Support unsubstituted type variables with both a default and a bound or constraints

[FyZzyss]

Change Summary

Add support for Bound and Default together in TypeVar:

from pydantic import BaseModel

class MyBaseModel(BaseModel):
    field: str

class ExtendedBaseModel(MyBaseModel):
    another_field: str

T = TypeVar('T', bound=MyBaseModel, default=ExtendedBaseModel)

Related issue number

fixes #9418

Checklist

• The pull request title is a good summary of the changes - it will be used in the changelog
• Unit tests for the changes exist
• Tests pass on CI
• Documentation reflects the changes where applicable
• My PR is ready to review, please add a comment including the phrase "please review" to assign reviewers

Selected Reviewer: @sydney-runkle

[codspeed-hq]

CodSpeed Performance Report

Merging #10789 will not alter performance

_{Comparing FyZzyss:9418-support-equals-bound-and-default (22a2bd4) with main (0157e34)}

Summary

✅ 44 untouched benchmarks

[github-actions]

Coverage report

Click to see where and how coverage changed

File	Statements	Missing	Coverage	Coverage (new stmts)	Lines missing
pydantic/_internal
_generate_schema.py
Project Total

This report was generated by python-coverage-comment-action

[FyZzyss]

please review

[sydney-runkle]

@FyZzyss,

Thanks for your contribution! We'll review this soon - working on pushing v2.10 out now, but I'm anticipating we can get this into v2.11 :)

[sydney-runkle]

Left some initial thoughts... I think maybe we could support an identical bound and default. From the original issue, this would require some addition like:

if bound is not None and default is not None and bound != default:
    raise RuntimeError(...)

Happy to get @Viicos' thoughts here too, he's got some great insight on detailed types like this.

[Viicos]

I think it's safe to support all combinations with bounds, constraints and defaults. In any case, we prioritize the default type over constraints or bound:

T1 = TypeVar('T1', bound=int, default=SubInt) --> generates a schema for `SubInt`
T2 = TypeVar('T2', int, str, default=int) --> generates a schema for `int`

Having a different bound and default shouldn't be an issue. The only requirement -- from a static type checking perspective -- is that default should be a subtype of bound. But such checks are out of scope for Pydantic (see #10462 (comment)).

This should work:

    def _unsubstituted_typevar_schema(self, typevar: typing.TypeVar) -> core_schema.CoreSchema:
        try:
            has_default = typevar.has_default()
        except AttributeError:
            # Happens if using `typing.TypeVar` on Python < 3.13
            pass
        else:
            if has_default:
                return self.generate_schema(typevar.__default__)  # pyright: ignore[reportAttributeAccessIssue]

        if typevar.__constraints__:
            return self._union_schema(typing.Union[typevar.__constraints__])

        if typevar.__bound__:
            schema = self.generate_schema(typevar.__bound__)
            schema['serialization'] = core_schema.wrap_serializer_function_ser_schema(
                lambda x, h: h(x), schema=core_schema.any_schema(),
            )
            return schema

        return core_schema.any_schema()

[FyZzyss]

Totally agree that it's safe.
I started this PR as supporting equals bound/default, but recognized that it's unnecessary limitation.

This code only works for unsubstituted models, so we simply prioritize the default type (expected behavior for me).
Python Type checker will check if default is not a subtype of bound in TypeVar.

I will take Viicos suggestion

[Viicos]

Python will check and won't start if default is not a subtype of bound in TypeVar.

It doesn't, only static type checkers enforce it :)

[sydney-runkle]

This LGTM otherwise, just one minor suggestion :)

[Viicos]

Thanks!