not installed>
deltalake:            <not installed>
fsspec:               2023.3.0
gevent:               <not installed>
hvplot:               0.8.4
matplotlib:           3.7.1
numpy:                1.24.3
openpyxl:             3.0.10
pandas:               1.5.3
pyarrow:              11.0.0
pydantic:             <not installed>
pyiceberg:            <not installed>
pyxlsb:               <not installed>
sqlalchemy:           1.4.39
xlsx2csv:             <not installed>
xlsxwriter:           <not installed>

[stinodego]

Thanks for opening an issue.

Indeed the __reduce__ method is no longer correct.

However, if I simply remove that method, deepcopy seems to work just fine. Could you elaborate on your point 2? I cannot reproduce your example if __reduce__ is deleted.

[zechigan]

Sorry, updated my example. @stinodego you are right, removing __reduce__ did solve both of my problems!