It seems like we wont be able to use the new OGNL byte code, as we will end up using reflection in most cases anyway. The first problem is the value stack, given and expression "A.B", it is always evaluated as "Root.A.B" where Root is the CompundRoot. That would be translated into root.getA().getB(), the problem is that root is a wrapper of a list of objects, so we never know the type of getA() which makes it impossible to compile it into byte code.
Enter the hack: I managed to get the compilation nodes from OGNL, so I would use reflection to get X=Root.A, and then generate the bytecode for X.B, this kinda works but defeats the whole purpose as the expressions could not be cached properly, so it is useless. Even if we use the hack described above, expressions like A[C].B would need to use reflection as the type of A[C] cannot be determined (if you thought: use the hack again on A[C], go ahead and smack yourself around a bit :) ) so we are out of luck on this, until invokedynamic is supported by the jvm(http://jcp.org/en/jsr/detail?id=292) (good article: http://blog.headius.com/2008/09/first-taste-of-invokedynamic.html) musachy -- "Hey you! Would you help me to carry the stone?" Pink Floyd --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@struts.apache.org For additional commands, e-mail: dev-h...@struts.apache.org
