fix: dataclass wrapper was not always called

[PrettyWood]

Change Summary

In a stdlib dataclass without __post_init__, __post_init__ call is not added in the generated __init__.
So when we inherit from a stdlib dataclass and we add a __post_init__, it won't be called. So all its related code + validation will be ignored.
Now it should be okish (not ok because it's a mess).

In v2, we should rely on stdlib dataclasses and go with a more functional approach

Related issue number

fix #4477

Checklist

• Unit tests for the changes exist
• Tests pass on CI and coverage remains at 100%
• Documentation reflects the changes where applicable
• changes/<pull request or issue id>-<github username>.md file added describing change
  (see changes/README.md for details.
  You can skip this check if the change does not need a change file.)
• My PR is ready to review, please add a comment including the phrase "please review" to assign reviewers

[samuelcolvin]

LGTM, I'll fix the markdown in the change file when building history (no need for extra builds).

[samuelcolvin]

otherwise LGTM, please update.

[PrettyWood]

The fix is wrong. I'm pushing something new that handles properly inheritance and stuff (I added a new test).
But it's not acceptable yet. The check on is_builtin_dataclass needs to be more resilient (it's a mess honestly...)

[samuelcolvin]

😞 I agree, let me know when it's ready to review.

[samuelcolvin]

@PrettyWood what do you think, maybe better to wait for another patch release and get the other changes released?

[PrettyWood]

@samuelcolvin it's good! Sorry I had client calls all day. Hard to do that at the same time 😓

[samuelcolvin]

@samuelcolvin it's good! Sorry I had client calls all day. Hard to do that at the same time sweat

I'm sure! Sorry about this.

[samuelcolvin]

LGTM, thanks so much.

[PrettyWood]

I feel like __pydantic_model__ and all should be attached to the dataclass wrapper and not the dataclass itself. But I'm too tired to think about it now. Hopefully we don't have new bugs with dataclasses in v1 and we'll rely directly on stdlib in v2

[samuelcolvin]

agreed, the whole idea of the Argument type (pydantic/pydantic-core#203) is that we can just call

args, kwargs = schema_validator.validate_python(input)
return dc(*args, **kwargs)

and never touch the dataclass.

[PrettyWood]

Yup! That's great 👍