Hi, I'm looking for ways to improve the performance of processing rather large instances (> 10MB). Currently takes ~ 1 minute on workstation type hardware.
After profiling a bit I noticed most of the time is spent in a child of XSLTEngineImpl::getSourceTreeFromInput namely parserLiaison.parseXMLStream. I noticed a little deeper that XalanSourceTreeContentHandler::characters maintains whitspace characters that end up being text nodes. I was thinking I'd like to drop them by preventing calls to m_document->createTextNode. It seems that would speed things up a lot since most of the time spent seems to be in node creation. But I get the sneaking suspicion that I'm I missing the boat on something. Is there another way to prevent the creation of whitspace nodes or any other low hanging fruit I'm not aware of? Thanks Jim
