Cannot generate Avro schema when using Generic Classes

[maver1ck]

Describe the bug
I'm trying to generate Avro Schema for model with Generic Class.

As a result I'm getting following error:

ValueError: Type ~T for field additonal_info is unknown. Please check the valid types at https://marcosschroh.github.io/dataclasses-avroschema/fields_specification/#avro-field-and-python-types-summary

To Reproduce

from typing import TypeVar, Generic
from dataclasses_avroschema import AvroModel
import pydantic


T = TypeVar('T')

@pydantic.dataclasses.dataclass
class UserInfoGeneric(Generic[T], AvroModel):
    name: str
    age: int
    additonal_info: T

@pydantic.dataclasses.dataclass
class EmailInfoGeneric(AvroModel):
    email: str
    is_verified: bool

@pydantic.dataclasses.dataclass
class UserEmailInfo(UserInfoGeneric[EmailInfoGeneric], AvroModel):
    pass

print(UserEmailInfo.fake())

print(UserEmailInfo.avro_schema())

Expected behavior

I'd like to see behaviour similar to example where I'm not using Generic classes.

from typing import TypeVar, Generic
from dataclasses_avroschema import AvroModel
import pydantic

@pydantic.dataclasses.dataclass
class EmailInfo(AvroModel):
    email: str
    is_verified: bool

@pydantic.dataclasses.dataclass
class AddressInfo(AvroModel):
    street: str
    city: str
    state: str
    zip: int

@pydantic.dataclasses.dataclass
class UserInfo(AvroModel):
    name: str
    age: int
    additonal_info: AddressInfo | EmailInfo

print(UserInfo.fake())

print(UserInfo.avro_schema())

[marcosschroh]

What you are trying to do it will never work as one type is a generic one. In avro schemas there are not generics. I will see if I can find a python trick to solve it. PR is welcome

[marcosschroh]

This feature seems quite complicated to achieve. It is related to a typing issue python/typing#629

[Robbie-Palmer]

@marcosschroh if using pydantic it's readily available through the __pydantic_generic_metadata__ field: Docs
e.g.

class MyModel(BaseModel, Generic[T]):
    def operate_on_assigned_type(self) -> T:
        generic_type_class: T = self.__class__.__pydantic_generic_metadata__["args"][0]
        return generic_type_class.foo()

[marcosschroh]

That can be an option, but the original issue is related to dataclasses not pydantic. The pydantic solution is a sort of hack to accomplish this, which it is ok but ideally it should come from the typing module so It will work with all flavors (dataclasses, pydantic and faust.