The Document Object Model (DOM) provides a standard interface for interacting with XML documents. Xalan-Java supports this interface. For large XML documents, however, this may involve considerable overhead, since Xalan-Java must create one or more Java objects for each Node in the document DOM tree.

For the majority of cases -- your input and output are URLs, files, or streams, and you use the default Liaison (DTMLiaison class) and XML parser -- Xalan-Java avoids this overhead by implementing the Document Table Model (DTM), a "pseudo-DOM" that uses integer arrays in place of the DOM. For larger input and output trees, the performance improvements can be very significant.

The XSLTProcessorFactory static getProcessor() method sets up an XSLT processor to use the default DTM liaison and XML parser. If your input or output turns out to be a DOM Node, the XSLTProcessor automatically switches to the XercesLiaison class and the Xerces DOM parser. You can explicitly instantiate an XSLTProcessor to use XercesLiaison and the Xerces DOM parser as follows:

org.apache.xalan.xslt.XSLTProcessor xsltProc = org.apache.xalan.xslt.XSLTProcessorFactory.getProcessor(
         new org.apache.xalan.xpath.xdom.XercesLiaison());

For more information, see Processing DOM input and Producing DOM output.

The DTM parser always attempts to expand eentityreferences in the source tree and stylesheet tree. If you attempt to turn off entity reference expansion -- with DTMLiaison setShouldExpandEntitityRefs(false) -- DDTMLiaisondoes not make the setting and issues a message to that effect. If you want to turn off entity rreferenceexpansion, instantiate the XSLTProcessor to use XercesLiaison and the Xerces DOM parser.