Is there a way to unify type parameters ?

Question

+[Accidentally posted in the Java Programming forum here: http://forum.java.sun.com/thread.jspa?threadID=5262928&tstart=0]+
The following code is pretty self contained. I will use some Haskell to explain the problem, cause it's easier (for me) to read.
For those foreign to Haskell, the basic syntax of these signatures is:
function_name :: ArgType1 -> ArgType2 -> .... -> ArgTypeN -> ReturnTypeWith list (or in this case ArrayList) arguments as:
[ArgTypeN]It starts off with the definition of:
BinaryFunction<T,U,X>With main method:
apply :: T -> U -> XThen is defined:
fold :: X -> [U] -> XAnd inside fold, apply is called such that X = T
apply :: X -> U -> XThis only compiles if the first argument to apply is cast to T, with the warning "BinaryFunction.java uses unchecked or unsafe operations."
My question is whether X can be unified with T manually at the call-site or in the definition of fold, in a safe way. Otherwise, I would like to know if it's acceptable to leave the cast in, since it's obviously ok.
Thanks. =)
package lib.functional;
import lib.functional.FunctionApplicationError;
import java.util.ArrayList;
public abstract class BinaryFunction<T,U,X>
     abstract public X apply(T t, U u) throws FunctionApplicationError;
     private X fold(X x, ArrayList<U> us)
            for (U u : us)
                 x = apply((T) x, u); // *************************
            return x;
}Edited by: Dorice on Feb 11, 2008 3:25 AM
Changed title... Hopefully this is clearer.

Answer 1

Dorice wrote:
It should be okay as long as I only use BinaryFunction<X, U, X>() when I need fold, right?If you only need X U X, why define T (or why define X)?
The main problem here is that, once you defined your instance, the generic types are defined. So, maybe ejp's advice would be a solution for your problem as you expect the X to be a T to define X extends T.
This would be different, if BinaryFunction was instantiated each time before invoking apply.
Oh yeah, I forgot to (re)ask, can I still play ball with the cast in place?What for? The cast will be no cast, as the warning says. It works accidentally at runtime in case X subclasses T. Thus, the type safety you define by Generics only is pretended and factually pointless.