Arndt Jonasson <arndt.jonasson@g...> wrote in
<1183642341.069392.220360@w...>:
> Say I have an XML document, where I want the contents of
> certain elements
> to be a textual pointer to another element, based on the
> contents of yet another element. Somewhat silly example:

[snip]
 
> The string "/top/friends/friend[name1=$./../name]" could
> be what I want, except that the "$." thing is an
> invention.

XSLT defines several extensions to XPath, including
current() function. Taking a look at XPointer/XLink/XQuery
is probably a much better idea, though.

However. Are you certain you want to do something like that
at all? That is, do you want to allow pointers to anything
in the document (and perhaps *beyond* the document, if you
allow document() or somesuch as well)? Perhaps it would be
a better idea to define a more structured grammar
for 'links', and let your application logic handle the
actual processing.

For example:

<moredata><friends/></moredata>

Upon seeing that, your applications would know they should
look up the corresponding friends. If you needed to link
something else as well, you would define more elements and
the logic for processing them as well:

<moredata><friends/><borks foobarcity="42"/></moredata>

(Supposedly linking all the Joe's borks with foobarcity of
42 or more... whatever that means.)

-- 
Pavel Lepin