emacs-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Fixing ill-conditioned regular expressions. Proof of concept.


From: Alan Mackenzie
Subject: Re: Fixing ill-conditioned regular expressions. Proof of concept.
Date: Thu, 26 Feb 2015 20:01:08 +0000
User-agent: Mutt/1.5.22 (2013-10-16)

Hello, Stefan.

On Thu, Feb 26, 2015 at 02:12:32PM -0500, Stefan Monnier wrote:
> >> >     R*\(\)R*
> >> > , but anybody who writes such regexps deserves what she gets.
> >> What is it that I deserve to get?
> > You deserve, perhaps, to lose (match-beginning 1) and (match-end 1),
> > which were ill-defined anyway.

> Why do you think so?  They seem perfectly well-defined to me.
> They're just always equal to one another, of course, but to the extent
> that the regexp syntax only forces me to put "named positions" in pairs,
> if I need a single position, it's fairly natural to just use \(\).

I really did mean R*\(\)R*, with R being the same on both sides of the
\(\), but the *s possibly being +s.  _That_ is nasty and undefined.

> > Have you really written a regexp like this (apart from for testing
> > purposes)?.  If so, what's it for?

>    grep '\\\\(\\\\)' **/*.el

> finds 27 matches.  Taking one example from the list:

>    lisp/emacs-lisp/smie.el:     ((looking-at "\\s(\\|\\s)\\(\\)")

> what this does is to let me use (match-beginning 1) to figure out which
> of the two alternatives was matched.  I could have written this as

>        ((looking-at "\\s(\\|\\(\\s)\\)")

> but this would be (marginally) slower, because we'd always push
> a "group-start" marker before try to match "\\s)", whereas with the
> other rule, we only do that when we know "\\s)" has matched.

OK.

> > By the way, how do you see the prospects of this file becoming
> > incorporated into Emacs at some stage?

> To be honest, I haven't looked at it at all, yet.
> The vague understanding I have of what it might be sounds interesting.
> It's just a patch trying to cover up the worst aspects of the
> current regexp engine, but since there doesn't seem to be much interest
> in improving/overhauling the regexp engine, maybe it's a good stop-gap.

Thanks.  I'll continue working on it, adding a decent set of test cases
too.

>         Stefan

-- 
Alan Mackenzie (Nuremberg, Germany).



reply via email to

[Prev in Thread] Current Thread [Next in Thread]