Extra dataclass_transform features

typing.dataclass_transform is a bit strange. It is a compromise to fulfill the urgent need to support dataclass-like objects (mostly in pydantic and attrs) to some extent without adding a generic and flexible mechanism for defining your own dataclass_transform.

If your use case deviates from what dataclass_transform explicitly supports, it can be difficult or impossible to work around that.

Luckily, dataclass_transform accepts arbitrary keyword arguments at runtime to allow type checkers to add their own little hacks on top of the standard ones.

basedpyright currently supports these options:

• skip_replace: setting this to True disables the synthesis of the __replace__ method.

  Pyright (and basedpyright) assumes that classes produced by a dataclass transform define a __replace__ method, as long as the class is not marked with init=False and does not define a custom __init__ method. However, typing.dataclass_transform doesn't require that the __replace__ method is defined.

  In addition, __replace__ messes with the variance inference of frozen dataclasses (see this discourse thread for details). Here's a recipe you can use to work around this:

  from typing import dataclass_transform
  
  @dataclass_transform(skip_replace=True, frozen_default=True)
  def frozen[T: type](t: T) -> T:
      return dataclass(frozen=True, slots=True)(t)
  
  # check that this enables covariance:
  @frozen
  class Box[T]:
      value: T
  
  box1: Box[str] = Box("test")
  box2: Box[str | int] = box1
  

All of these options require the enableBasedFeatures configuration option to be set to true.