Scala: Difference between 'type A = XXX' and 'final type A = XX'?

问题

assuming I have an abstract type AA and concrete type XXX:

trait AA {
  type A = XXX
  final type B = XXX
}

In this case in any subclass of AA, both type A and B cannot be overriden, so it appears that the keyword final is completely redundant. Is this statement correct?

回答1:

It's hard to prove that they're exactly identical, but I'm going to argue that they are, minus a few useless quirks.

Useless quirks

First and most obviously, they give different error messages. But that's not all: it's technically possible to override A, you just can't override it to anything other than XXX:

trait A1 extends AA {
  override type A = XXX  // Compiles, but doesn't really do anything.
}

On the other hand, you can't ever override B:

trait A2 extends AA {
  override type B = XXX  // Does not compile.
}

Are there any useful differences?

Again, I'm going to argue that there aren't. In a very detailed answer to the question Is it possible to override a type field, StackOverflow user 0__ notes that

type T = C inevitably fixes T, which would kind of correspond to making a method final.

and

You can now easily see that it must be forbidden to further 'override' T

followed by some explanation of how the type system would be inconsistent if you could override T to a different type. See that answer for the details.

来源：https://stackoverflow.com/questions/55151774/scala-difference-between-type-a-xxx-and-final-type-a-xx