Lint for using Option::as_deref

[jan-auer]

With Rust 1.40.0, Option::as_deref has landed.

It would be great to have a clippy lint that detects Option::as_ref chained by Option::map<F> where F: FnOnce(&T) -> &<T as Deref>::Target. This would then automatically account for cases where AsRef::as_ref or a custom method like String::as_str are used.

By far the most compelling case is to apply it to Option<String> or Option<PathBuf>, with probably lots of code bases now containing opt.as_ref().map(String::as_str) or similar constructs.

Example in:

let opt = Some(" hello ".to_owned());
let trimmed = opt.as_ref().map(String::as_str).map(str::trim);

Example out:

let opt = Some(" hello ".to_owned());
let trimmed = opt.as_deref().map(str::trim);

There are probably more cases that could be detected, like ? chaining or match statements, but I think those will get gradually less useful to be worth the effort.

Icing on the cake would be compatibility to rustfix :)

[jan-auer]

I have to say that I'm highly interested in attempting to implement this if someone agrees that this is useful. Realistically though, I might not get to it as I know myself, so I guess this is up for grabs.

It would be great to have a clippy lint that detects Option::as_ref chained by Option::map where F: FnOnce(&T) -> &::Target. This would then automatically account for cases where AsRef::as_ref or a custom method like String::as_str are used.

Aren't we going to get false positives if we check based on the signature of the closure? The code for as_def is self.as_ref().map(|t| t.deref()). IMO, we lint should only generate warnings for this code or special cases that are equivalent.

[jan-auer]

You’re right. Especially with closures, there could be false positives. I missed that.

I wonder if there is a nice way to avoid whitelisting all well-known cases...