Hi ChengLin, just out of curiosity, have you compared the performance by testing the document with an XSL copy transformation? By this you see what performance the specific XSL transformation is requiring.
Especially with XSLT it is sometimes possible to improve the performance of a stylesheet a lot. For the performance test, you might even write yourself a small Java test called by command line, which executes the transformation. The difference between the execution time of the simple copy transformation and the more complex UOF (or Word2003) is the time you might improve by using different algorithms. Bests, Svante ChengLin wrote: > It's great idea to upgrade. > We're working on Chines UOF transformation by xslt-based filter. But the > performance is poor. Just like Word2003 XML transformation, it needs about 9s > to open/save a 16page file with Intel 1.6G DualCore CPU. :( > So does OOo have a plan to upgrade and improve the performance? > > Regards! > ------------------ > ChengLin > 2008-06-11 > > ------------------------------------------------------------- > 发件人：Vyacheslav Sedov<[EMAIL PROTECTED]> > 发送日期：2008-06-10 20:07:34 > 收件人：dev<[email protected]> > 抄送： > 主题：Re: [xml-dev] XSLTC in XML Filter? > > by the way - i can help with migration xslt-based filters to XSLT 2.0 > > with almost ten years of xslt experience > > On Tue, Jun 10, 2008 at 1:18 PM, Svante Schubert > <[EMAIL PROTECTED]> wrote: >> Hi, >> >> Ian Shields wrote: >>> On Fri, 2008-05-30 at 17:06 +0800, ChengLin wrote: >>>> Hi, >>>> >>>> Is it possible to use XSLTC [1] in XML Filter to speed up the >>>> transformations? >> It is a matter of XSLTC reliabilty, if the result has a similar quality as >> with the Xalan processor (from the same xalan.jar). I had once problem with >> the XHTML export perhaps it has changed, havn't checked for while and I must >> admit forgot about my bug-id. >> >> But instead of switching to XSLTC, I would personally rather switch to >> Saxon-B 9.0 [2] providing XSLT 2.0, XPath 2.0 and XQuery 1.0. >> >>>> [1] http://xml.apache.org/xalan-j/xsltc_usage.html >>>> >>> I have a large and somewhat complex filter and it runs much faster the >>> second time I use it after starting OOo than the first. I imagine OOo is >>> either caching something or compiling it. >>> >> Yes, there is a caching. The stylesheets are being parsed only once and than >> cached after first usage. >> >> But even Java start up might have its part. >> Not sure if OpenJDK 6.0 build b10 [3] has the same performance improvements >> as the Java SE 6 update 10 [4], but it will improve a lot in the future. >> >> Cheers, >> Svante >> >> [2] http://saxon.sourceforge.net#F9.0B >> [3] http://download.java.net/openjdk/jdk6/ >> [4] >> http://java.sun.com/developer/technicalArticles/javase/java6u10/index.html >> >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: [EMAIL PROTECTED] >> For additional commands, e-mail: [EMAIL PROTECTED] >> >> > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
