Bob Jamison wrote: > Works great! We replaced the Xerces1.4 jar in our app, and > it worked with only minor modifications. So it is definitely > used in in a real-world app here.
Cool. > The only major difference is how we access TEXT_NODEs > within an element. For example, if you had a tag like: > [...] There have been some bug fixes in the DOM implementation that is in Xerces 1.x that have not been merged into the Xerces2 codebase. So I went ahead and synchronized the source from Xerces 1.x into the Xerces2 branch in CVS. > ..as now it apparently treats TEXT_NODEs, > ENTITY_REFERENCE_NODEs, and CDATA_SECTION_NODEs > separately. Once we take this into account, everything works fine. > Is this proper operation? Well, certainly CDATA sections and text nodes are kept as separate nodes from each other. (This was the reason for having the setCoalescing method on the JAXP DocumentBuilder interface.) However, the """ entity refs appearing in your document instance are not reported as general entity references. So they should not be represented as entity ref nodes in the tree. But I checked the code and Xerces2 had a bug relating to the entity ref nodes. We added an option that character entity references and built-in entity refs (e.g. """) could be communicated from the scanner. The default values for these should have been false but, due to a bug, the built-in entity refs were getting to the outside, ignoring the default value of this feature. I've checked in the fix for this bug into CVS. If you are able to extract and build Xerces2, I'd be interested in knowing if your problem went away. -- Andy Clark * IBM, TRL - Japan * [EMAIL PROTECTED] --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
