Hi Hallvard,
> > I'm just wondering why you think that a custom location step is requried.
> > e.g. your example could be reworked a little as follows...
> >
> > ext:related-thing(thing)[@color='red']
>
> When designing the XML structure you have to anticipate the navigation
> pattern, to ensure that everything interesting is underneath in the
> hierarchy. This is often not feasible, so I find it useful to have custom
> functions for jumping across the XML tree.
I wholeheartedly agree.
> Requiring that the function be
> the first step is limiting, i.e. cannot be used for just navigating along
> an idref and nothing more.
So the same function with the same arguments
could evaluate to different things in different contexts?
> In my particular application, I combine XML and javascripting, so the
> custom function can be written in javascript and embedded in the file.
Wow!
> > Or further location steps can be appended as needed such as
> >
> > ext:related-thing(thing)[@color='red']/foo/bar
> >
> > Without having to change the XPath grammar etc. Would that do what you
> > need?
>
> For a specific purpose this is usually possible, but when all navigation
> patterns are taken together, it's not always possible to design a tree
> where navigation is always downward. Hence, there is a need for custom
> "jumps" in the XML tree. In some application, the speed may also improved
> this way.
Yes, and the question is: Do you really care where you jump from?
To me, your idea sounds like invoking a method on an object,
whereas the traditional functions are more like stand-alone functions in a
programming language.
The alternative to a method invocation is to pass the "this" object in an
argument (see the Perl and the Python programming languages).
So instead of
/ path / to / somewhere /
ext:and-now-something-completely-different( 'zz' ) /
here-we-stand
You could write
ext:and-now-something-completely-different-II(
/ path / to / somewhere,
'zz'
) / here-we-stand
Provided that the /path/to/somewhere is needed for evaluating the function
in the first place.
> BTW, isn't this part of XPath 2.0, for these purposes?
Will have to re-read that spec a few times, I confess.
> Hallvard
Jan
_______________________________________________
Jaxen-interest mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/jaxen-interest
