Why can't a value class wrap another value class?

Question

Is there a reason for the restriction that one value class can't wrap another value class?

It seems an obvious thing to want to do, and extends the benefit further down the class hierarchy. If class B(val x: A) extends AnyVal can be treated as compile-time-only, and uses class A behind the scenes, then it would be nice if class C(val y: B) extends AnyVal could do the same. It's implemented transparently as B, which is implemented transparently as A at runtime in the JVM, so all the efficiency and gc overhead advantages carry over to class C.

Why the restriction?

Answer 1

This is an implementation restriction. It's lifted in Dotty:

$ dotr
Starting dotty REPL...
scala> case class C1(i: Int) extends AnyVal
// defined case class C1
scala> case class C2(c1: C1) extends AnyVal
// defined case class C2
scala> val x = C2(C1(1))
val x: C2 = C2(C1(1))