RE: [xsl] Elements and functions available

["Michael Kay" <mike@xxxxxxxxxxxx>]

> Out of interest, can someone comment on the cost of calling 
> these functions?  I imagine element-available() is very 
> cheap, but I'm wondering about function-available() with user 
> created extensions...

Performance is always processor-dependent. If the argument to
function-available is a literal, then Saxon is going to optimize away
something like

if (function-available('f:x')) then f:x() else ()

at compile time either to a call on f:x() or to ().

On the other hand, with a run-time argument, function-available() in Saxon
is probably very expensive. Much the same applies to element-available,
since Saxon does allow user-written extension elements.

Michael Kay
http://www.saxonica.com/

> 
> Given a transform that makes use of many custom extensions, 
> is it reasonable to do:
> 
> <xsl:choose>
>    <xsl:when test="function-available( )">
>      normal execution
>    </
>    <xsl:otherwise>
>      test execution
> 
> ...to allow you to run the transform away from the host application?
> Or would all the calls to function-available() adversely 
> affect normal running of the transform?
> 
> The alternative (and good practice anway I think) is to move 
> all of the calls to extensions into one or maybe two physical 
> stylesheets, and then you can use an alternative primary 
> stylesheet that imports a different stylesheet where the 
> calls to the extensions have been replaced with sample values....
> 
> Either way, I'm just wondering if I should try to avoid lots of
> function-available() calls?
> 
> 
> 
> 2008/10/22 Jesper Tverskov <jesper@xxxxxxxxxxx>:
> > I have tested the little used XSLT functions element-available() and
> > function-available() and made a stylesheet that can create a report 
> > for most XSLT processors, testing also for EXSLT extensions 
> and Saxon 
> > extensions.
> >
> > I have made 8 reports available and a summary, "Elements 
> and functions 
> > available in XSLT processors", 
> > http://www.xmlplease.com/elements-functions.
> >
> > Feel free to suggest improvements.
> >
> > Best regards,
> > Jesper Tverskov
> >
> > http://www.xmlplease.com
> >
> >
> 
> 
> 
> --
> Andrew Welch
> http://andrewjwelch.com
> Kernow: http://kernowforsaxon.sf.net/