Re: [xslt] scoped result tree fragments

[Daniel Veillard <veillard redhat com>]

On Thu, Jul 08, 2004 at 02:08:02PM -0700, Mark Vakoc wrote:
> 
> Seems like there is a lot of opportunity for libxslt to make use of the
> dictionary.  Since the xslt RVT, transform context, and result document(s) all
> share the same dictionary use cases like mine where a lot of external data from
> extensions is used would benefit the most.

  Actually using the dictionnary for names internal to libxslt processing
is what boosted performances most, for example finding a variable from
the current set is done with pointer equality tests instead of !strcmp()
I think template matching tests too. String interning #1 effect is not
memory consumption, it's quicker comparisons (assuming you're always from
the pool and a pool never deliver duplicates).

> I would guess areas like xsltCopyNode, xsltCopyText, etc.. would be the first
> to attack.  And rumor is you take patches.  Any immediate concerns?

  The only concern is that I think I tried to switch xsltCopy* then
ran into troubles so switched back. But it was when I was just starting
using dicts with libxslt . Now is a good time to restart this, as things 
look stable now, and sure we take (good) patches :-) !

  thanks,

Daniel

-- 
Daniel Veillard      | Red Hat Desktop team http://redhat.com/
veillard redhat com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/