ImmutableValidatedObject: Support nested Mapping types

[talyz]

Since typeguard was updated in #1569, the use of deployment.keys (+ a few required options in nixops-gce, like bootstrapImage), has been broken. This is because typeguard now inspects the type of mapping containers (Mapping[str, KeyOptions] and such) and nested such containers weren't handled by ImmutableValidatedObject.

This also improves the handling of nested sequences and fixes their associated test.

fixes #1570

[pSub]

This fixes #1570 for me. Is there anything that prevents this from being merged and released? Maybe the mypy-ratchet check?

Would be great to have a release with this bug fixed!

[roberth]

This rejects strings, but in the new code all Sequences are matched, because

>>> isinstance("hi", collections.abc.Sequence)
True

Could you exclude strings from the match, to be sure, or perhaps explain why strings don't occur here if that's the case?

[talyz]

No, not all Sequences are matched, only those explicitly annotated as such. This seems more in line with what is expected from the comments, tests and existing code, afaict.

>>> import typing
>>> import collections.abc
>>> class test_class:
...     test_string: str
...     test_tuple: typing.Tuple[test_class]
...     test_list: typing.List[test_class]
...     test_sequence: typing.Sequence[test_class]
...     test_dict: typing.Dict[str, test_class]
...     test_mapping: typing.Mapping[str, test_class]
...
>>> for k, v in typing.get_type_hints(test_class).items():
...     match typing.get_origin(v):
...             case collections.abc.Sequence | collections.abc.Mapping:
...                     print(f"{k}, {v}")
... 
test_sequence, typing.Sequence[__main__.test_class]
test_mapping, typing.Mapping[str, __main__.test_class]

[roberth]

Suggested change

	new_value: Any = []
	new_value: List[Any] = []

or maybe this could be more specific.

If the mypy-ratchet job still fails after that, I won't complain.

[talyz]

Did this and added Dict to the mapping one + renamed them to avoid variable redefinition. The ratchet still fails locally for me, though. Maybe it's time to disable it?

[roberth]

Maybe it's time to disable it?

I think we should keep it around, so that regressions in type coverage don't go unnoticed, but also it's ok to override it when we don't seem to have a good way to improve the types in the change.

[talyz]

Yeah, it's just not very helpful, since it doesn't actually point to where you have undeclared types. Does it even do what it's supposed to? I really can't tell. It doesn't exactly make it easier to contribute to NixOps, though.

[roberth]

Code looks ok, fixes the problem, has tests.