Re: PEG Parser

[Noah Lavine]

Hello,

I tried looking at it a bit. First of all, your cleanups are awesome. :)

Unfortunately, I need to ask for help in order to work on this more. I
thought I would first move the code-generating functions to their own
module. It seems like this should be a simple and obviously-correct
transformation, because I didn't change any code - just moved it to
its own module, and replaced its definition by a (use-modules ...)
clause. Yet I have somehow created an error, and I don't see why. If
you have time, could someone please explain why the attached patch
does not work? I am afraid there is some interaction between modules
and syntax generators that I don't understand.

The patch will apply to the wip-mlucy branch.

Thanks,
Noah

On Fri, Feb 18, 2011 at 5:03 PM, Andy Wingo <address@hidden> wrote:
> On Wed 02 Feb 2011 01:26, Noah Lavine <address@hidden> writes:
>
>> Here it is! All of the unhygienic syntax is gone, is a series of only
>> 20 commits. :-) The peg.test tests should all pass after each one of
>> these commits.
>
> Thanks!  You've probably seen that I've applied this to wip-mlucy, which
> we should probably rename wip-peg.  I've also added on a number of
> cleanups of my own, some of which I will push out shortly when my ISP
> figures out the route to git.sv.gnu.org again (hah).
>
> The branch still needs some work before it can go in.  I have a feeling
> that it should probably be split into two modules -- one providing the
> things that peg-sexp-compile needs (minus `peg' patterns perhaps?) and
> another that uses the "base" library to define a PEG grammar.  Perhaps?
> In any case we need to not have the entire thing in one big ol'
> eval-when.
>
> Also, the documentation needs some help, and perhaps the patterns need
> some tweaking -- for example (& pat) makes more sense than (body & pat
> 1) or the like.
>
> I think Michael's work was pretty great, especially considering the
> scope of the problem.  It has the potential to have so wide an impact
> that we should focus on making it have exactly the right interface
> before we merge it in.
>
> Onwards, upwards, etc.!
>
> Andy
> --
> http://wingolog.org/
>

0001-Move-PEG-code-generators-into-their-own-module.patch
Description: Binary data