reveal_locals is hella inconsistent / top level variables not listed / narrowed type inconsistent with reveal_type

[KotlinIsland]

a: int
b = 1
if isinstance(a, bool):
    reveal_locals()  # "b: int", I would expect "a: bool, b: int"
    reveal_type(a)  # "a: bool" ✔
    
def foo(a: int) -> None:
    b: int
    reveal_locals()  # "a: int, b: int", this is correct, but inconsistent with top level
    reveal_type(a)

a: int = 1
reveal_locals() # nothing 😥

[erictraut]

I would not expect reveal_locals to emit the narrowed types of the locals, whereas I would expect reveal_type to take narrowing into account. This appears to be what mypy is doing. This is consistent with pyright, BTW.

It looks like reveal_locals is not currently displaying locals with an inferred type (such as b in your example above), which looks like a bug.

[JonathonReinhart]

I was also surprised to see the difference between reveal_type and reveal_locals with regard to type narrowing.

def test() -> None:
    foo = None
    if foo is None:
        foo = "never None"

    reveal_locals()     # Revealed local types are:
                        #    foo: Union[builtins.str, None]

    reveal_type(foo)    # Revealed type is 'builtins.str'

I would not expect reveal_locals to emit the narrowed types of the locals, whereas I would expect reveal_type to take narrowing into account.

Interesting. Why do you feel that this is the correct behavior?

If this is working as intended and not a bug, then perhaps the docs for reveal_locals() should be updated to explain this.