dataclass decorator: unexpected keyword argument using mypy

[BayerSe]

Bug

Output of python -c "import pydantic.utils; print(pydantic.utils.version_info())":

             pydantic version: 1.9.0
            pydantic compiled: True
                 install path: /home/bay2rng/PycharmProjects/semisched/packages/scheduler/.venv/lib/python3.9/site-packages/pydantic
               python version: 3.9.9 (main, Nov 16 2021, 03:08:02)  [GCC 9.3.0]
                     platform: Linux-5.4.0-90-generic-x86_64-with-glibc2.31
     optional deps. installed: ['typing-extensions']

While translating some models implemented using Python's standard dataclass decorator into Pydantic dataclasses, mypy 0.931 started complaining about unexpected keyword arguments

from dataclasses import dataclass
from pydantic.dataclasses import dataclass as pd_dataclass


@dataclass
class ModelA:
    parameter: int


@pd_dataclass
class ModelB:
    parameter: int


a = ModelA(parameter=1)
b = ModelB(parameter=1)

Output of mypy test.py

test.py:16: error: Unexpected keyword argument "parameter" for "ModelB"

[pecigonzalo]

It seems the return type has some problem, and likely affecting VSCode auto-completion as well which I believe was meant to be fixed in #2721.

If you hover "b" in this example, you will see its type is detected as (variable) b: Dataclass.

[ajkaijanaho]

The dataclass decorator is annotated as returning a Dataclass object, which defines the Pydantic dataclass additions, losing any information about the original class that is being decorated. The dataclass transformation defined in #2721 is only applied to ModelMetaclass, which is not the metaclass of dataclasses.

I know of no way to annotate correctly a decorator that modifies a class's methods and fields. You either do what Pydantic does or you annotate like your decorator does not change the class.

[adriangb]

This works on both v1.10.X and v2 without issues even with mypy --strict, so it seems like it's been resolved.