Mapping Enumeration values causes diverging implicit expansion error on Scala 2.10.0-M2, worked with 2.8 and 2.9

[scabug]

I'm not sure if this is really a bug but atleast it seems like unintended consequence.

Using MyEnumeration.values.map(p => (p -> DoSomehtingWith(p)) causes diverging implicit expansion error.
This used to work on 2.8 and 2.9 and you can go around with MyEnumeration.values.toList.map but that seems strange and unnecessary.

More verbose explanation here: http://stackoverflow.com/questions/9487425/enumeration-and-mapping-with-scala-2-10

[scabug]

Imported From: https://issues.scala-lang.org/browse/SI-5534?orig=1
Reporter: Janne Sinivirta (vertti)

[scabug]

@hubertp said:
Adriaan: Why is this not critical, since it is a regression?

[scabug]

@adriaanm said (edited on Aug 7, 2012 9:00:27 AM UTC):
This is a library issue, probably caused by scala/scala@2d0f82898d
newCanBuildFrom for SortedSet is competing with the normal canBuildFrom.

Since there's an easy workaround (add a toList) and unlikely to be the compiler's fault, lowering priority and re-assigning.

object Phrase extends Enumeration {
  type Phrase = Value
  val PHRASE1 = Value("My phrase 1")
  val PHRASE2 = Value("My phrase 2")
}

class Entity(text:String)

object Test {
    val myMapWithPhrases = Phrase.values/*.toList*/.map(p => (p -> new Entity(p.toString))).toMap
}

[scabug]

@adriaanm said:
the other workaround would be to provide an instance of Ordering at Entity:

implicit object ord extends Ordering[Entity] { def compare(x: Entity, y: Entity) = ??? }

since values is a SortedSet, that's needed for its canBuildFrom

Stefan, I think we can close, but I'll let you have a second look since you know in more detail what's going on.

[scabug]

@szeiger said:
I wish I knew what's going on. This canBuildFrom stuff is a hack (overriding newCanBuildFrom in SortedSet) upon a hack (needing both canBuildFrom and newCanBuildFrom in the first place) and I don't fully understand the latter.

The reason for needing the SortedSet hack is that the standard canBuildFrom for SortedSets only works for sets with a user-specified ordering (even though SortedSet itself has no such requirement).

I think it's a legitimate bug that should be fixed but off the top of my head I can't think of a good solution that won't further complicate the implementation (with the risk of creating new bugs in slightly different situations).

[scabug]

@retronym said:
Sounds like this needs a library change; we'll have to defer to 2.11 due to the forward binary compatibility regime.

[scabug]

@gkossakowski said:
Unassigning and rescheduling to M7 as previous deadline was missed.