Rust: rewrite `...` to `..=` as an idiom lint for Rust 2018 edition

Created on 24 May 2018 · 10Comments · Source: rust-lang/rust

Current status

Implemented in https://github.com/rust-lang/rust/pull/51149 but currently not marked as automatically applicable due to this issue which is explained below

A-edition-2018-lints A-rust-2018-preview C-enhancement P-high T-compiler WG-epoch

Source

nikomatsakis

👍1

Most helpful comment

One remaining issue of a..=b pattern is #48501. When stabilizing a..=b, to avoid the confusing interpretation of &a...b, in the following example:

fn main() {
    match &5 {
        &3...6 => {}
        _ => {}
    }
}

if we just s/.../..=/g the code will error with

error: the range pattern here has ambiguous interpretation
 --> src/main.rs:3:10
  |
3 |         &3..=6 => {}
  |          ^^^^^ help: add parentheses to clarify the precedence: `(3 ..=6)`

Note that pattern_parentheses (#48500) is still unstable 🙃. However, given that default binding modes (#42640) is already stable, an even more idiomatic fix would probably be

fn main() {
    match &5 {
        3..=6 => {}   // no &, no (), no whatever
        _ => {}
    }
}

kennytm on 26 May 2018

❤1 👍1

All 10 comments

Do we have an existing lint for this?

nikomatsakis on 24 May 2018

If we hope to someday get rid of ... in patterns (the current description of #28237 describes it as "(silently) deprecated"), wouldn't we want a future-incompatible lint rather than an idiom lint? (... in expressions is already illegal.)

zackmdavis on 25 May 2018

❤1

One remaining issue of a..=b pattern is #48501. When stabilizing a..=b, to avoid the confusing interpretation of &a...b, in the following example:

fn main() {
    match &5 {
        &3...6 => {}
        _ => {}
    }
}

if we just s/.../..=/g the code will error with

error: the range pattern here has ambiguous interpretation
 --> src/main.rs:3:10
  |
3 |         &3..=6 => {}
  |          ^^^^^ help: add parentheses to clarify the precedence: `(3 ..=6)`

Note that pattern_parentheses (#48500) is still unstable 🙃. However, given that default binding modes (#42640) is already stable, an even more idiomatic fix would probably be

fn main() {
    match &5 {
        3..=6 => {}   // no &, no (), no whatever
        _ => {}
    }
}

kennytm on 26 May 2018

❤1 👍1

@zackmdavis

If we hope to someday get rid of ... in patterns (the current description of #28237 describes it as "(silently) deprecated"), wouldn't we want a future-incompatible lint rather than an idiom lint? (... in expressions is already illegal.)

I classified it as an idiom lint because the code will not cause a hard error in the new edition, even if it is frowned upon.

nikomatsakis on 30 May 2018

👍1

@zackmdavis if you've got a moment, can you write up a listing of what's left after #51149 landed?

alexcrichton on 11 Jul 2018

if you've got a moment, can you write up a listing of what's left after #51149 landed?

@alexcrichton: the comment above by @kennytm should still be addressed; the lint suggestion is marked as maybe-incorrect for this reason. If pattern-parentheses are stabilized (which itself reportedly needs more testing), using them for a correct suggestion when needed will be very easy.

zackmdavis on 12 Jul 2018

Ok great, thanks!

alexcrichton on 12 Jul 2018

visited for T-compiler triage. It seems like it was blocked on stabilization of the pattern-parentheses feature. So what are the next steps, and who wants to own this?