Re: [xsl] Optimizing Trax

["Mark R. Diggory" <mdiggory@xxxxxxxxxxxxxxxxx>]

Pietschmann is quite right here about DOM memory usage, but, even if you 
use SAXTransformer, SAXSource and SAXResult, I still believe there is a 
DOM tree generated internal to the transformer to handle the walking the 
tree on various axis. I sure different implementations approach dealing 
with this with different optimizations. "Sources" and "Results" really 
only provide alternate means of IO and not underlying DOM creation.

There is a chance that instantiating the DOM and handing it as a source 
may be more memory consuming than letting the Transformer build it, but 
this would be very implementation specific.

One alternative is to use a SAX XMLFilter to process out content you may 
not need prior to doing your transform, a tool I've just started to use 
http://www.devsphere.com/xml/saxdomix/index.html for transforming/filter 
processing very large XML documents more efficiently.

It provides an a couple utilities over SAX and DOM that allow you to 
only build DOM fragments for the subtrees of the Document you'll 
actually be transforming against. It currently has a simple facility 
that reads the XSLT template for matches and keys to build a list of 
elements, it then passes matched sax events onto a DOMBuilder to build a 
DOM fragment, this DOM fragment can then can be transformed against 
using the same XSLT.

-Mark Diggory

Schwartz, Rechell R, ALABS wrote:
> Based on the various responses I received I attempted to use SAX to transform an XML file to a SAXResult, and then to read in the SAXResult as
> an input source to a second transformation, and transform the document to
> HTML to be sent to the browser. I received an error complaining that the ContentHandler of the SAXResult was null. Obviously, I don't' understand how to set/use  a ContentHandler, and I obviously don't understand how to convert a SAXResult to a SAXSource. Here is the relevant code snippet in my JSP page. Any help for a newbie like me would be greatly appreciated, as I am at my wits end. Also, if this is not the right newsgroup for this problem, please direct me to the right place.
>
> String xsl = "/stylesheets/data.xsl";
> String xsl2 = "/stylesheets/format.xsl";
> Templates templates = (Templates)application.getAttribute("templates");
> Templates templates2 = (Templates)application.getAttribute("templates2");
> if (templates == null) {
> 				TransformerFactory tfactory = TransformerFactory.newInstance();
> 				templates = tfactory.newTemplates(new StreamSource(getClass().getResourceAsStream(xsl)));
> 				application.setAttribute("templates", templates);				
> 			}
> if (templates2 == null) {
> 		        TransformerFactory tfactory = TransformerFactory.newInstance();
> 				templates2 = tfactory.newTemplates(new StreamSource(getClass().getResourceAsStream(xsl2)));
> 				application.setAttribute("templates2", templates2);
> 			}
> Transformer transformer = templates.newTransformer();
> SAXResult saxResult = new SAXResult();
> transformer.transform(new StreamSource(new StringReader(hand.detailresult)), saxResult);
> transformer = templates2.newTransformer();
> transformer.transform(new StreamSource(saxResult.getSystemId()), new StreamResult(out));
> 			
> -----Original Message-----
> From: Rob Rohan [mailto:me@xxxxxxxxxxxx] 
> Sent: Friday, June 20, 2003 3:52 PM
> To: xsl-list@xxxxxxxxxxxxxxxxxxxxxx
> Subject: RE: [xsl] Optimizing Trax
>
> On Fri, 2003-06-20 at 12:22, Schwartz, Rechell R, ALABS wrote:
>
> > How do I tranform the document using SAX-- i.e., how do I change this
> > line?
> >
> > transformer.transform(new DOMSource(document), new StreamResult(out));
>
> This is kind of more a java question than an XSLT question. Here is a
> snip-it that I use in one of my programs - it's a touch kludgey but if
> you know java you can decipher the needed parts - if you can't I
> recommend "Java & XML" second edition by Brett McLaughlin from O'Reilly
>
> public long transform(java.io.InputStream xmlin, java.io.InputStream
> xslin, java.io.OutputStream output, java.io.File xslURI, java.io.File
> xmlURI) throws Exception {
>        
> 	strResult = null;
> 	if(params == null)
> 		params = new Vector();
> 	   
>        String media = null;
>        
>        //set the factory to the specified one       
> 	System.setProperty("javax.xml.transform.TransformerFactory",  XSLTFactory);
>               
>        //get an instance and make a result object       
> 	tfactory = javax.xml.transform.TransformerFactory.newInstance();
>        
> 	strResult = new javax.xml.transform.stream.StreamResult(output);
>        
> 	stf = (javax.xml.transform.sax.SAXTransformerFactory) tfactory;
>            
>        StreamSource ss = new StreamSource(xslin);
>        if(xslURI != null) ss.setSystemId(xslURI);
>        
>        stylesheet = tfactory.newTemplates(ss);
>        
>        transformer = stylesheet.newTransformer();
>
>        //set params       
> 	for(int i=0; i < params.size()-1; i++)
>         	transformer.setParameter((String) params.elementAt(i), (String)params.elementAt(i+1));
>
>         //do the transformation  
> 	StreamSource input = new StreamSource(xmlin);
>         if(xmlURI != null) input.setSystemId(xmlURI);
>
>      
>         transformer.transform(input, strResult);
> ....
>
>
>
> > Rechell Schwartz
> >
> > -----Original Message-----
> > From: J.Pietschmann [mailto:j3322ptm@xxxxxxxx]
> > Sent: Friday, June 20, 2003 3:13 PM
> > To: xsl-list@xxxxxxxxxxxxxxxxxxxxxx
> > Subject: Re: [xsl] Optimizing Trax
> >
> >
> > Schwartz, Rechell R, ALABS wrote:
> >
> > > I am using Trax to transform my xml files, and am finding the
> >
> > performance
> >
> > > very slow especially for large files. Are there any ways of optimizing
> >
> > the
> >
> > > performance such as by pre-compiling the stylesheets?
> >
> > Yes, but your time is probably wasted elsewhere.
> >
> >
> > > transformer = tFactory.newTransformer(new 
> > StreamSource(getClass().getResourceAsStream(xsl)));
> > > transformer.transform(new DOMSource(document), new
> > StreamResult(out));
> >                              ^^^^^^^^^^^^^^^^
> > Use a profiler to check where your program wastes time:
> > - DB access or whereever you get your raw data
> > - DOM construction
> > - transformation
> > If it is the latter, check your XSL for inefficient constructs.
> > If it is DOM construction, use SAX. Also DOM uses a lot of memory,
> > which may be a problem in itself (more GC).
> >
> > J.Pietschmann
> >
> >
> > XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list
> >
> >
> > XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list


XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list