dataclass with custom function returning dataclass.Field

[PythonFZ]

I have a function similar to the following

import dataclasses

def field(
    *,
    my_data: str,
    **kwargs,
) -> dataclasses.Field:
    return dataclasses.field(**kwargs, metadata={"my_data": my_data})

and would like to use it as

@dataclasses.dataclass
class MyCls:
    data: float = field(my_data="my_data")

But with this code I get Type "Field[Unknown]" is not assignable to declared type "float".

The code above is only a utility as the full example would look something like this where I have the same issue:

import dataclasses
import znfields
from typing import dataclass_transform

def getter(self, name) -> str:
    return f"{name}:{self.__dict__[name]}"

def setter(self, name, value) -> None:
    if not isinstance(value, float):
        raise ValueError(f"Value {value} is not a float")
    self.__dict__[name] = value

def myfield():
    return znfields.field(getter=getter, setter=setter)

@dataclass_transform(field_specifiers=(myfield,))
@dataclasses.dataclass(kw_only=True)
class Node:

    def __init_subclass__(cls):
        return dataclasses.dataclass(cls, kw_only=True)
    

class MyNode(Node):
    x: float = myfield()

[erictraut]

The stdlib dataclasses.field symbol requires special-case handling in a type checker. If you define your own variant, a type checker doesn't know to apply this special-case handling logic.

One workaround is to leverage the typing.dataclass_transform mechanism in the type system. For details, refer to this documentation. This facility allows one to create classes that have dataclass-like semantics. That would look something like this:

from dataclasses import dataclass, field
from typing import Any, dataclass_transform

def my_field(*, my_data: str, **kwargs) -> Any:
    return field(**kwargs, metadata={"my_data": my_data})

@dataclass_transform(field_specifiers=(my_field,))
def my_dataclass[T](cls: type[T], /) -> type[T]:
    return dataclass(cls)

@my_dataclass
class MyCls:
    data: float = my_field(my_data="my_data")