On Mon, Oct 10, 2005 at 10:45:54AM -0400, Matt Fowles wrote: > Perhaps a better approach would be to perform a bit > of static analysis on the grammar and look for left recursions at > creation time (I believe that is a known problem). Then we can just > warn once up front and go on our merry way recursing at matchtime.
I think this approach falls back into my category of "unless someone wants to take on the task of implementation it's not likely to be resolved anytime soon..." :-) :-) In a dynamic environment like the one we have (parrot/perl) I find it's really hard to say what constitutes "creation time", especially since we can refer to subrules dynamically ( <$foo>, <::$foo>, etc. ) and our grammars have inheritance in which a rule defined by a parent grammar can invoke a rule defined later in a child grammar. As with closures, we could probably punt on performing any analysis of rules with dynamic references or that refer to rules outside of the current grammar. But given that we can't really know until matchtime what subrule will be invoked by an expression like <binary>, it's hard for me to imagine a whole lot we can do in the way of static analysis beforehand. However, that could easily just be a limit of my imagination. Pm
