Inject[] does not let the interface show up.

[mozesa]

Hello @alecthomas,

first of all let me to say Thank you! for this outstanding package.

I try to utilize the feature of PEP 593 which actually works fine, but let speak the pictures.

As you can see, when I use Inject[], the PyCharm doesn't show the interface's methods.

But it works when I use the Annotated[interface: Type[T], ...].

Thanks for your help in advance!

[jstasiak]

Hey @mozesa. I'll answer in @alecthomas's stead as I maintain this project right now – I think you need to take this up with the PyCharm folks, I don't use it and I'm afraid I can't help much here, everything seems good on our side.

[mozesa]

Hello @jstasiak,

Thanks for your reply.
I will create a ticket on PyCharm's side.

I keep open this issue and later will update.

[jstasiak]

Sure thing!

[mozesa]

issue created.

[JosXa]

Yes, the JetBrains folks unfortunately lag very much behind updating their type checker. I could go on another rant, but I'll be civil and leave it at that.

The problem is that the Annotated feature has not been created with "regular" __class_getitem__ classmethods, but instead another one of these _SpecialForms. I just spent 30 minutes trying to come up with a workaround, but I couldn't come up with anything better than def PyCharmCompatibleInject(typ: T) -> T: ... which could somehow patch that missing behavior, but it looks ugly as hell: def foo(bar: PChInject(Bar)). meh.

[halbow]

Hi,

I'm still having this issue with Pycharm (mypy works fine) with the Annotation Alias Inject[Something].
A workaround for me was to use Annotated[Something, _inject_marker] (using the Annotated keyword directly instead of the annotation aliasInject).

@jstasiak as it seems Pycharm will not fix this issue soon, would you consider making _inject_marker public ? That would allow to have a proper workaround for tools not supporting the alias from PEP-0593 yet

[davidparsson]

would you consider making _inject_marker public ?

I think that is a reasonable request, allowing a workaround for those that think it's worth it. We should make the _noinject_marker public at the same time, and ideally update the documentation.

[davidparsson]

I have opened a PR for this in #269.