From: Robert Feldt <feldt@...>
Date: 2001-12-12T23:40:16+09:00
Subject: [ruby-talk:28345] Re: Rendering UML diagrams?

On Wed, 12 Dec 2001, Pierre-Charles David wrote:

> > I adopted the lexer from irb, and the top-level parser from rtags. I
> > then added a recursive-descent parse for the stuff I neded. The
> > combination works very well, and is about 1700 lines of Ruby. However,
> > it will definitely need tidying up over time, as it carries with it
> > some or its irb heritage.
> 
> Just an idea: it seems to me that lots of projects need a Ruby parser,
> but everybody is reimplementing his own (probably
> incomplete/buggy/inefficient). The Ruby interpreter already does all
> this work, so would not it be simpler to "open" the parser and to
> provide an interface to client programs? I'm thinking of something
> like 'Ruby.parse(aString)' which would return a (simplified) AST (with
> comments kept to allow RubyDoc and friends). IIRC, Smalltalk has
> something like this, correct?
> 
> I have no idea how difficult it would be to implement, but it would
> greatly simplify the development of all kinds of tools. If the current
> interpreter structure is too complicated to hack, maybe this feature
> could be included in Rite?
> 
Yes, I agree that it would be a good thing. However, its been discussed
before and Matz has said he's not willing to open up such internals. Maybe
things are changing for Rite; I don't know.

Before going for a full RubyInRuby parser I did Ruth which wraps the
internal parse trees into Ruby objects that you can manipulate. Its in the
rubyvm.sf.net CVS if you want to take a look. Have'nt touched it in a
while though.

Regards,

Robert