[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

Invalid characters in Java strings retained when marshallingJAXB-bound o

• From: Tony Nassar <tnassar@palantir.com>
• To: "'xml-dev@lists.xml.org'" <xml-dev@lists.xml.org>
• Date: Sun, 8 Aug 2010 11:18:10 -0700

Forgive me for the possible reposting; my company's domain name changed, and I don't think this ever appeared on the list.

I was surprised to find that no exception is thrown when strings containing characters invalid in XML (e.g. \x00, \x01, \x0c, etc.) are serialized to XML by means of the JAXB API. My code is simple boilerplate, but I'm including it here, in case it matters (Groovy, not Java; though it shouldn't matter). I'm defaulting to the version of Xalan and Xerces that ship with the JDK. I've attempted to marshal into a SAXResult, and then go through an identity transform I got from Saxon, but since there's no XML input to be parsed, the invalid characters are never detected.

Of course I can write a CharacterEscapeHandler, I'm just surprised that I have to. Am I doing something wrong, is this a bug in the JAXB implementation, or is this the expected/desired behavior?

Marshaller m = context.createMarshaller()
m.setProperty( Marshaller.JAXB_FORMATTED_OUTPUT, true)
m.setProperty(Marshaller.JAXB_ENCODING, charset.name()) 
m.setProperty("com.sun.xml.bind.characterEscapeHandler", new NioEscapeHandler(charset.name()))
def handler = factory.newTransformerHandler()
def f = new File(output, "jaxb (${charset.name()}).xml")
handler.setResult(new StreamResult(f))
f.withOutputStream { stream ->
	// Supposedly, marshalling to a Stream will cause the XML declaration to be omitted. That's not actually happening!
	m.marshal(testStrings, handler)
}