Re: The identity transform and attributes

On 21/01/2008, Abel Braaksma <abel.online@xxxxxxxxx> wrote:
> > David Carlisle wrote:
> > attribute::node() and self::node() both select attribute nodes.
> > node() on its own is short for child::node() which does not select
> > attribute nodes because there are never any attributes in the child
> > axis, it is teh child axis that does not include attributes, not the
> > node() node test.
>
> That's a great insight.

+1 to that.

I guess this explains why the allowed axes in match patterns are child
and attribute.  So when you have:

match="foo"

you really have:

match="child::foo"

...to distinguish between foo being an attribute or element?  The same
logic applies to node() in that:

match="node()"

is really

match="child::node()"

so to get attributes as well you need to explicitly say that:

match="attribute::node()|child::node()"

...so the only way for match="node()" to match all node types
including attributes would be if node() wasn't automatically converted
to child::node(), but some other all encompassing axis.

Assuming this is correct so far, the real mind bender is:

apply-templates select="parent::foo"

being matched by:

template match="child::foo"

Is that clear is anyones mind?  If so, please share :)


thanks
-- 
Andrew Welch
http://andrewjwelch.com
Kernow: http://kernowforsaxon.sf.net/