fix for #8512

[Anvil]

Change Summary

This is hopefully a fix for #8512, that I've reported a few days ago.

Basically, it makes PlainValidator.__get_pydantic_core_schema__ reuse other annotations informations.

If I did anything wrong, please let me know.

Related issue number

fix #8512

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: @lig

[Anvil]

please review

[codspeed-hq]

CodSpeed Performance Report

Merging #8567 will not alter performance

_{Comparing Anvil:plain-validator-and-serializer (1c222e9) with main (8060fa1)}

Summary

✅ 10 untouched benchmarks

[sydney-runkle]

These changes look good overall - I've requested a small addition to your test!

[davidhewitt]

I think we can avoid introducing a wrapper here and just do

Suggested change

	serialization = core_schema.wrap_serializer_function_ser_schema(function=lambda v, h: h(v), schema=schema)
	serialization = schema.get('serialization')

[Anvil]

Hello.

Thanks for the advice. I originally tried that, but ended up with some type checking errors:

Typecheck................................................................Failed
- hook id: typecheck
- exit code: 1

/home/anvil/git/pydantic/pydantic/functional_validators.py
  /home/anvil/git/pydantic/pydantic/functional_validators.py:163:68 - error: Impossible d’affecter l’argument de type « SerSchema | IncExDictOrElseSerSchema | IncExSeqOrElseSerSchema | None » au paramètre « serialization » de type « SerSchema | None » dans la fonction « with_info_plain_validator_function » (reportGeneralTypeIssues)
  /home/anvil/git/pydantic/pydantic/functional_validators.py:167:85 - error: Impossible d’affecter l’argument de type « SerSchema | IncExDictOrElseSerSchema | IncExSeqOrElseSerSchema | None » au paramètre « serialization » de type « SerSchema | None » dans la fonction « no_info_plain_validator_function » (reportGeneralTypeIssues)
2 errors, 0 warnings, 0 informations

Should I cast the value ?

[Anvil]

ha, pardon my french.

PDM, version 2.11.2
pdm run ruff pydantic tests docs/plugins
pdm run ruff format --check pydantic tests docs/plugins
164 files already formatted
pre-commit 3.6.0
pre-commit run typecheck --all-files
Typecheck................................................................Failed
- hook id: typecheck
- exit code: 1

/home/anvil/git/pydantic/pydantic/functional_validators.py
  /home/anvil/git/pydantic/pydantic/functional_validators.py:163:68 - error: Argument of type "SerSchema | IncExDictOrElseSerSchema | IncExSeqOrElseSerSchema | None" cannot be assigned to parameter "serialization" of type "SerSchema | None" in function "with_info_plain_validator_function" (reportGeneralTypeIssues)
  /home/anvil/git/pydantic/pydantic/functional_validators.py:167:85 - error: Argument of type "SerSchema | IncExDictOrElseSerSchema | IncExSeqOrElseSerSchema | None" cannot be assigned to parameter "serialization" of type "SerSchema | None" in function "no_info_plain_validator_function" (reportGeneralTypeIssues)
2 errors, 0 warnings, 0 informations

[adriangb]

with_info_plain_validator_function accepts a SerSchema: https://github.com/pydantic/pydantic-core/blob/4da7192ffc104cd6c424d102bc1c9c5bfdad543e/python/pydantic_core/core_schema.py#L2183

But some schemas, e.g. dict have other serializers: https://github.com/pydantic/pydantic-core/blob/4da7192ffc104cd6c424d102bc1c9c5bfdad543e/python/pydantic_core/core_schema.py#L1724-L1733

@davidhewitt I don't recall what the deal was with the inc/exc stuff but I'd suggest one of two things:

• They get added to the SerSchema union if this usage is valid
• In this PR we add some assert schema['type'] in ... or assert schema['type'] not in ...

[davidhewitt]

Potentially instead of the assert we can fall back to the slower wrap serializer proposed here just to be safe?

[Anvil]

I dont get all the details (yet?), but if I can be of any use, let me know.

[sydney-runkle]

Ah, adding the inc/exc stuff to the SerSchema union isn't easily done - for ex:

IncExSeqOrElseSerSchema = Union[IncExSeqSerSchema, SerSchema]
IncExDictOrElseSerSchema = Union[IncExDictSerSchema, SerSchema]

We would end up having recursive defs in the SerSchema union, as this is how those types are defined. For now, I think it makes sense to do a type check, and perhaps we could add an issue to the pydantic-core board if we want to figure out a way to add those types to the SerSchema Union in the long term.

[sydney-runkle]

Alright, a proposed fix:

schema = handler(source_type)

# if the serialization schema is of type 'include-exclude-dict' or 'include-exclude-set',
# we need to wrap it in a serializer function schema
# otherwise, we can leave it as is (which is more efficient)
serialization = schema.get('serialization')
if serialization and serialization['type'] in ('include-exclude-dict', 'include-exclude-set'):
    serialization = core_schema.wrap_serializer_function_ser_schema(function=lambda v, h: h(v), schema=schema)

info_arg = _inspect_validator(self.func, 'plain')
if info_arg:
    func = cast(core_schema.WithInfoValidatorFunction, self.func)
    return core_schema.with_info_plain_validator_function(
        func,
        field_name=handler.field_name,
        serialization=serialization,  # type: ignore
    )
else:
    func = cast(core_schema.NoInfoValidatorFunction, self.func)
    return core_schema.no_info_plain_validator_function(func, serialization=serialization)  # type: ignore

This requires some # type: ignore statements, which isn't ideal. However, it's not possible to check if something is an instance of SerSchema, as that's a union of types. Furthermore, checking against the individual types is difficult bc they're often TypedDict types, which aren't friendly with isinstance checks.

[sydney-runkle]

Actually, let's go with:

schema = handler(source_type)

_ser_schema_serialization_types = (
    *get_args(core_schema.ExpectedSerializationTypes),
    'function-plain',
    'function-wrap',
    'format',
    'to-string',
    'function',
    'model',
)

# if the serialization schema is of type 'include-exclude-dict' or 'include-exclude-set',
# we need to wrap it in a serializer function schema
# otherwise, we can leave it as is (which is more efficient)
serialization = schema.get('serialization')
if serialization:
    if serialization['type'] in ('include-exclude-dict', 'include-exclude-set'):
        serialization = core_schema.wrap_serializer_function_ser_schema(
            function=lambda v, h: h(v), schema=schema
        )
    else:
        assert serialization['type'] in _ser_schema_serialization_types

info_arg = _inspect_validator(self.func, 'plain')
if info_arg:
    func = cast(core_schema.WithInfoValidatorFunction, self.func)
    return core_schema.with_info_plain_validator_function(
        func,
        field_name=handler.field_name,
        serialization=serialization,  # type: ignore
    )
else:
    func = cast(core_schema.NoInfoValidatorFunction, self.func)
    return core_schema.no_info_plain_validator_function(func, serialization=serialization)  # type: ignore

Or something along those lines. @Anvil, could you please implement that change + we can get this across the line today?

Just as a side note - this approach still feels a bit unclean to me, so I'm going to mark this as something to come back to and see if we can clean up.

[sydney-runkle]

Alright, after discussing a bit more with @adriangb, I think that my above suggestion is ugly enough that we shouldn't use that alternative.

@Anvil, let's go with what you currently have, and I'll create an issue to address the performance of the

serialization = core_schema.wrap_serializer_function_ser_schema(function=lambda v, h: h(v), schema=schema)

line that you've added. I'd rather have the bug fix now and improve the performance down the line.

Thanks for all of your help on this!