Re: Unscientific performance results

> It looks to me as if there is something non-linear in the recognition of
> long comments by Coffeepot's Earley parser.  If I understand correctly,
> on deterministic grammars Earley should in principle be linear in cost
> (to the size of the grammar plus the size of the input) -- it makes me
> wonder whether there is something non-deterministic about the way the
> ixml grammar defines comments.

When I have a moment, I will publish the BNF grammar that is actually
used by the Earley parser. Perhaps you’ll see something I’m missing.
Perhaps the act of getting it ready to publish will focus my mind
sufficiently that I can see it before I publish it! :-)

                                        Be seeing you,
                                          norm

--
Norm Tovey-Walsh
Saxonica