Wrong warning for multiple type parameters if argument passed is a variable

[pksadiq]

Bug Report
When the function parameter is annotated to have some combination of multiple types with |, mypy validation warns
if a variable is passed to the called function.

To Reproduce

from typing import Any

def show(data: dict[int | str, dict[str, str]]):
    print(data)

data = {0: {"message": "Hello world"}}

# Warning
show(data)

# No warning
show({0: {"message": "Hello world"}})

Gist url: https://mypy-play.net/?mypy=latest&python=3.12&gist=2e3c8b121347ac9547e8e84eba7f1017

Expected Behavior

There should be no warning as it's a valid code

Actual Behavior

mypy ./test.py 
test.py:10: error: Argument 1 to "show" has incompatible type "dict[int, dict[str, str]]"; expected "dict[int | str, dict[str, str]]"  [arg-type]
Found 1 error in 1 file (checked 1 source file)

Your Environment

• Mypy version used: 1.13.0 (compiled: yes) as packaged in Debian sid
• Mypy command-line flags: None
• Mypy configuration options from mypy.ini (and other config files): None
• Python version used: 3.12.8

[brianschubert]

Hi! This is https://mypy.readthedocs.io/en/stable/common_issues.html#invariance-vs-covariance

Like many mutable generic containers, dict is invariant with respect to its type parameters. That means that even though int is a subtype of int | str, dict[int, V] is not a subtype of dict[int | str, V]. To see why that would be unsafe, consider this example:

def foo(x: dict[int | str, str]):
    x["a"] = "bad"

d: dict[int, str] = {1: "good"}
foo(d)

for key in d:
    assert isinstance(key, int)  # boom!

The reason the second example works is because bidirectional inference allows mypy to infer a better type for the dict literal by peeking at the type of the parameter that its being passed to. To make the first example type check, you can try adding an explicit type annotation for data or modifying the signature of show to accept a type that's compatible with dict[int, dict[str, str]].

[pksadiq]

I see. I can't modify the function signature. I shall annotate my variable instead.

Thanks for the explanation.