Metals does not infer concrete type of path dependent types

[pityka]

Describe the bug

In this snippet the identifier y in the lambda has type A{type T = Int} as verified by the compiler if type ascription is present. However if type ascription is not present, then metals infers it to A{type T = x.T }.

//> using scala 3.4.0
object Obj {
  trait A { self =>
    type T
    def fun(
        body: A { type T = self.T } => Unit
    ) =
      ()
  }
  val x: A { type T = Int } = ???

  // hover shows A{type T = x.T }
  x.fun { y =>
    y
    ()
  }
  // compiler infers as A of T = Int
  x.fun { (y:A{type T = Int}) =>
    y
    ()
  }

}

Expected behavior

Metals infers the same type as the compiler.

Operating system

macOS

Editor/Extension

VS Code

Version of Metals

v1.2.2

Extra context or search terms

No response

[tgodzik]

Thanks for reporting! We'll take a look at it!

[pityka]

After some thoughts and discussion on the scala discord, I am unsure whether there is an issue here. One can argue that the type of y is A{type T = x.T } which is equivalent to A{type T = Int}.

Auto-complete for e.g. extension methods do work in an extended example:

[tgodzik]

Looks like this is an issue in the compiler, mentioned by @rochala

[pityka]

I found a simpler case.
In the line val z = x.a, if I hover on 'z' then it shows the type as x.T, if I hover on 'a' then it shows the type as Int.

//> using scala 3.4.0
object Obj {
  trait A { self =>
    type T
    def a  : T = ???

  }
 
  val x: A { type T = Int } = ???

  val z = x.a


}

[kasiaMarek]

Hi, @pityka, sorry for nudging but are you still planning to continue your work on this issue or would you rather someone took over?

[pityka]

I do plan, i did not forget. Sorry I was distracted. As this is hobby I can not commit for a deadline so it is also fine if somebody else finishes if you find bandwidth and need it before I get to it. On 25 Apr 2024, at 09:01, Katarzyna Marek ***@***.***> wrote: Hi, @pityka, sorry for nudging but are you still planning to continue your work on this issue or would you rather someone took over? —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you were mentioned.Message ID: ***@***.***>

[kasiaMarek]

No worries, take your time, I'm just making sure it's not forgotten about 😄