Re: Bug: (xercesc_2_5) DOMImplementationLS::createDocument / DOMWriter::serialize produces invalid output

[Gareth Reakes]

Hi,
Creating a new DOMDocument, supplying any namespace-argument different 
than zero, results in DOM-tree/serialization inconsistencies (see 
examples below).

With xml-parsed DOMDocuments, the document's root element contains the 
xmlns-attribute, if given in the xml-file. In contrary, newly created 
DOMDocuments' root elements never contain the xmlns-attribute, 
although a namespace-argument was given to createDocument, but the 
DOMWriter-serialized output shows the xmlns-attribute, as expected. 
This is inconsistent.
 

This is correct behavior. A DOM tree does not need an xmlns attribute to 
tell it what namespace an element is in. In fact, namespace attrs are a 
parse time thing. They are completely ignored after parse.

Even worse: adding the xmlns-attribute manually to the root element 
after creation, the serialized output shows the xmlns-attribute 
twice, which is illegal with respect to the XML specification.
 

This is indeed incorrect. Could you provide the xml file and code that 
produces this?

The only workaround would be to always set the namespace-argument of 
DOMImplementationLS::createDocument to zero and to add the 
xmlns-attribute manually.

Here's some examples that show the spot:
implementation: valid DOMImplementationLS-object-pointer
writer:         valid DOMWriter-object-pointer
namespace:      XMLCh-representation of a namespace name
root_name:      XMLCh-representation of the root element name
xmlns:          XMLCh-representation of "xmlns"
CASE 1:
DOMDocument *doc = 
	implementation->createDocument(namespace, root_name, 0);
DOMElement *elem = doc->getDocumentElement();
XMLCh *serialized = writer->serialize(elem);

o elem contains no attributes (e.g. elem->hasAttributes() returns 
false resp. elem->getAttributes() is empty)
o serialized shows the provided xmlns-attribute
 

This is correct behavior. The DOMWriter "fixes" the xmlns on the way out 
so you can parse it again and get the same document.

CASE 2:
DOMDocument *doc = 
	implementation->createDocument(namespace, root_name, 0);
DOMElement *elem = doc->getDocumentElement();
elem->setAttribute(xmlns, namespace);
XMLCh *serialized = writer->serialize(elem);

o elem contains the xmlns-attributes
o serialized shows the xmlns-attribute TWICE, which is invalid!
 

What is namespace in this case? Are you aware that it has to be 
something specific to count as a namespace attr? It is a common mistake 
to think that xmlns attrs are required when you are creating a document 
in memory. They are not. The serializer will sort that out for you. The 
only time you would create them is if you wanted to bind specific 
prefixes or give a specific default namespace.

Cheers,
Gareth
--
Gareth Reakes, Managing Director      Parthenon Computing
+44-1865-811184                  http://www.parthcomp.com 

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]