misleading "try adding parentheses" in match with comma

[ChrisJefferson]

Consider the following (incorrect) code.

enum E { A,B,C }

fn x(e:E) -> i32 {
    match e {
        A,B => 1,
        C => 2
    }
}

The comma in the match is obviously supposed to be a |. However, rustc recommends I rewrite to (A,B), which makes the situation worse, as the compile error becomes expected enum E, found tuple.

[oli-obk]

is obviously supposed

I disagree. It could be either of these. At the point where we're suggesting this we don't know the type of the match or any other details.

We could suggest both (A, B) and A | B to make it clear that the user should put some thought into fixing the error.

[11Takanori]

This seems like a nice first issue for a beginner like me to contribute to, so I'd be happy to tackle this one.

[ChrisJefferson]

I don't know where the suggestions are mades but in this case (A,B) is obviously wrong, as it doesn't type check, while A|B does :)

[oli-obk]

I don't know where the suggestions are mades but in this case (A,B) is obviously wrong, as it doesn't type check, while A|B does :)

We have no way to distinguish between your code sample and the following code sample.

fn x(e: (u8, u8)) -> i32 {
    match e {
        A,B => 1,
        C => 2
    }
}

The error is happening during parsing. This means we have no type information available. All we know is that we're parsing a pattern. Think about it this way, what can you tell about the following code snippet:

    match e {
        A,B => 1,
        C => 2
    }

Just looking at that without looking at the surroundings tells you nothing, except that A, B and C are either constants or enum variants if the programmer didn't ignore a warning about lowercase variable names.

[ZerothLaw]

@11Takanori Just checking in - have you made progress on this issue? Willing to pitch in on it.

I'd like to slightly modified the initial code with:

enum E {A, B, C}

fn foo(e: E) -> u32 {
    match e {
        E::A,E::B => 0, 
        E::C => 1
    }
}

Error message is the same, but you can see that the compiler has more information, and still giving the wrong message.

[Krout0n]

Is anyone working on this issue? I'm a first-timer but want to try.