Pyright/Pylance typing error when overriding __pydantic_extra__ type

[zEdS15B3GCwq]

Initial Checks

• I confirm that I'm using Pydantic V2

Description

Similar to #9147 and #7094, but those seem to be focused on TypeAdapter and I'm unsure how closely those are related. Sorry if this is a duplicate.

See code below. Pylance warns that __pydantic_extra__ should also be specified when instantiating the class. I believe this warning should not appear.

Example Code

from pydantic import BaseModel, ConfigDict


class Model(BaseModel):
    __pydantic_extra__: dict[str, int]

    a: int

    model_config = ConfigDict(extra="allow")


m = Model(a=1) # Argument missing for parameter "__pydantic_extra__" Pylance (reportCallIssue)

Python, Pydantic & OS Version

pydantic version: 2.6.4
        pydantic-core version: 2.16.3
          pydantic-core build: profile=release pgo=true
                 install path: C:\Test\.venv\Lib\site-packages\pydantic
               python version: 3.12.2 (tags/v3.12.2:6abddd9, Feb  6 2024, 21:26:36) [MSC v.1937 64 bit (AMD64)]
                     platform: Windows-11-10.0.22631-SP0
             related packages: typing_extensions-4.11.0
                       commit: unknown

[sydney-runkle]

@zEdS15B3GCwq,

Thanks for reporting this! Yep, looks like a bug. Great first issue, for anyone interested!

[Viicos]

This is unfortunate, as dataclass_transform (which is used to make type checkers aware of BaseModel acting like a dataclass) does not provide any way to "ignore" fields from the synthesized __init__ method.

A workaround is to define your model as:

from pydantic import ConfigDict, Field


class MyModel(BaseModel):
    __pydantic_extra__: dict[str, int] = Field(init=False)

    a: int

    model_config = ConfigDict(extra="allow")

Ideally, this feature could have been implemented as a config value, e.g. ConfigDict(extra_type=int) or similar, but the current implementation (having to annotate __pydantic_extra__) makes it nice as you get type checking support as well.

[blueTurtz]

@sydney-runkle I can pick this up.

[sydney-runkle]

@Viicos,

What do you think about this approach here (we might need some docs changes):

from pydantic import BaseModel, ConfigDict
from pydantic.fields import PrivateAttr as _PrivateAttr


class Model(BaseModel):
    __pydantic_extra__: dict[str, int] = _PrivateAttr()

    a: int

    model_config = ConfigDict(extra="allow")


m = Model(a=1)

Thinking about this approach on the heels of #9293

[sydney-runkle]

Also, here's the relevant docs for reference: https://docs.pydantic.dev/latest/concepts/models/#extra-fields

[Viicos]

Looks good as a workaround, so the docs can probably be updated with the added PrivateAttr. #9293 already applied the change on the BaseModel.__pydantic_extra__ annotation :)

[blueTurtz]

#9187 (comment)
Using mypy I found that the following issue is raised when using the Example Code from the link above

(pydantic-env) ➜  tmp mypy 9187.py
9187.py:12: error: Missing named argument "__pydantic_extra__" for "Model"  [call-arg]
Found 1 error in 1 file (checked 1 source file)

[dmontagu]

When #9659 gets merged, I think the solution to this will be to do:

from pydantic import BaseModel, Field


class MyModel(BaseModel, extra='allow'):
    __pydantic_extra__: dict[str, int] = Field(init=False)

(or similar, with whatever type hint you want on the pydantic extra)