Subj : Re: Embedding E4X support with SpiderMonkey
To   : netscape.public.mozilla.jseng
From : Peter Wilson
Date : Wed May 04 2005 08:27 pm

Brendan Eich wrote:
> Peter Wilson wrote:
> 
>> For example:
>>  >     new XML('<mytag></mytag>')
>>  > claims an XML syntax error, possibly down to the XML namespace being
>>  > unsupported.
>>
>> Answered part of this myself : JS_SetOptions(..., JSOPTION_XML) which 
>> seems to get the basic XML engine working.
> 
> 
> 
> Yes, that's required for backward compatibility.  JS handles the
> 
> <!-- comment hiding hack
> script goes here
> // -->
> 
> trick I perpetrated in 1995 to get inline scripts wedged into HTML 
> without showing in older browsers than Netscape 2 (beta, at that time).
> 
> 
>> For further clarification - my existing DOM/Xpath without having to
>> rebuild another shadow of the tree. (Note - my XML parser etc is
>> written in C++ and integrated as a set of native classes that shadow
>> parts of that tree on demand). Having to serialise that back to text -
>> then have the SpiderMonkey engine re-parse it into a tree isn't going to
>> be efficient, especially in a server-side application.
> 
> 
> 
> See https://bugzilla.mozilla.org/show_bug.cgi?id=270553.  I have a patch 
> I've been neglecting on account of higher priority work, that adds an 
> API to Gecko's DOM.  It implements the rest of ECMA-357 -- new 
> XML(document), .domNode/.domNodeList, etc.  You could use it for the 
> same purpose with your DOM impl, but I'm sure it will need some revamping.
> 
> How about I get the patch into that bug and get the DOM hookup working 
> with Gecko, then you try that E4X/DOM API out with your DOM code?

Sounds like a great starting point.

Will the API cover the XPath selection as well? I effectively have 
analogues of XML (XmlGeneric and a set of sub-classes) and XMLList 
(XmlNodeSet) in my code - node effectively has node.xpath and 
node.select to query the tree. It takes an entire XPath expression and 
creates an XmlNodeSet. If anyone is interested:

http://www.whitebeam.org/library/guide/TechNotes/xmlproc.rhtm
http://www.whitebeam.org/library/guide/TechNotes/xpath.rhtm

I only create JS instances of XML elements when referenced by JS code so 
I don't have to duplicate the entire tree in JS if the script is only 
going to change one node.

I guess ideally an API that passes over the XPath query and accepts a 
node-set back would be good. It's a pity the E4X doesn't use standard 
XPath syntax - it may have been better to have a '.select' method and 
stick with the syntax than invent something new.

If you let me know when you've got the patch ready I'll take a look and 
work out how to proceed.

As always, thanks very much Brenden
Pete

> 
> /be

.