Looks interesting! Have you implemented any more complex grammars with this, e.g. Python or Javascript? Would be interested to see how they perform.
Also, does it support multi-stage parsing, i.e. generation of a token tree and then a parse tree from it?