Oh yes, I see: if your input tree is 5 levels deep then the output will be 2**5 times the size of the input. Since the XSLT code is trivial, the elapsed time is going to be dominated by serialization cost, which explains why the cost of parsing doesn't affect it much. So perhaps your figures show that the Xalan serializer is faster than the Saxon serializer, which  (if true) is interesting to know but doesn't prove very much about real workloads. To confirm this, try transforming to a dummy SAXResult and see how it affects the figures.
 
Michael Kay
http://www.saxonica.com/


From: saxon-help-bounces@lists.sourceforge.net [mailto:saxon-help-bounces@lists.sourceforge.net] On Behalf Of Jose Antonio Illescas del Olmo
Sent: 09 October 2007 10:18
To: Mailing list for SAXON XSLT queries
Subject: Re: [saxon] Where is the problem?

Thank you Mr. Kay,

  Its true that isn't a realistic transformation (can't test my real transformations under xalan because using XSLT 2.0)

  The high time isn't identity transform (you can see in my first, or second, email)

	<xsl:template match="@*|node()">
           <xsl:copy>
              <xsl:apply-templates select="@*|node()"/>
           </xsl:copy>
           <xsl:copy> <!-- another copy to increment measures -->
              <xsl:apply-templates select="@*|node()"/>
           </xsl:copy>
	</xsl:template>
            ...using a second xsl:copy to increment all measures