Don't show suggestion if slice pattern is not top-level

[long-long-float]

Close #120605

Don't show suggestion to add slicing ([..]) if the slice pattern is enclosed by struct like Struct { a: [] }.

For example, current rustc makes a suggestion as a comment. However, the pattern a: [] is wrong, not scrutinee &self.a.
In this case, the structure type a: Vec<Struct> and the pattern a: [] are different so I think the pattern should be fixed, not the scrutinee.
If the parent of the pattern that was the target of the error is a structure, I made the compiler not show a suggestion.

pub struct Struct {
    a: Vec<Struct>,
}

impl Struct {
    pub fn test(&self) {
        if let [Struct { a: [] }] = &self.a {
//             ^^^^^^^^^^^^^^^^^^   ------- help: consider slicing here: `&self.a[..]`
            println!("matches!")   
        }
    }
}

Note:

• I created PatInfo.history to store parent-child relationships for patterns, but this may be inefficient.
  • I use two fields parent_kind and current_kind instead of vec. It may not performance issue.
• Currently only looking at direct parents, but may need to look at deeper ancestry.

[rustbot]

r? @Nadrieril

rustbot has assigned @Nadrieril.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[Nadrieril]

Hi! Thanks for your contribution.

Let me know with @rustbot ready when you've found a solution that doesn't overflow or if you need help.

@rustbot author

[Nadrieril]

My first question is: the example you show gives a correct suggestion: to match on self.a with a slice pattern you really do need to use self.a[..] or self.a.as_slice(). This is different from #120605.

Could you add two tests: one with the example you mention, and one with the example in #120605 so we can see how your code handles them?

[long-long-float]

The reason why I use self.a without [..] is that the errors displayed are the same in the code with and without [..].
The following example shows that there is no suggestion for the error for test.rs:8:29.
Note that the error for test.rs:8:16 was added while resolving this issue (I think this error and suggestion is correct).

# self.a without [..]
-> % rustc +stage1 test.rs
error[E0529]: expected an array or slice, found `Vec<Struct>`
 --> test.rs:8:16
  |
8 |         if let [Struct { a: [] }] = &self.a {
  |                ^^^^^^^^^^^^^^^^^^   ------- help: consider slicing here: `&self.a[..]`
  |                |
  |                pattern cannot match with input type `Vec<Struct>`

error[E0529]: expected an array or slice, found `Vec<Struct>`
 --> test.rs:8:29
  |
8 |         if let [Struct { a: [] }] = &self.a {
  |                             ^^ pattern cannot match with input type `Vec<Struct>`

error: aborting due to 2 previous errors

For more information about this error, try `rustc --explain E0529`.

# self.a with [..]
-> % rustc +stage1 test.rs
error[E0529]: expected an array or slice, found `Vec<Struct>`
 --> test.rs:8:29
  |
8 |         if let [Struct { a: [] }] = &self.a[..] {
  |                             ^^ pattern cannot match with input type `Vec<Struct>`

error: aborting due to 1 previous error

For more information about this error, try `rustc --explain E0529`.

Could you add two tests: one with the example you mention, and one with the example in #120605 so we can see how your code handles them?

Yes, I will add them today.

[long-long-float]

I added a test for the issue. Could you review?

@rustbot ready

[Nadrieril]

@rustbot author

[Nadrieril]

Thank you!

@bors r+ rollup

[bors]

📌 Commit 04c966b has been approved by Nadrieril

It is now in the queue for this repository.

[Nadrieril]

Oops wait, could you squash all commits into one please?

@bors r-

[long-long-float]

I squashed all commits. Thank you for reviewing.

[Nadrieril]

👍

@bors r+ rollup

[bors]

📌 Commit 78e94cb has been approved by Nadrieril

It is now in the queue for this repository.