document.evaluate and createNSResolver miss on namespaced elements

• Component: DOM
• Status: RESOLVED FIXED

[Paul Everitt]

I am using the nightly builds to gain JS access to XPath selection of nodes.  In the test case I create an XHTML2 using a string in JS and calling the DOMParser.  To match on namespaced elements, I create a resolver with xmldoc.createNSResolver and pass it into the xmldoc.evaluate call.

Unfortunately the XPath selection matches nothing in WebKit.  It works in Firefox.

[Paul Everitt]

Created attachment 8457 [details]
createNSResolver test case

A test case with HTML and JS that works in Firefox but fails in the WebKit nightly from May 20 2006.

[Anders Carlsson]

Created attachment 8680 [details]
Patch

[Darin Adler]

Comment on attachment 8680 [details]
Patch

-Step::Step(AxisType axis, const String& nodeTest, const Vector<Predicate*>& predicates)
+Step::Step(AxisType axis, const String& nodeTest, const Vector<Predicate*> &predicates)

The above is not a good change.

+    Parser *parser = Parser::current();

We put the * next to the type.

+    if (!parser->m_currentNamespaceURI.isNull()) {
+        m_namespaceURI = parser->m_currentNamespaceURI;
+        parser->m_currentNamespaceURI = String();
+    }

Why the if here? Seems that check should be omitted.

+            $$ = new String($1->deprecatedString().mid(colon + 1));

This could use String::substring and then you would not need to bother with the conversion to/from DeprecatedString.

r=me

[Darin Adler]

I got an idea:

-    if (m_nodeSetPosition + 1 >= m_nodeSet.size())
+    if (m_nodeSetPosition + 1 > m_nodeSet.size())

This could also be fixed by removing the +1 and keeping the =.

[Paul Everitt]

I just tried the test case in the most recent nightly and it now passes.  Thanks Anders!

[Lucas Forschler]

Mass moving XML DOM bugs to the "DOM" Component.