Stop Xalan Transform From Resolves Entities

Question

I'm trying to write a DOM tree to an XML file, so I've been using Xalan to do a transformation. My code is quite simple:
DOMSource source = new DOMSource( domDocument );
StreamResult result = new StreamResult( new File( "out.xml" ));
TransformerFactory tf = TransformerFactory.newInstance();
Transformer transformer = tf.newTransformer();
transformer.setOutputProperty( OutputKeys.METHOD, "xml" );
transformer.setOutputProperty( OutputKeys.ENCODING, "utf-8" );
transformer.transform( source, result )
The problem is that my DOM document has character entity references that the transformer seems intent on converting. For example XML that looks like this in the DOM tree:
<value>Here is a weird character: &eacute;</value>
Gets output like this in the xml file:
<value>Here is a weird character: ?</value>
Since I'm not actually transforming anything, just outputting the XML, I need to preserve any character entity references. Oddly enough the source XML file contains the &amp; entity, which gets converted to a regular & character when the file is parsed into the DOM tree. It's the only character entity that isn't stored as an actual EntityReference object in the tree, but the only one that gets converted back to &amp; on a transform. Any help would be greatly appreciated.

Answer 1

Yes, the character entities are declared in the DTD. At one point in development I was including it with:
transformer.setOutputProperty( OutputKeys.DOCTYPE_PUBLIC,domDocument.getDoctype().getPublicId());
transformer.setOutputProperty( OutputKeys.DOCTYPE_SYSTEM,domDocument.getDoctype().getSystemId());
transformer.transform( source, result );
(I'm writing this code from memory; sorry if there are any syntax errors)
The code compiled and did correctly add the DOCTYPE entry to the file, but the references are still being converted. It does matter for us that the references remain intact. Eventually the XML will go into a publishing system that uses a different set of proprietary character codes for that software; utf-8 won't work :-(