dataclasses.dataclass, custom __init__ and dataclasses.replace false positive.

[squeaky-pl]

Bug Report

I am not sure if this is a mypy bug or by design or maybe a missing feature, but at least looking at the runtime behavior it seems like a bug to me. I asked on gitter and one user responed that it seems to be a bug/oversight in the implementation.

I want to do converters with dataclasses, and the way to do this with dataclasses is to provide custom __init__ i.e. consider the following example.

To Reproduce

Playground link

import dataclasses

@dataclasses.dataclass
class WrapsInt:
   value: int
   
   def __init__(self, value: int | str):
       self.value = int(value)

wraps_int = WrapsInt("3")
print(repr(wraps_int.value))

wraps_int2 = dataclasses.replace(wraps_int, value="4")
print(repr(wraps_int2.value))

At runtime it will print and boths will be integers:

3
4

This is because dataclasses.replace always calls init. But mypy seems to take the type information from the class body only instead when analyzing dataclasses.replace call.

Expected Behavior

No error, mypy recognizes __init__ implementation when analyzing dataclass.replace calls.

Actual Behavior

main.py:16: error: Argument "value" to "replace" of "WrapsInt" has incompatible type "str"; expected "int"  [arg-type]
Found 1 error in 1 file (checked 1 source file)

Your Environment

• Mypy version used: 1.15.0, master branch
• Mypy command-line flags: none
• Mypy configuration options from mypy.ini (and other config files): default
• Python version used: 3.12