Enum variants are suggested in RHS of match arm

[jonas-schievink]

enum MyEnum {
    VariantA,
    VariantB,
}

fn f(e: MyEnum) -> Option<u8> {
    match e {
        MyEnum::VariantA => Some(0),
        MyEnum::VariantB => $0
    }
}

We suggest both MyEnum::VariantA and MyEnum::VariantB here, but they don't really make any sense, since the expected type is Option<u8>.

[jonas-schievink]

Huh, apparently we don't complete VariantB here at all, I thought that was the misfiring completion:

enum MyEnum {
    VariantA,
    VariantB,
}

fn f(e: MyEnum) -> Option<u8> {
    match e {
        MyEnum::VariantA => Some(0),
        Varia$0
    }
}

[jonas-schievink]

Yeah the original issue was just that this code was triggering because for some reason ctx.expected_type is MyEnum there instead of the correct Option<u8>:

rust-analyzer/crates/ide-completion/src/completions/expr.rs

Lines 155 to 160 in b6b17c7

	if let Some(hir::Adt::Enum(e)) =
	ctx.expected_type.as_ref().and_then(|ty| ty.strip_references().as_adt())
	{
	super::enum_variants_with_paths(acc, ctx, e, |acc, ctx, variant, path| {
	acc.add_qualified_enum_variant(ctx, variant, path)
	});

[flodiebold]

Yeah, I actually looked into this a bit when you posted it first, should have written some notes. We get the wrong expected type because we treat this like the match e { $0 } case. It wasn't super trivial to fix because in match e { Variant => $0 }, the cursor is not 'in' the match arm. I'm a bit rusty on handling syntax trees though, so maybe there's some nice easy way to detect this case 😅