Higher Kinded Types Example Refactored

[deusebio]

Bug report

The typing inferred via higher-kinded types does not seem to be correct

What's wrong

I've started to play around with the Higher-Kinded Types (I'd really love to be able to type even with type classes) that are provided in the library, following the blog post here (By the way...terrific blog!). One (seemingly basic) thing that I tried to do, was to turn the apply_function into a class method of HasValue, rather than a stand-alone function. However, I've then stumbled into a number of issues.

If one simply puts the function into the class, making it a method, you'd end up with the following conflict:

The erased type of self "returns.primitives.hkt.KindN[_InstanceKind-1, _ValueType1, Any, Any]" is not a supertype of its class "test_mypy_working.HasValue[_ValueType`1]"

yes, right. The HasValue class does not inherit from KindN. Therefore, I have made HasValue extends the KindSupports, also avoiding to re-inherit this in the child classes Bag and Box .

class HasValue(SupportsKind1[_InstanceKind, _ValueType], Generic[_InstanceKind, _ValueType]):
    @abc.abstractmethod
    @property
    def value(self) -> _ValueType:
        """Returns a value property."""

    @abc.abstractmethod
    def with_value(
        self: _InstanceKind,
        new_value: _NewValueType,
    ) -> Kind1[_InstanceKind, _NewValueType]:
        """Creates a new instance with a changed value."""

    @kinded
    def apply_function(
            self: Kind1[_InstanceKind, _ValueType],
            callback: Callable[[_ValueType], _NewValueType],
    ) -> Kind1[_InstanceKind, _NewValueType]:
        new_value = callback(self.value)
        return self.with_value(new_value)


@dataclasses.dataclass
class Box(HasValue["Box", _ValueType]):
    value: _ValueType
    length: int
    width: int
    height: int

    def with_value(self, new_value: _NewValueType) -> "Box[_NewValueType]":
        return Box(new_value, self.length, self.width, self.height)


@dataclasses.dataclass
class Bag(HasValue["Bag", _ValueType]):
    value: _ValueType
    brand: str
    model: str

    def with_value(self: "Bag", new_value: _NewValueType) -> "Bag[_NewValueType]":
        return Bag(new_value, self.brand, self.model)

When doing this, however mypy complains with the following message:

error: Incompatible return value type (got "KindN[_ValueType, _NewValueType, Any, Any]", expected "KindN[_InstanceKind, _NewValueType, Any, Any]")

which I can't quite understand. How does it get the KindN[_ValueType, ...] from the self.with_value method that returns KindN[_InstanceKind, ...]

Despite the error, if in the script I place the following lines, mypy nevertheless INDEED DOES the correct inference

box = Box(value=10, length=1, width=2, height=3)
reveal_type(box.apply_function(str))

Revealed type is "hkt.Box[builtins.str]"

How is that should be

I suppose the error should not appear, unless I'm missing some points here. I'm still new to these features of this library and I'm still wrapping my head around it

Love dry-python? ❤️
Do you want your bugs to be solved faster?

Please consider supporting us:
👉 https://github.com/sponsors/dry-python

-->

System information

• python version: 3.7.6

• returns version: 0.19.0

• mypy version: 0.941

• mypy-extensions version: 0.4.3

• hypothesis version (if any):

• pytest version (if any):