match_like_matches_macro: strip refs in suggestion

[matthiaskrgr]

fixes #6503

changelog: match_like_matches_macro: strip refs in suggestion (#6503)

[rust-highfive]

r? @ebroto

(rust-highfive has picked a reviewer for you, use r? to override)

[alex-700]

@matthiaskrgr thanks for PR! But this implementation is not accurate enough. It should also think about a using the variable after matches! macro. For example the following code does not compile and using reference here is meaningful.

enum A { X, Y }

fn foo(x: A) {}

fn main() {
    let z = A::X;
    let _ = matches!(z, A::Y);
    foo(z);
}

[ebroto]

@alex-700 makes a good point, although I was a bit confused at the beginning by the example because A is a variable binding and not lowercase. Since the scrutinee is not a reference anymore and the pattern is not a reference pattern, the binding moves the value. The same case slightly modified:

struct S(i32);

fn fun(_val: Option<S>) {}

fn main() {
    let val = Some(S(42));
    
//    let _res = match &val {
//        Some(_a) => true,
//        _ => false,
//    };

    let _res = matches!(val, Some(_a));
    fun(val);
}

(playground)

---

The current implementation would also fail if the match uses a reference pattern, so match ergonomics does not apply. Consider the following case:

struct S(i32);

fn fun(_val: Option<S>) {}

fn main() {
    let val = Some(S(42));
    
//    let _res = match &val {
//        &Some(ref _a) => true,
//        _ => false,
//    };

    let _res = matches!(val, &Some(ref _a));
    fun(val);
}

(playground)

This would yield a mismatched types error

error[E0308]: mismatched types
  --> src/main.rs:11:30
   |
11 |     let _res = matches!(val, &Some(ref _a));
   |                         ---  ^^^^^^^^^^^^^ expected enum `Option`, found reference
   |                         |
   |                         this expression has type `Option<S>`
   |
   = note:   expected enum `Option<S>`
           found reference `&_`

[matthiaskrgr]

I think I got all the options covered now..?
There's still some cases where we could remove a & but the suggestion will still compile.

[ebroto]

r? @llogiq (I'm leaving the team, so I'm reassigning my PRs to other active members)

[llogiq]

From my point of view, this looks like a very solid improvement, also the tests seem to cover all cases. Thank you!

@bors r+

[bors]

📌 Commit 39f39d5 has been approved by llogiq

[bors]

⌛ Testing commit 39f39d5 with merge a982ab4...

[bors]

☀️ Test successful - checks-action_dev_test, checks-action_remark_test, checks-action_test
Approved by: llogiq
Pushing a982ab4 to master...