Subj : Different handling of XML literals in Spidermonkey and Rhino
To   : netscape.public.mozilla.jseng
From : Martin Honnen
Date : Sat Mar 26 2005 06:36 pm

Section 11.1.4 XML Initialiser of E4X says:

An XML initialiser is an expression describing the initialization of an 
XML object, written in a form of a literal. It may specify an XML 
element, an XML comment, an XML PI, or a CDATA section using ordinary 
XML syntax.

Consequently I have tried the following with Spidermonkey, using 
literals for the different kind of nodes:

js> var god = <god>Kibo</god>;
js> god.nodeKind()
element
js> var comment = <!-- Kibology for all. -->;
js> comment.nodeKind()
comment
js> var pi = <?process Kibology="on"?>;
js> pi.nodeKind()
processing-instruction
js> var cdata = <![CDATA[Kibology for all.<br>All for Kibology.]]>;
js> cdata.nodeKind()
text

Thus with Spidermonkey it seems a literal doesn't have to be well-formed 
markup in the sense of the XML specification (where you would always 
need to have a root element) but you can use a snippet of markup for the 
specific node types.

However when I try the same with Rhino (Rhino 1.6 release 1 2004 11 30) 
I am only able to use a literal for an element successfully, for the 
other attempts to have a literal solely creating a comment or a 
processing instruction of a CDATA text node Rhino gives errors or hangs:

js> var god = <god>Kibo</god>;
js> god.nodeKind()
element
js> var comment = <!-- Kibology for all. -->;
comment.nodeKind()
js: "<stdin>", line 3: uncaught JavaScript runtime exception: TypeError: 
Cannot call method "nodeKind" of undefined

js> var pi = <?process Kibology="on"?>;
pi.nodeKind()

hangs

js> var cdata = <![CDATA[Kibology for all.<br>All for Kibology.]]>;
cdata.nodeKind()

hangs


Trying to understand which behaviour is correct I have looked closer at 
section 11.1.4 XML, the production for XML initialiser is

XMLInitialiser :
   XMLMarkup
   XMLElement

so any attempts not to use an element in the literal should be covered 
by the production XMLMarkup on which the sections says:

The production XMLInitialiser : XMLMarkup is evaluated as follows:
1. Let markup be a string literal containing the same sequence of 
characters as XMLMarkup
2. Return a new XML object created as if by calling the XML constructor 
with argument markup (section 13.4.2)

which then means the markup string is passed to ToXML as defined in 
section which would require to parse
   <parent xmlns="currentdefaultnamespaceURI">markup</parent>

Looking at that it seems Spidermonkey behaves correctly while Rhino has 
bugs, it should be able to handle the literals/initializers for 
comments, processing instructions, and cdata sections.

However, I am not sure how the default for 
XML.ignoreComments/ignoreProcessingInstructions come into play here and 
what is really supposed to happen.

What do other think, does Rhino have a bugs? For instance when I change 
the ignoreComments setting I still get the error:

Rhino 1.6 release 1 2004 11 30
js> XML.ignoreComments
true
js> XML.ignoreComments = false;
false
js> var comment = <!-- Kibology for all. -->;
comment.nodeKind()
js: "<stdin>", line 3: uncaught JavaScript runtime exception: TypeError: 
Cannot call method "nodeKind" of undefined

-- 

	Martin Honnen
	http://JavaScript.FAQTs.com/

.