INT: Support "Create associated function" for path Self::func()

[dima74]

changelog: Support Create associated function intention for path Self::func()

[Kobzol]

Could it also work with <this struct>::Foo? For example:

struct Foo;

impl Foo {
    fn foo() {
        Foo::bar();
    }
    // create fn bar() {}
}

Maybe we could just always detect/find an existing impl block and use it if it's available? That seems like it could unify the "OtherType::foo()", "::foo()" and "Self::foo()" use-cases.

[Kobzol]

It looks like in the current usage, existingImpl and item are exclusive? It could be modeled by some enum, but it's probably overkill.

[dima74]

Could it also work with <this struct>::Foo?

It already works but creates new impl. As you said, it would be nice to find an existing impl, and I plan to do it next. Btw for Self::foo() we can support the intention even if Self is not ADT, since Self resolves to an impl

[Kobzol]

Ok, in that case LGTM :)

[artemmukhin]

LGTM. Regarding this intention in general, I noticed that it does not reuse existing impl blocks when generating new associated functions or methods, e.g.:

struct  S;

impl S {
    fn foo() {}
}

// new `impl S` block for `bar` will be added instead of reusing the previous one

fn foo() {
    S::bar/*caret*/(1, 2);
}

IMO it would be nice if the intention could avoid generating unnecessary impl blocks in such cases, for example when there is only one existing impl block located in the same file. @dima74 WDYT?

[dima74]

IMO it would be nice if the intention could avoid generating unnecessary impl blocks in such cases, for example when there is only one existing impl block located in the same file. @dima74 WDYT?

Yeah I plan to do this next

[dima74]

bors r=ortem

[bors]

Build succeeded:

• check

[neonaot]

Judging by the tests, this is the expected behavior, but I opened the issue before looking at the tests :D

#9466