Must_use and linear types

language design
21

I feel like the more the 2021 deadline approaches, the more everyone wants to fit everything into an edition change, but we should try to remember that edition-boundary changes aren't free.

I think the status quo works well enough for 90% of cases.

2 Likes
22

Given the discussion about let _ misusage, it does seem like that could become an antipattern, at which point drop ought to be the go-to replacement to quiet #[must_use] values. In the very specific case of Result, however, drop does not feel or read right: I don't like how drop(fs::remove_dir_all(...)) looks like.

I remember having added a custom extension trait on Result to have something akin to what @kornel mentioned:

fs::remove_dir_all(...).ignore_err();

and I do think (again, given how shotgunny let _ appears to be) that adding that to the std (core) lib wouldn't hurt:

impl<E> Result<(), E> {
    /// Succint, and yet conveys a very clear meaning, even to beginners.
    fn ignore_err (self) // or .ignored() % bikeshed
    {}
}
6 Likes
23

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.