Generic with dataclasses

[Andarius]

I have the following dataclass with a generic property and I'm not sure on how to go with this.

I have the data_type property which overlaps the PossibleTypes type such as

from typing import Union, Any, TypeVar, Generic
from dataclasses import dataclass

PossibleTypes = Union[str, int, float]


def validate_type(data_type: Any, value: str) -> PossibleTypes:
    if data_type is str:
        return str(value)
    elif data_type is int:
        return int(value)
    elif data_type is float:
        return float(value)
    else:
        raise NotImplementedError()


T = TypeVar('T', str, int, float)


@dataclass
class CommandOption(Generic[T]):
    data_type: T

    def validate(self, value: str) -> T:
        return validate_type(self.data_type, value)

But it raises:

  test-ignore.py:26:16 - error: Expression of type "PossibleTypes" cannot be assigned to return type "T@CommandOption"
    Type "PossibleTypes" cannot be assigned to type "T@CommandOption" (reportGeneralTypeIssues)

What is the correct way of typing either validate_type or CommandOption to avoid raising an error ?

Thanks !

[erictraut]

I'm not 100% sure I understand what you're trying to do here, but if my assumptions are correct, here's a potential solution:

from typing import TypeVar, Generic
from dataclasses import dataclass

T = TypeVar("T", str, int, float)

def validate_type(data_type: type[T], value: str) -> T:
    if data_type in T.__constraints__:
        return data_type(value)

    raise NotImplementedError()

@dataclass
class CommandOption(Generic[T]):
    data_type: type[T]

    def validate(self, value: str) -> T:
        return validate_type(self.data_type, value)

[Andarius]

Thanks @erictraut for the quick answer , that's exactly it !

[Andarius]

As a side note: is it possible to make it work with datetime.date ? (for T = TypeVar("T", str, int, float, datetime.date))

[erictraut]

It needs to be a bit more complicated if you want to support classes that require arguments for their constructors, but it's possible. Here's a variant that handles that.

T = TypeVar("T", str, int, float, date)

def validate_type(data_type: type[T], value: str) -> T:
    if issubclass(data_type, date):
        return data_type(2022, 1, 1)

    if issubclass(data_type, (str, int, float)):
        return data_type(value)

    raise NotImplementedError()