bug#60953: The :match predicate with large regexp in tree-sitter font-lock seems inefficient

[Dmitry Gutov]

On 01/02/2023 15:10, Eli Zaretskii wrote:
> > Date: Tue, 31 Jan 2023 20:16:01 +0200
> > Cc: casouri@gmail.com, 60953@debbugs.gnu.org
> > From: Dmitry Gutov <dgutov@yandex.ru>
> >
> > > Can you describe what that function should do?  I don't think I have a
> > > clear idea of that.
> >
> > In Lisp that function could be implemented as
> >
> >     (defun buffer-substring-match (regexp &optional start end
> >                                    inhibit-modify)
> >       (string-match regexp
> >                     (buffer-substring (or start (point-min))
> >                                       (or end (point-max)))
> >                     inhibit-modify))
> >
> > Meaning, it matches the regexp against the buffer substring, with the
> > string-start and string-end anchors working.
> >
> > But it would be implemented in C, meaning we could avoid the extra
> > consing and funcall overhead.
>
> Now I'm confused, because I thought the C functions we were
> considering all fit the above description.

Except they don't match \` and \' at START and END. Only at actual 
point-min and point-max.

I suppose the new function could be implemented with narrowing as well. 
If we decide it's the best method.

> > Anyway, it seems like it might be too late as an addition to Emacs 29.
> > And we can implement the match predicate using narrowing for this
> > release, to be updated later.
>
> Right, this is not a catastrophe, IMO.  The work on TS-based modes is
> just beginning.

I also don't see any particular slowdown from altering BEGV and ZV in C. 
So the current method might be the fastest possible anyway.