`self` argument is not assignable to variable typed as `Self`

[DetachHead]

from typing import Self

class Foo:
    def foo(self) -> None:
        asdf: Self = self

main.py:5: error: Incompatible types in assignment (expression has type "Foo", variable has type "Self")  [assignment]
Found 1 error in 1 file (checked 1 source file)

playground

[JamesParrott]

I agree this could be improved. FYI, typing self: Self works:

playground

[brianschubert]

I took a stab at making self have an implicit type of Self instead of Foo to see what would happen (for posterity sake, the relevant logic is here).

It turns out there's a test against that:

mypy/test-data/unit/check-selftype.test

Lines 1815 to 1830 in f6520c8

	[case testTypingSelfAttrOldVsNewStyle]
	from typing import Self, TypeVar
	T = TypeVar("T", bound='C')
	class C:
	x: Self
	def foo(self: T) -> T:
	return self.x
	def bar(self: T) -> T:
	self.x = self
	return self
	def baz(self: Self) -> None:
	self.x = self
	def bad(self) -> None:
	# This is unfortunate, but required by PEP 484
	self.x = self # E: Incompatible types in assignment (expression has type "C", variable has type "Self")

I'm not sure I understand the comment about this being a PEP 484 requirement. It seems that pyright gets away making the type of an unannotated self compatible/identical to Self: Pyright playground

[hauntsaninja]

There's a sentence in PEP 484 about not needing to annotate self and treat it as the type of the containing class. Looks like that sentence is in the current spec too and we should maybe change that: https://typing.readthedocs.io/en/latest/spec/annotations.html#annotating-instance-and-class-methods

I don't think that's too much of an issue though, I seem to remember folks were concerned about performance. Let me dig up a reference...

[hauntsaninja]

Yeah, see #14075 (closing this as a duplicate of that)

[hauntsaninja]

Let's also change the spec to be clearer about this python/typing#1860