use stc to improve deno lint

[KnorpelSenf]

deno lint is awesome because it's super fast. However, there are things I really wish it could do, but it can't. The prime example for this is no-floating promises.

function foo() { return Promise.resolve() }

foo() // forgot to use await!

The reason why Deno isn't able to lint such things properly is because it doesn't run tsc. Thus, there's no way to infer the return type of foo, so the promise is floating without a warning. Using tsc is not really an option due to the performance cost.

I suggest to use https://github.com/dudykr/stc which lets us do most of these things correctly. It isn't mature enough to replace tsc inside deno check IMO, but it could serve as the basis for much better linting as it would enable deno lint to access type information.

What's your opinion on this?

[dsherret]

stc is not being worked on anymore.

stc: Donny confirmed the article direction and mentioned no longer working on stc.

https://www.joshuakgoldberg.com/blog/rust-based-javascript-linters-fast-but-no-typed-linting-right-now/

[dsherret]

I also want no-floating-promises. It would be great if TypeScript would just add it microsoft/TypeScript#13376 (comment)

[KnorpelSenf]

Oh no. I was not aware.

Would that mean that Deno would start running tsc during lint?

[dsherret]

The TypeScript issue would probably mean another compiler option where it points out unawaited promises. So it would be pointed out with deno check.

[KnorpelSenf]

Smart!

[KnorpelSenf]

@dsherret have you considered adding opt-in type-based linting, such as deno lint --check or perhaps deno lint --types, which supports more rules at the expense of worse perf?

[bartlomieju]

@KnorpelSenf weve considered it and it's possible, but we don't have bandwidth to work on it now. Contributions are welcome though :)

[dsherret]

Contributions are welcome though :)

There should be agreement on something so large before it's implemented. It will be a lot of work and last time it was discussed Ry was against it.

[bartlomieju]

Contributions are welcome though :)

There should be agreement on something so large before it's implemented. It will be a lot of work.

Yes, for sure. A proposal would be a good contribution first.

[dsherret]

Btw, there is a list of type aware lints here: denoland/deno_lint#1138

[KnorpelSenf]

Thanks for explaining! I currently also have too much on my list to work on this, but it's good to know that we can push in that direction if someone finds the time.