Right direction?

[Peter Ahé]

Your patch is line wrapped.  Could you try to resend it as an attachment?

Cheers,
Peter

On 5/26/07, Matthias Ernst <ernst.matthias at gmail.com> wrote:
> Hi,
>
> I've created an experimental feature, namely cascading for void method
> calls as described on my blog @ http://mernst.org/blog/rss.xml
> (Firefox only :-(  Patches at the end, too.
>
> I would be glad to hear some comments what you think about the patches
> (not the semantics but the implementation) - did I miss something? A
> hint how I would find out the receiver type for unqualified method
> calls without duplicating the lookup strategy in Attr#visitApply would
> be appreciated, too!
>
>
> Thanks
> Matthias
>
>
>
> --- comp/Attr.java.orig.txt     2007-05-26 13:55:53.000000000 +0200
> +++ comp/Attr.java      2007-05-26 14:15:51.000000000 +0200
> @@ -1315,6 +1315,18 @@
>                                restype.tsym);
>              }
> +            // as a special case, o.m() where m is a void method, has type of o
> +            if(restype == Symtab.voidType &&
> !isStatic(TreeInfo.symbol(tree.meth))) {
> +                if(tree.meth.tag != JCTree.SELECT) {
> +                    log.warning(tree.meth.pos(), "Cannot (yet)
> cascade unqualified method call");
> +                } else {
> +                    JCExpression receiver = ((JCFieldAccess)
> tree.meth).selected;
> +                    restype = receiver.type;
> +                }
> +            }
> +
> +
> +
>              // Check that value of resulting type is admissible in the
>              // current context.  Also, capture the return type
>              result = check(tree, capture(restype), VAL, pkind, pt);
>
> --- comp/TransTypes.java.orig.txt       2007-05-26 13:56:46.000000000 +0200
> +++ comp/TransTypes.java        2007-05-26 14:15:51.000000000 +0200
> @@ -583,7 +587,8 @@
>          tree.args = translateArgs(tree.args, argtypes, tree.varargsElement);
>
>          // Insert casts of method invocation results as needed.
> -        result = retype(tree, mt.getReturnType(), pt);
> +        // mernst: do not use mt.resType as cascading may have changed it
> +        result = retype(tree, types.erasure(tree.type), pt);
>      }
>
>      public void visitNewClass(JCNewClass tree) {
>
> --- jvm/Gen.java.orig.txt       2007-05-26 14:00:07.000000000 +0200
> +++ jvm/Gen.java        2007-05-26 14:16:38.000000000 +0200
> @@ -1664,14 +1664,24 @@
>   *************************************************************************/
>
>      public void visitApply(JCMethodInvocation tree) {
> +
>         // Generate code for method.
>         Item m = genExpr(tree.meth, methodType);
> +
> +        // in case of a cascade, duplicate the receiver
> +        boolean cascaded = ((MethodType)tree.meth.type).restype ==
> Symtab.voidType && tree.type != Symtab.voidType;
> +        Item stackItem = cascaded ? items.makeStackItem(tree.type) : null;
> +        if(cascaded) stackItem.duplicate();
> +
>         // Generate code for all arguments, where the expected types are
>         // the parameters of the method's external type (that is, any implicit
>         // outer instance of a super(...) call appears as first parameter).
>         genArgs(tree.args,
>
> TreeInfo.symbol(tree.meth).externalType(types).getParameterTypes());
>         result = m.invoke();
> +
> +        // in case of a cascade, the receiver is top of stack now
> +        if(cascaded) result = stackItem;
>      }
>
>      public void visitConditional(JCConditional tree) {
>