Use Literal for MatchSingleton

[cdce8p]

MatchSingleton can only ever be True, False, or None.
The current Optional[bool] is inaccurate for that as it also allows implicit booleans: 1 != 2.

https://docs.python.org/3.10/library/ast.html#ast.MatchSingleton

[srittau]

LGTM

[github-actions]

According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉

[Akuli]

So why should match_singleton.value = (foo != bar) be an error in code that manipulates AST?

[srittau]

It should basically not matter whether something is annotated bool or Literal[True, False], both should be treated as equivalent. (Although I think at least mypy is buggy in that regard.) So this change was a no-op (minus type checker bugs).

[Akuli]

I am assuming the current "buggy" mypy, where Literal[True, False] is a subtype of bool, and expressions generally result in type bool.

[srittau]

Here's the corresponding mypy issue: python/mypy#6113.

[srittau]

@Akuli Just to clarify: I'm not married to this change. If it causes problem we can roll this back. I was just assuming that the current implementation was causing problems for @cdce8p, so this would be better.

[cdce8p]

I had the issue the other way round. Got an error when I tried to assign Optional[bool] to Literal[True, False, None]. It seems this does indeed needs a fix in mypy. Another reason for Literal was that it's a bit more expressive especially for hints in IDEs.