Overzealous warning from scala compiler: Auto-application to `()` is deprecated

Question
1

I’m copying this message here from discord, as it may need additional discussion and reflection.

Can someone help me understand this compiler warning.

[warn] /Users/jnewton/Repos/regular-type-expression/cl-robdd-scala/src/test/scala/bdd/BddTestSuite.scala:93:36: Auto-application to `()` is deprecated. Supply the empty argument list `()` explicitly to invoke method size,
[warn] or remove the empty argument list from its definition (Java-defined methods are exempt).
[warn] In Scala 3, an unapplied method like this will be eta-expanded into a function.
[warn]           case _: BddTerm => assert(bdd.size() == 1)
[warn]                                    ^

Is this really a warning, telling me to do what I’m already doing? or is it rather just an info message and should not be a warning at all? Or am I reading the message wrong?

Is the message telling me to omit the () or to keep them there?

I’m confused.

Here is the offending code

  def genSamples(): Set[Bdd] = {
    ...
  }

  test("size") {
    withNewBddHash {

      val samples = genSamples()

      for (bdd <- samples) {
        bdd match {
          case _: BddTerm => assert(bdd.size() == 1)
          case b: BddNode => assert(bdd.size() <= 1 + b.positive.size() + b.negative.size()) // maybe less because some nodes or terms might be shared
        }
      }
    }
  }

The size method is indeed defined using parens ()

sealed abstract class Bdd {
  ...
  
  def size():Int = {
    ...
  }
...
}
2

Can’t reproduce: Scastie - An interactive playground for Scala.

What is that withNewBddHash {? Feels like something that would do dark magic at compile / runtime; which may be related to the error.

3

Nothing strange here. Do you think the compiler gets confused inside call-by-name code?

  val maybeNodeHash = new DynamicVariable[Option[Cache]](None)

  // wrapper for all code which needs to calculate with Bdds.  The Bdd
  //   calculations need a hash map to enforce structural identity.
  //   Attempt to construct a Bdd outside the dynamic extend of this
  //   function results in an error.
  def withNewBddHash[A](code: => A): A = {
    maybeNodeHash.withValue(Some(newCache())) {
      code
    }
  }
4

Do you think the compiler gets confused inside call-by-name code?

Inside call by name, no.
But maybe the DynamicVariable may be related. Anyways, without a reproducible example is impossible to help.

5

To answer the question, this warning is straightforward, but parens are surprisingly tricky.

The caret is pointing at the apply paren for assert, which is wrong.

I’d recommend -Vprint:parser,typer,refchecks which will show what the trees really look like. If you turn on -Werror and it’s emitting this warning, it will error out after typer.

I’ve been always cross-compiling with Scala 3, in case one compiler gives better messaging.

6

assert is probably a scalatest macro so I guess that it’s generating code that the compiler doesn’t like anymore. You could try using the latest scalatest version and see if that fixes it.

2 Likes
7

I’ve updated to scalatest 3.2.15, and the error seems to still be there.

libraryDependencies += "org.scalactic" %% "scalactic" % "3.2.15"
libraryDependencies += "org.scalatest" %% "scalatest" % "3.2.15" % "test"
8

I reported the issue with scalatest

9

The workaround suggested by the developers is to use Predef.assert() rather than assert().

10

Looks like they also merged a fix.

1 Like