Support injecting via generics #6
Conversation
|
I pushed a major update, the mechanism is now more or less at a point where I wanted to get. This is ready for a more detailed review. |
src/sciline/container.py
Outdated
| return wrapper | ||
| def _bind_free_typevars(tp: type, bound: Dict[TypeVar, type]) -> type: | ||
| if isinstance(tp, TypeVar): | ||
| return bound[tp] |
There was a problem hiding this comment.
Shouldn't it also be bound.get(tp, tp) in case it's not bound to anything, or is that not possible at all?
There was a problem hiding this comment.
I want it to fail in this case, so __getitem__ works (to get KeyError). I could use what you suggest, to get an UnsatisfiedRequirement elsewhere. Not sure which is better? Or maybe we should have another specific exception like UnboundTypeVar?
There was a problem hiding this comment.
I think raising UnboundTypeVar makes more sense because that's what you'll get from static type checker also.
There was a problem hiding this comment.
Then I think it shouldn't allow TypeVar as a return annotation at all, since it will not be injected to anyone, right....?
| args = get_args(key) | ||
| if args in subproviders: | ||
| raise ValueError(f'Provider for {key} already exists') | ||
| subproviders[args] = provider |
There was a problem hiding this comment.
Should we forbid this kind of nested generic arguments...?
VT = TypeVar("VT")
Bike = NewType("Bike", object)
Car = NewType("Car", object)
class Time(sl.Scope[VT], float): ...
def provide_bike_time() -> Time[Bike]:
return Time(1)
def provide_car_time() -> Time[Car]:
return Time(0.1)
def provide_list_of_time(bike_time: Time[VT]) -> List[Time[VT]]
return [bike_time]
It should be
TimeT = TypeVar("TimeT")
def provide_list_of_time(bike_time: TimeT) -> List[TimeT]
return [bike_time]
And then we can compute nested generic type like; List[Time[Car]] or List[Time[Bike]].
There was a problem hiding this comment.
Yes, good point, will have to think about whether this will be required. I opened #9.
|
If there is a provider that loads file and if the file is replaced, we need to rebuild the dask delayed again. |
I don't think we need to do anything? |
|
Now I remembered that once the |
Yes, I am indeed not allowing this currently, to avoid such complications. |
This replaces #4.
Details in
Containerare a proof of concepts and likely still wrong in some cases. Will refactor and add tests after initial discussion.complex_workflow_test.pyprovides an example of how actual client code might look. Note the slightly verbose creation of parametrized/scoped domain types, but the simple/trivial way of using functions with generics in their type hints withContainer.