LSP doesn't show label completion inside Record after spreading

[uh-kay]

Steps to reproduce:

pub type Foo {
  Foo(wibble: String, wobble: String)
}

pub fn main(foo) {
  Foo(..foo, wib )
                ^ no autocomplete shown
}

pub fn baz(foo) {
  Foo(wib )
         ^ autocomplete shown
}

Gleam v1.14

[lpil]

Hello! I've been unable to reproduce this. Could you test with nightly or v1.15.0-rc1 please 🙏

[ankddev]

@lpil But shouldn't it show wibble: completion here? Or have I misunderstood the issue?

[giacomocavalieri]

This is interesting, if I try and recreate your setup @lpil I get no completions in the tests:

pub fn main() {
  Wibble(..todo, w)
//                ^Trigger completions here
}

pub fn wibble() { todo }

The snapshot I get back from the LS has no completions:

 4 │+pub fn main() {
 5 │+  Wibble(..todo, w|)
 6 │+}
 7 │+
 8 │+pub fn wibble() { todo }
 9 │+
10 │+
11 │+----- Completion content -----

I think this might be due to the incomplete record update resulting in a syntax error. While typing in the editor it relies on the previous (now outdated) result of a previous compilation.
But the code in tests results in a syntax error without ever being successfully compiled first. So it has no info to go by to produce completions.
I think this would be best addressed by not making it a syntax error, it would make it a lot easier to then add completions reliably.

[lpil]

Sorry, I don't follow. Why would we want to do that given that in an editor it does provide completions?

But shouldn't it show wibble: completion here? Or have I misunderstood the issue?

@ankddev That's not a feature of the Gleam LS, but it could be! Please open an issue.

[uh-kay]

Hello! I've been unable to reproduce this. Could you test with nightly or v1.15.0-rc1 please 🙏

I mean it doesn't show the wibble: completion just like @ankddev said. That's what I'd like the LS to show.

[GearsDatapacks]

That's not a feature of the Gleam LS, but it could be! Please open an issue.

@lpil it is, I implemented it a while back in #4490

[giacomocavalieri]

Sorry, I don't follow. Why would we want to do that given that in an editor it does provide completions?

It makes it really hard to test, the change is actually pretty straightforward (#5439) but I can't think of a good way of testing this