- From: Joseph Kesselman <keshlam@us.ibm.com>
- Date: Tue, 5 Mar 2002 13:01:52 -0500
- To: "Arnold, Curt" <Curt.Arnold@hyprotech.com>
- Cc: "'www-dom-ts@w3.org'" <www-dom-ts@w3.org>
>when we attempted to use pre-release >versions of at Xalan-J 2.3, there were still problems with the large number >of document() calls being made in the transforms Current versions of Xalan should be able to handle many more documents at once than earlier versions, so I do think it's worth trying again. (If you're using global variables I'd recommend trying the current CVS code, since I just discovered and fixed an annoying bug in 2.3.1's handling thereof.) If we still fail please do post a bug report, with a pointer to the current testcase source so whoever tackles this can reproduce the failure on their own machine. Possibly relevant tip: If you're using for-each to step through a series of documents, there's a special-purpose optimization in Xalan which lets you tell it to discard these documents from memory (and recover the storage) after that iteration of the for-each loop is complete, rather than caching them for possible re-examination as is the default. To select this mode of operation, insert the <?xalan:doc-cache-off?> processing instruction as an immediate child of the <xsl:for-each/> element. This PI should be harmless for other XSLT processors. (Yes, we're aware that a more centralized and automated approach to managing the document cache is needed.) ______________________________________ Joe Kesselman / IBM Research
Received on Tuesday, 5 March 2002 13:02:40 UTC