Re: master 8bee4060ea4 2/2: Add peg.el as a built-in library

[Eli Zaretskii]

Ping!  Eric, any progress with this?

> From: Eric Abrahamsen <eric@ericabrahamsen.net>
> Date: Tue, 02 Apr 2024 20:03:05 -0700
> Cc: Stefan Monnier <monnier@iro.umontreal.ca>
> 
> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
> 
> > Eli Zaretskii <eliz@gnu.org> writes:
> >
> >>> branch: master
> >>> commit 8bee4060ea42c61e52ebe6487ff97bc095261050
> >>> Author: Eric Abrahamsen <eric@ericabrahamsen.net>
> >>> Commit: Eric Abrahamsen <eric@ericabrahamsen.net>
> >>> 
> >>>     Add peg.el as a built-in library
> >>>     
> >>>     * lisp/progmodes/peg.el: New file, taken from ELPA package.
> >>>     * test/lisp/peg-tests.el: Package tests.
> >>>     * doc/lispref/peg.texi: Documentation.
> >>
> >> Thanks.  A few comments about this:
> >
> > Thanks for the all the notes. I should have this sorted out later today.
> 
> Hi Stefan, I'm still working on integrating peg.el in to Emacs, and am
> not having any luck getting peg-tests.el to compile. Actually, any use
> of `peg-parse', compiled or interactive, seems to fail. Say you've got
> this simple example from the manual:
> 
> (peg-parse
>   (number sign digit (* digit))
>   (sign   (or "+" "-" ""))
>   (digit  [0-9]))
> 
> Inside `peg-parse', the only way this code is going to work correctly is
> if it takes the `with-peg-rules' branch, because these pexs aren't
> defined as rules yet. But the call to `peg-normalize' always succeeds,
> so we always take the `peg-run' branch, and the code fails on undefined
> rules.
> 
> (defmacro peg-parse (&rest pexs)
>   (if (and (consp (car pexs))
>            (symbolp (caar pexs))
>            (not (ignore-errors (peg-normalize (car pexs)))))
>       ;; `pexs' is a list of rules: use the first rule as entry point.
>       `(with-peg-rules ,pexs (peg-run (peg ,(caar pexs)) 
> #'peg-signal-failure))
>     `(peg-run (peg ,@pexs) #'peg-signal-failure)))
> 
> `peg-normalize' called on a cons applies `peg--macroexpand' to the list,
> which calls its "head" implementation, and we end up with
> 
> (call number sign digit (* digit))
> 
> `peg' sees that and assumes this has been defined as a rule.
> 
> So `peg-normalize' should be failing on a list like this, but I don't
> know how, exactly.
> 
> I'm hoping you have some pointers!
> 
> Thanks,
> Eric
> 
> 
>