Re: Re: software design clarification

On Fri, 2007-05-04 at 00:05 -0700, Nathaniel Smith wrote:
> On Fri, May 04, 2007 at 12:44:09AM -0400, Gilles J. Seguin wrote:
> > > I know I've brought this up before, and apologies for providing
> > > another distraction from, you know, actually writing said code... but
> > > isn't this yet more reason to consider just integrating Elsa?  It's
> > > 2-clause BSD, it already parses C++ well, Mozilla is investing heavily
> > > in using it to build automated refactoring tools...
> > > 
> > > It's at, of course:
> > >   http://www.cs.berkeley.edu/~smcpeak/elkhound/sources/elsa/
> > 
> > an old project, we are not crazy, if the thing would work we would have
> > used it since long time.
> > 
> > from documentation,
> > «The template design is still somewhat incomplete.»,
> > means can not used it if using template.
> 
> Yeah, I dunno -- just skimming the source there is clearly a bunch of
> template support code, and 'cc_type.html' says that template support
> is incomplete and that template instantiation should be added while
> 'design.html' appears to document their template instantiation
> support... I can't really tell what the status is.  (NB, it looks like
> the URL above is out of date; the code is still being updated and the
> latest version is at the cubewano.org site I linked later.)
> 
> In any case, isn't the question more about what is easiest to fix?
> >From the outside, it looks like Elsa is parsing huge programs that
> have clobbered every other alleged C++ parser (aside from the 3
> C++ compilers), and supporting fancy data-flow analysis and accurate
> symbol resolution, while you guys are still working on just getting
> symbol tables going.  But, obviously, I have not looked into these
> things as closely as you guys.

That means you have the thing running.
Parse something that have template, for example it use Boost.
I will be very receptive for sure.

> > In brief, parsing C++ require backtracking that is it, that is all.
> 
> Huh?  Are you claiming that the GLR-parser approach used by Elsa
> cannot be made sufficient for C++ code?  That sounds incorrect to me,
> but I'm not an expert.  Really I am just guessing what your sentence
> means, though.

Need to check that one, can be parsing already validated code.