On Fri, Feb 22, 2008 at 06:05:29PM +0100, Bjorn Reese wrote:
> Daniel Veillard wrote:
> 
> >   Hum, unfortunately this looks like a different problem, it seems
> > to take a lot of cycles and I'm afraid there is something serious going
> > on and that can be only studied and fixed by debugging the generated
> > automata.
> >   reducing the expression and the input to provide a minimal automata
> > showing the problem is a good first step.
> 
> Place the following two lines in a file, and run testRegexp on it:
> 
> =>((a[0-9]+|b[0-9]+)c)
> b12345678901234567890123

  thanks a lot of an even simpler test case, allowed me to nail it down
while in the train yesterday,

> If you add more digits to the second line, it will slow down even
> further at what looks like an exponential rate.

  yes, a serious problem when building regexp automata, I chased it yesterday
in the train, found and fixed it. I'm just surprized it didn't break way
more things, commited in revision 3705. I suggest rerunning your regression
tests if possible as this is really a core change, i think it's safe, but
need to be heavilly tested,

Daniel

-- 
Red Hat Virtualization group http://redhat.com/virtualization/
Daniel Veillard      | virtualization library  http://libvirt.org/
[EMAIL PROTECTED]  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine  http://rpmfind.net/
_______________________________________________
xml mailing list, project page  http://xmlsoft.org/
[email protected]
http://mail.gnome.org/mailman/listinfo/xml

Reply via email to