Re: [Spirit-general] Why doesn't Spirit use an Earley parser?

[Ben H. <jam...@go...>]

tyler wrote
> I'm thinking of writing a Spirit-like library based on a Earley parser.
> Are there any reasons that would be a bad idea?

I'm very interested in this. I also found the following an interesting read:
http://blogs.perl.org/users/jeffrey_kegler/2013/04/is-earley-parsing-fast-enough.html

To be blunt, it would be nice to have tools that go beyond LL and LALR(1).
To quote Dick Grune and Ceriel J.H. Jacobs in PARSING TECHNIQUES A Practical
Guide "In summary, LR(k) parsers are the strongest deterministic parsers
possible and they are the strongest linear-time parsers known, with the
exception of some non-canonical parsers; ... They react to errors
immediately, are paragons of virtue and beyond compare, but even after 40
years they are not widely used."

Although the Earley algorithm is a little left field, it would be
fascinating to see how it performs in reality and would be a lot easier to
implement than LALR(k).

Regards,

Ben Hanson



--
View this message in context: http://boost.2283326.n4.nabble.com/Why-doesn-t-Spirit-use-an-Earley-parser-tp4676550p4676952.html
Sent from the spirit-general mailing list archive at Nabble.com.