object.__setattr__ triggering PLC2801 outside of dunder methods

[my1e5]

Related to false positive on PLC2801 "unnecessary dunder... for setattr" #9584

object.__setattr__ is the recommended way to set attributes on frozen dataclasses. This was addressed in #9584 but PLC2801 is triggering when object.__setattr__ is in non-dunder methods.

See - https://play.ruff.rs/1a9602da-78ce-44c9-bbfe-b17b147da04b

from dataclasses import dataclass, field


@dataclass(frozen=True)
class Foo:
    a: int
    b: int
    c: int = field(init=False)

    def __post_init__(self):
        object.__setattr__(self, "c", self.a + self.b)

    def foo(self):
        object.__setattr__(self, "c", self.a - self.b)    # <-------

    def __str__(self):
        object.__setattr__(self, "c", self.a * self.b) 

14:9: PLC2801 Unnecessary dunder call to `__setattr__`. Mutate attribute directly or use setattr built-in function.

ruff v0.4.6

[AlexWaygood]

Thanks for the report!

I'm curious, what's the purpose of mutating attributes on a frozen dataclass from arbitrary non-dunder methods? Wouldn't the idiomatic thing here be to make it a regular (not frozen) dataclass?

[AlexWaygood]

Having said that, we may want to consider relaxing the rules here anyway. object.__setattr__ can be useful in various other situations (though it's quite niche) where you want to make sure that you're avoiding triggering custom logic in descriptor or __setattr__ methods

[my1e5]

I'm curious, what's the purpose of mutating attributes on a frozen dataclass from arbitrary non-dunder methods?

So my code looks something like this

def __post_init__(self):
    self._init_foo()
    self._init_bar()
    self._init_baz()

Basically just breaking up the __post_init__ step for readability.