Method __class_getitem__

pydantic/main.py:904–969

(
        cls, typevar_values: type[Any] | tuple[type[Any], ...]
    )

        """

    def __class_getitem__(
        cls, typevar_values: type[Any] | tuple[type[Any], ...]
    ) -> type[BaseModel] | _forward_ref.PydanticRecursiveRef:
        cached = _generics.get_cached_generic_type_early(cls, typevar_values)
        if cached is not None:
            return cached

        if cls is BaseModel:
            raise TypeError('Type parameters should be placed on typing.Generic, not BaseModel')
        if not hasattr(cls, '__parameters__'):
            raise TypeError(f'{cls} cannot be parametrized because it does not inherit from typing.Generic')
        if not cls.__pydantic_generic_metadata__['parameters'] and Generic not in cls.__bases__:
            raise TypeError(f'{cls} is not a generic class')

        if not isinstance(typevar_values, tuple):
            typevar_values = (typevar_values,)

        # For a model `class Model[T, U, V = int](BaseModel): ...` parametrized with `(str, bool)`,
        # this gives us `{T: str, U: bool, V: int}`:
        typevars_map = _generics.map_generic_model_arguments(cls, typevar_values)
        # We also update the provided args to use defaults values (`(str, bool)` becomes `(str, bool, int)`):
        typevar_values = tuple(v for v in typevars_map.values())

        if _utils.all_identical(typevars_map.keys(), typevars_map.values()) and typevars_map:
            submodel = cls  # if arguments are equal to parameters it's the same object
            _generics.set_cached_generic_type(cls, typevar_values, submodel)
        else:
            parent_args = cls.__pydantic_generic_metadata__['args']
            if not parent_args:
                args = typevar_values
            else:
                args = tuple(_generics.replace_types(arg, typevars_map) for arg in parent_args)

            origin = cls.__pydantic_generic_metadata__['origin'] or cls
            model_name = origin.model_parametrized_name(args)
            params = tuple(
                dict.fromkeys(_generics.iter_contained_typevars(typevars_map.values()))
            )  # use dict as ordered set

            with _generics.generic_recursion_self_type(origin, args) as maybe_self_type:
                cached = _generics.get_cached_generic_type_late(cls, typevar_values, origin, args)
                if cached is not None:
                    return cached

                if maybe_self_type is not None:
                    return maybe_self_type

                # Attempt to rebuild the origin in case new types have been defined
                try:
                    # depth 2 gets you above this __class_getitem__ call.
                    # Note that we explicitly provide the parent ns, otherwise
                    # `model_rebuild` will use the parent ns no matter if it is the ns of a module.
                    # We don't want this here, as this has unexpected effects when a model
                    # is being parametrized during a forward annotation evaluation.
                    parent_ns = _typing_extra.parent_frame_namespace(parent_depth=2) or {}
                    origin.model_rebuild(_types_namespace=parent_ns)
                except PydanticUndefinedAnnotation:
                    # It's okay if it fails, it just means there are still undefined types