Operations on NamedTuples do not work with singleton types due to match type refusing to reduce about scala3 HOT 7 OPEN

Yes, widening singleton types of paths that can be substituted later creates established unsoundness issues. See #19746

from scala3.

Thanks for the reference. A possibly viable workaround is to use a wrapper type:

import scala.language.experimental.namedTuples

class Wrapper[T](x: T)
object Wrapper:
  type Extract[W <: Wrapper[?]] = W match
    case Wrapper[t] => t

object Test:
  val f: (name: String, age: Int) = ???
  val f2 = Wrapper(f)
  val x: NamedTupleDecomposition.Names[Wrapper.Extract[f2.type]] = ("name", "age")

Not sure if there's a better way or if that can help suggest some more systematic fix.

from scala3.

@smarter what if we explicitly create a Widen type operation? So Names will be implemented as Widen[X] match ...

from scala3.

That would reopen the soundness hole of #19746 if Widen[this.type] can reduce to different types depending on the context.

from scala3.

I'm not sure that will happen, because of the lazy nature of singleton ops operations.

from scala3.

Though note that the fix for #19746 (0a3497b) specifically target singleton types coming from parameters (since those can be substituted), so maybe something like a widenSingletonOfNonParamRef operation? It's not clear to me that this is enough if the underlying type of a non-param singleton is a GADT, in which case maybe the original hole isn't fixed.

from scala3.

in which case maybe the original hole isn't fixed.

Indeed: #20515.

from scala3.