The first line in the Call Tree should have a "Tree Time %" of 100%.
If you do profiling while debugging, that might throw off the timer
if you pause and continue. Otherwise, we'd be interested if you could
post the XSL and XML that caused this. There is a little attachment
checkbox above the field you type in; you can use it to attach the
documents. Or, you can email them to stylus-field-report (at)
datadirect.com and we'd be happy to take a look and give you an
explanation.
Subject:Strange Author:Valérie Schneider Date:11 Apr 2005 04:32 AM
I just do profiling, not debugging.
Here are the XSL and the XML.
The stylesheet is word2html.xsl, which transforms WordprocessingML files to HTML.
It's a component of the Microsoft Word 2003 XML Viewer.
Subject:Strange Author:Tony Lavinio Date:12 Apr 2005 11:29 AM
This is a very interesting problem, and not one that can easily be solved - and here's why. The XSLT you have has lots of variable definitions at the top level - outside of any template. Since XSLT is declarative, the implementation defines the order in which steps are executed, and in this case, it appears that Xalan-J (and Saxon and even our internal engine) process those _before_ starting the root template. So the execution times for them aren't under <xsl:template match="/">. It's like a tree with multiple roots. Unfortunately, there isn't much we can do, since we can only report the time as the XSLT engine gives it to us. However, the various times for the individual trees rooted at each xsl:variable are themselves accurate.
Subject:Strange Author:Valérie Schneider Date:13 Apr 2005 04:10 AM
I thought it might be because of the top-levels variables.
But this is the only input file that gives such a result.
I do profiling with the same stylesheet and others xml files and the first line in the Call Tree has always a "Tree Time %" of 100%.
If Xalan-J processes top-levels variables before starting the root template, I should see this result fort almost all input files (there is only 6 of the 117 top-levels variables which depend on the source tree, the others are fixed).
Attached another input file, if you're interested.
I will try to obtain the times spent on each top-level variables.