bug: cannot instantiate PEP 593 type alias

[adriangb]

from dataclasses import dataclass
from typing import Annotated
from pydantic import BaseModel

@dataclass
class DataclassModel:
    foo: str = "123"

DataclassAlias = Annotated[DataclassModel, "bar baz"]

DataclassAlias()  # works

class PydanticModel(BaseModel):
    foo: str = "123"

PydanticAlias = Annotated[PydanticModel, "bar baz"]

PydanticAlias()  # error
#   File ".../typing.py", line 680, in __call__
#     result.__orig_class__ = self
#   File "pydantic/main.py", line 347, in pydantic.main.BaseModel.__setattr__
# ValueError: "PydanticModel" object has no field "__orig_class__"

[MichaelTiemannOSC]

I just hit something throwing the same error. Pydantic version 1.8.2 (because 1.9.0 doesn't work for other reasons).

[iodsfx3d]

I'm getting a similar error while using generics

T = TypeVar('T', bound=BaseModel)

class Paginated(Generic[T], BaseModel):
    total: int
    total_pages: int
    page: int
    per_page: int
    items: List[T]

and I'm doing something like this when it throws the error

Paginated[PydenticModelSchema](
        total=qs._pagination.total_results,
        total_pages=qs._pagination.num_pages,
        page=qs._pagination.page_number,
        per_page=qs._pagination.page_size,
        items=qs.results,
    )

but it works if I remove the generic type and just use List[Any]

class Paginated(BaseModel):
    total: int
    total_pages: int
    page: int
    per_page: int
    items: List[Any]

Is it because of the same bug? I couldn't find anything related to this online

[PrettyWood]

@iodsfx3d you should use GenericModel like explained in the doc

from pydantic.generics import GenericModel
...

T = TypeVar('T', bound=BaseModel)

class Paginated(GenericModel, Generic[T]):
    total: T


Paginated[str](total='qwe')

[PrettyWood]

@adriangb Yep indeed! Annotated seems to add a dunder attribute which can't be set because of our custom __setattr__. I'll make a quick fix

[adriangb]

Awesome, thank you!