From: Aaron Patterson <aaron@...>
Date: 2009-04-14T02:17:01+09:00
Subject: Re: Using ruby for generic language parsing (or any	language-specific parsing libraries out there?)

On Tue, Apr 14, 2009 at 01:29:51AM +0900, Andrew S. Townley wrote:
> Hi,
> 
> I've been looking at the various parser generator options for Ruby on
> and off for a while now, and I was wondering if anyone had any
> experience/recommendations on existing solutions.
> 
> What I'd ideally like to have is a set of pre-defined parser libraries
> that defined an event-driven interface for handling productions.  This
> would theoretically allow you to simply plug in your own business logic
> into a robust, maintained and (hopefully) well-tested parser without
> actually having to go though the tedium and pain of defining a bunch of
> parsers yourself.
> 
> I'd like to be able to have something like this for SQL-92, Java and C,
> with potentially other languages in the future.  I'm guessing stuff like
> this has been developed for many of the IDEs out there that support
> these languages, but I've never seen something as complete as what I'm
> talking about via searching through google.
> 
> In the past, I've written parsers with Lex/Yacc, and I did do some work
> a couple of years ago with rex & racc, but for anything as complex as
> the above, I think rex/racc would be a) too time consuming/error prone
> (I found it incredibly difficult to debug and diagnose problems and
> generate reasonable error messages) and b) surely re-inventing the wheel
> for the nth time, where n > 10,000,000,000. :)

I've been using rex and racc for quite a while, and I'm pretty happy
with both of them.  I've found racc to be quite good in combination with
existing Yacc grammars.  I've written a pure ruby JavaScript parser[1]
and CSS parser[2][3] using racc.  I started on a C parser, but decided it
wasn't worth my time because I didn't want to deal with preprocessor
statements and we already have CAST[4].  Other non-trivial racc usage is
Ryan Davis's ruby parser[5] which is a ruby parser written with racc.

> I looked a bit at rparsec and treetop, but before I go down this path, I
> was hoping to tap some of the community's collective wisdom for pointers
> to similar initiatives.

Treetop looks neat.  Last time I looked at it though, it couldn't do
error recovery in grammars similar to the error token in Yacc.  That may
have changed since I looked at it last.

[1] http://github.com/tenderlove/rkelly/tree/master
[2] http://github.com/tenderlove/csspool/tree/master
[3] http://tinyurl.com/dkn8y4
[4] http://cast.rubyforge.org/
[5] http://rubyforge.org/forum/forum.php?forum_id=29842

-- 
Aaron Patterson
http://tenderlovemaking.com/