`ExplicitResultTypes` breaks compilation when using symbolic methods

[dhalilov]

Hi folks,

ExplicitResultTypes doesn't seem to work correctly when the def/val uses symbols, as it inserts : right next to the symbol, which fools the Scala parser. Adding a space before the : should resolve the issue.

I'm not sure if this is part of the known limitations mentionned in the documentation page, as I haven't found an issue describing the problem.

How to reproduce

Here is a minimal example, using Scala 3.5.2 and sbt 1.10.7.

src/main/scala/Main.scala:

object A {
  // Using the non-trivial example from the documentation to make sure that ExplicitResultsType kicks in
  def ! = 1.to(10).map(i => i -> i.toString()).toMap
}

@main def main(): Unit =
  println(A.!)

build.sbt:

lazy val root = project
  .in(file("."))
  .settings(
    scalaVersion := "3.5.2",
    semanticdbEnabled := true
  )

project/plugins.sbt:

addSbtPlugin("ch.epfl.scala" % "sbt-scalafix" % "0.13.0")

.scalafix.conf:

rules = [
  ExplicitResultTypes
]

Running sbt scalafix produces the following line:

def !: Map[Int, String] = 1.to(10).map(i => i -> i.toString()).toMap

which fails to compile. The following version:

def ! : Map[Int, String] = 1.to(10).map(i => i -> i.toString()).toMap

would compile correctly.

[tgodzik]

Thanks for reporting, looks like an issue in the presentation compiler. We need to fix it upstream, but should be a quick fix.

[bjaglin]

Supposedly fixed in the latest dotty snapshot thanks to scala/scala3#22485. It won't be backported to 3.5.x since it's EOL (and we follow the minor version of the sources to pick the presentation compiler), so I am afraid you will have to bump to 3.6.4 (assuming it does get in there - I am not familiar with the release/branching strategy).

[dhalilov]

Supposedly fixed in the latest dotty snapshot thanks to scala/scala3#22485. It won't be backported to 3.5.x since it's EOL (and we follow the minor version of the sources to pick the presentation compiler), so I am afraid you will have to bump to 3.6.4 (assuming it does get in there - I am not familiar with the release/branching strategy).

No worries, thank you all for the quick fix!

[tgodzik]

It should be ported back to 3.3.6.

[bjaglin]

It should albo be ported back to 3.3.6.

Right 👍 I didn't mention it because the report mentioned a Scala version after LTS, but I certainly hope we can backport that to LTS!

[bjaglin]

you will have to bump to 3.6.4 (assuming it does get in there - I am not familiar with the release/branching strategy).

FTR, I just had a look at 3.6.4-RC2 and it does not contain the fix, because the release branch for 3.6.4 was cut before the fix was merged. As a result, we'll have to wait for 3.7.0 to be integrated into scalafix, which means the bug will not be fixed for code built against 3.6.x unfortunately.