Completion stops working in macros

[Matelasse]

Specifically in prtinln! macro.
In the following code where ┃ is the cursor position

struct Test {
    i: char,
    t: i8,
    w: i8,
}
fn main() {
  let t = Test { i: 'a', t: 0, w: 0 };
  t.┃ // completion works here
  println!("{}", t.┃) // completion does not work here
}

[bjorn3]

That is expected, as something that looks like an expression to humans may not actually be an expression. It may in fact be a pattern or just some tokens.

[Matelasse]

@bjorn3 Well then can this be and few other macros be whitelisted, because you tend to want completion in println!, format!, ... and not seeing them makes you think that extension has stopped working.

[matklad]

Yeah, this is a known limitation at the moment. The crucial missing bit here is ability to correlate position in macro calls and in the expanded code. Shouldn't be too hard to implement, but will require quite a bit of plumbing throughout rust-analyzer.

[JoshMcguigan]

The example in the original post is now working, but it doesn't not work if the closing paren is left off.

struct Test {
    i: char,
    t: i8,
    w: i8,
}
fn main() {
  let t = Test { i: 'a', t: 0, w: 0 };
  t.┃ // completion works here
  println!("{}", t.┃ // completion does not work here, note the lack of closing paren
}

I made a small patch which fixes this here, but I'm not sure it makes sense to fix this issue at the macro expansion level as opposed to fixing it earlier in the parsing. Thoughts on this?

[flodiebold]

That happens because the code here only uses ( as delimiter if there's a matching ) in the end. We should probably be a bit more lenient there.

[edwin0cheng]

Yeah, that's definitely a bug in the place @flodiebold mentioned, and I think the approach in #3518 is much simpler than node/token based one used in there. I will try to refactor it later.

[matklad]

the basics are fixed in master