Feature: reveal_type should work from comments

[ramalho]

It would be great if reveal_type could be called from comments.

Mypy already looks at various comments to do its job, and reveal_type is not a runtime function, it only exists during a type check.

While experimenting with type hints in test code, I could write this:

a = some_func() # reveal_type(a)

Instead of:

a = some_func()
if typing.TYPE_CHECKING:
    reveal_type(a)

[JukkaL]

I'm not sure if this would be a net improvement. Now it's possible to comment out uses of reveal_type to temporarily silence them. What do others think?

Also, this may be tricky to support since the parser we use drops most comments, I think.

[ramalho]

Thanks for you response and your work on Mypy, @JukkaL.

The syntax for the comment can be different, to accommodate the current use of commenting out uses of reveal_type, and make it easier for the parser. Maybe this:

a = some_func() # type: reveal_type(a)

Before focusing on the syntax of the comment, what do people think of the idea of having a way to put calls to reveal_type in test suites in a way that does not interfere with running the tests, and also doesn't require the cumbersome if TYPE_CHECKING guards?

[msullivan]

It doesn't seem to me that it would be enough less cumbersome than `if TYPE_CHECKING` to be worth having and documenting as an extra feature

…

On Fri, May 22, 2020 at 11:35 AM Luciano Ramalho ***@***.***> wrote: Thanks for you response and your work on Mypy, @JukkaL <https://github.com/JukkaL>. The syntax for the comment can be different, to accommodate the current use of commenting out uses of reveal_type, and make it easier for the parser. Maybe this: a = some_func() # type: reveal_type(a) Before focusing on the syntax of the comment, what do people think of the idea of having a way to put calls to reveal_type in test suites in a way that does not interfere with running the tests, and also doesn't require the cumbersome if TYPE_CHECKING guards? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#8864 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACTC7N6A2XJ2LGNWBP6VODRS3AVHANCNFSM4NGJPS2Q> .

[ramalho]

reveal_type() is a great debugging and learning feature.

My use case is as a teacher and a writer: I have code that is tested with pytest, and is sprinkled with reveal_type to demonstrate Mypy's type inference. Perhaps having if TYPE_CHECKING guards is not too bad in normal Mypy usage, but in my examples it does add noise.

I am biased, but I do believe that features that help with learning should be carefully considered, even if they don't seem useful to those with a lot of experience with the tool.

[AlexWaygood]

@ramalho, does the recent addition of typing.reveal_type to the stdlib work for your use case? (It will be new in 3.11, but has also been added to the typing_extensions backports package.)

[ramalho]

No, it does not solve my use case. Please consider my initial feature request and my responses to the other comments. Thanks!

[JelleZijlstra]

The use case seems too marginal to justify the complexity, especially now that we have the runtime reveal_type() available so you shouldn't need if TYPE_CHECKING:.

[ramalho]

I agree @JelleZijlstra. I did not know that reveal_type() would no longer require the if. Thanks!