Suggestions when a single-field enum variant pattern is needed

[kpreid]

Beginners are often confused by the pattern where a set of types are combined using an enum, especially when the name of the enum variant is the same as the name of the variant's field's type. Even knowing how it works, I've myself forgotten to write out enum variant patterns containing struct patterns where they are needed. It would be useful if the compiler could give suggestions that understand this mistake.

Given the following code:

enum Foo {
    Bar(Bar),
}
struct Bar {
    x: i32,
}

fn oops_1(f: Foo) {
    match f {
        Foo::Bar { x } => todo!()
    }
}
fn oops_2(f: Foo) {
    match f {
        Bar { x } => todo!()
    }
}

The current output is:

error[E0769]: tuple variant `Foo::Bar` written as struct variant
  --> src/lib.rs:10:9
   |
10 |         Foo::Bar { x } => todo!()
   |         ^^^^^^^^^^^^^^
   |
help: use the tuple variant pattern syntax instead
   |
10 |         Foo::Bar(x) => todo!()
   |                 ~~~

error[E0308]: mismatched types
  --> src/lib.rs:15:9
   |
14 |     match f {
   |           - this expression has type `Foo`
15 |         Bar { x } => todo!()
   |         ^^^^^^^^^ expected enum `Foo`, found struct `Bar`

The compiler could instead recognize these cases, and produce the suggestion Foo::Bar(Bar { x }) => as well as a message specific to the situation:

• A variant pattern with fields which do not match the variant, but do match the fields of the variant's sole field.
• A pattern not of enum type where an enum pattern is expected, but the enum has a variant whose sole field is of the pattern's type.

@rustbot label +A-suggestion-diagnostics +C-enhancement +D-newcomer-roadblock

[kpreid]

This is now half completed (by #95386); the compiler now suggests wrapping oops_2's struct pattern with the necessary enum pattern. However, there is no appropriate suggestion for oops_1.