Thanks you. I am looking at using ttree (or a derivative of it, like what Randall's article describes, if I run into limitations).