Re: "Re: [xml] XSLTPROC is slow along with XSLT"

[Daniel Veillard <veillard redhat com>]

On Mon, Feb 16, 2004 at 03:00:23PM +0100, Kasimier Buchcik wrote:
> Hi,
>
> on 2/16/2004 12:19 PM Daniel Veillard wrote:
> > On Mon, Feb 16, 2004 at 06:59:36PM +0800, Vikrant Rathore wrote:
> > Content-Description: Mail message body
> >
> > > Hi daniel,
> > >
> > > Thanks for a prompt response. I did not expect so fast reply. The data file is 
> > > attached with this mail and the version info is
> >
> >
> >    Also XSLT requires to build a full tree for the input file,
> > XSLT is not designed to handle very big files, in general. It's
> > pure general basic information about XSLT.
>
> Just for the sake of information:
> In contrast m$xml seems to handle big files much faster, but from what I 
> learned this is due to the fact that it transforms to a stream. So if we 
> want a really fast xslt that handles big files, someone has to implement
> transformation that does not build a tree.

  There is a subset of XSLT stylesheets that theorically don't require a 
full tree and can be streamed. I know some people work on those tools :-)
But this is very specialized, you may change one line in the stylesheet and
suddenly this won't work so it's a bit risky to rely on this.
  I don't know if MSXML really tries to stream, maybe, I think that for
those class of transformation a good old Python script based on 
a reader and possibly on XPath is probably a simpler and more solid
solution to such problems :-)

Daniel

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