Higher Kinded Types Example Refactored
deusebio opened this issue · 0 comments
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, _ValueType
1, 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):