Mypy typing errors with "reusable" model_validator

[vicchi]

Initial Checks

• I confirm that I'm using Pydantic V2

Description

I'm at the final stages of migrating a v1 code base to v2. Functionally it's complete, the code runs as expected without any issues (the v1 to v2 migration guide was really useful). But I'm getting mypy errors when using my attempt to migrate "reused" v1 root_validator instances to v2 model_validator instances.

When I use a decorated model_validator inline in the model definition all is well. The model validates as expected and mypy is content and happy.

When I use my attempt at a reusable model_validator the model still validates as expected but mypy is not happy at all.

The "inline" test case I've put together, plus the output from mypy is:

#!/usr/bin/env python
# pylint: disable=all

from typing import Any, Dict, Optional

from pydantic import BaseModel, ConfigDict, model_validator

def normalise_known_invalids(values: Dict[str, Any]) -> Dict[str, Any]:
    if values:
        for key, value in values.items():
            value = value.strip()
            if value:
                if value.lower() in ['', 'n/a', '-']:
                    values[key] = None
                else:
                    values[key] = value

            else:
                values[key] = None

    return values


class InlineValidatorModel(BaseModel):
    fld: Optional[str] = None

    @model_validator(mode='before')
    @classmethod
    def normaliser(cls, values: Dict[str, Any]) -> Dict[str, Any]:
        return normalise_known_invalids(values)

    model_config = ConfigDict(str_strip_whitespace=True, extra='forbid')

def main():
    model = InlineValidatorModel()
    print(f'empty inline model: {model}')

    model = InlineValidatorModel(fld='OK')
    print(f'valid input inline model: {model}')

    model = InlineValidatorModel(fld='n/a ')
    print(f'invalid input inline model: {model}')

if __name__ == '__main__':
    main()

$ mypy model_test.py
Success: no issues found in 1 source file

But the mypy output from the "reused" variant, code in the example below, is:

$ mypy model_test2.py
model_test2.py:29: error: Argument 1 has incompatible type "Callable[[dict[str, Any]], dict[str, Any]]"; expected "ModelBeforeValidator | ModelBeforeValidatorWithoutInfo"  [arg-type]
Found 1 error in 1 file (checked 1 source file)

Also, this is possibly related to #6693.

Example Code

#!/usr/bin/env python
# pylint: disable=all

from typing import Any, Dict, Optional

from pydantic import BaseModel, ConfigDict, model_validator

def normalise_known_invalids(values: Dict[str, Any]) -> Dict[str, Any]:
    if values:
        for key, value in values.items():
            value = value.strip()
            if value:
                if value.lower() in ['', 'n/a', '-']:
                    values[key] = None
                else:
                    values[key] = value

            else:
                values[key] = None

    return values

class ReusedValidatorModel(BaseModel):
    fld: Optional[str] = None

    _normaliser = model_validator(mode='before')(normalise_known_invalids)
    model_config = ConfigDict(str_strip_whitespace=True, extra='forbid')


def main():
    model = ReusedValidatorModel()
    print(f'empty reused model: {model}')

    model = ReusedValidatorModel(fld='OK')
    print(f'valid input reused model: {model}')

    model = ReusedValidatorModel(fld='n/a ')
    print(f'invalid inputreused  model: {model}')

if __name__ == '__main__':
    main()

Python, Pydantic & OS Version

pydantic version: 2.5.2
        pydantic-core version: 2.14.5
          pydantic-core build: profile=release pgo=true
                 install path: /home/gary/scratch/.direnv/python-3.10.12/lib/python3.10/site-packages/pydantic
               python version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0]
                     platform: Linux-6.2.0-37-generic-x86_64-with-glibc2.35
             related packages: mypy-1.7.1 pydantic-settings-2.0.3 typing_extensions-4.8.0

[sydney-runkle]

@vicchi,

Good call. This is something we can fix pretty easily! I'll have a PR up shortly.