TypedDict not recognised as valid dict when passed as parameter to a function

[berzi]

Bug Report

Mypy reports an error when a TypedDict is passed as an argument to a function that is annotated as accepting dicts more generally.

To Reproduce

class TestDict(TypedDict):
    a: int


def myfunc(test: dict):
    print(test)


myfunc(TestDict(a=2))

Note that:

>>> type(TestDict(a=2))
<class 'dict'>

Expected Behavior

No error.

Actual Behavior

error: Argument 1 to "myfunc" has incompatible type "TestDict"; expected "Dict[Any, Any]"  [arg-type]
    myfunc(TestDict(a=2))
           ^

Your Environment

• Mypy version used: 0.960 (compiled: yes)
• Python version used: 3.9.10

[ShaneNolan]

A TypedDict object is not a subtype of the regular dict[...] type (and vice versa), since dict allows arbitrary keys to be added and removed, unlike TypedDict. However, any TypedDict object is a subtype of (that is, compatible with) Mapping[str, object], since Mapping only provides read-only access to the dictionary items:

def print_typed_dict(obj: Mapping[str, object]) -> None:
    for key, value in obj.items():
        print('{}: {}'.format(key, value))

print_typed_dict(Movie(name='Toy Story', year=1995))  # OK

[1]

[berzi]

TypedDict can also be instantiated with total=False, which would effectively make its keys read-write. That would still not Make it 100% compatible with dict but it seems worth mentioning.

So does this mean I just have to change my typing of dict to Mapping? That's fair on one hand, but on the other i think this is an easy enough mistake to make. Maybe the message can be made more explicit about why it doesn't type-check.

Also, external libraries more often than not use dict rather than mapping, and i can't control their annotations. If i want to annotate my code to specify more exactly what a dict coming from such a library contains, and TypedDict is not correct, then what should i use?

[davidism]

I'm not sure the explanation makes sense. If the function can accept any dict with arbitrary keys, then a TypedDict with specific keys seems to satisfy that requirement. This also came up in pallets/flask#4695, TypedDict fails both as a return value for dict and as the signature of a decorated function.

[erictraut]

This isn't a bug in mypy. It's doing the correct thing here. PEP 589 is very clear that a TypedDict instance is not type compatible with dict and gives the reasons why. A TypedDict instance is compatible with Mapping[str, Any].

[JelleZijlstra]

Mypy is following the spec here, though I have some reservations about whether the spec got this right.