Re: [xml] Node structure questions...

[Daniel Veillard <veillard redhat com>]

On Thu, Apr 03, 2003 at 12:55:02PM -0800, Jerome Pesenti wrote:
> Greetings,
>
> Why is the value of an attribute a list of nodes? Is

 yes

> it to be able to represent the succession of text
> and entity references? If so, wouldn't it be more

  also because it matches the XPath datamodel.

> space savy to have a light structure for those instead
> of the full xmlNode structure (10 pointers of so)?

  I have a lighter structure for xmlNs and it's a MAJOR
headache for XPath processing.

> Also, is that an actual XML requirements, can't the
> entity references be all resolved?

  No, libxml was designed as an editing toolkit, not a mere parsing
tool.

> Doing some profiling on libxml, it seems that building
> a tree without attributes is very fast but that having
> a lot of attributes is very costly.... I just thought
> I would mention it....

  Attribute processing is more complex than standard node 
processing, they have types, their content needs to be 
modified, there is defaulting, etc ... Adding an attribute
to a tree is at least as expensive as adding a sub element.

Daniel

-- 
Daniel Veillard      | Red Hat Network https://rhn.redhat.com/
veillard redhat com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/