Comments on: Foiled by Erasure!

By: Ricky Clarkson

Ricky Clarkson — Mon, 05 Nov 2007 21:09:50 +0000

Erasure itself isn’t really a bad thing; Haskell erases types but doesn’t have this kind of problem. You can still do type-based dispatch in Haskell, but at compile-time (typeclasses). Scala appears to exhibit the same problem as Java here, but has something equivalent to typeclasses – implicits, so you can still do most things you’d like to be able to.

I suppose you could split MyBean up into the bean plus a couple of EventSubscribers.

By: Stephen Colebourne

Stephen Colebourne — Mon, 05 Nov 2007 09:34:18 +0000

The naive way for an improved compiler to handle this would be:

public void onEvent(Object event) {
. if (event instanceof String) {
. . onEvent((String) event);
. } else if (event instanceof StatusEvent ) {
. . onEvent((StatusEvent) event);
. } else {
. . throw new ClassCastException(”Invalid use of generics”);
. }
}

It would appear at first glance that this could be done without erasing erasure. The question is whether it would perform well enough for you?

By: Bob Lee

Bob Lee — Mon, 05 Nov 2007 05:33:48 +0000

I think this might be more indirectly related to erasure. I mean, method calls are bound at compile time. You have all the type information you need for this to work. Actually, you have all the information you’d need at runtime, too. I’ll bet the real problem is that calls to the methods with raw types would be ambiguous.