Re: [xsl] XPath that returns nothing

[Michael Kay <mike@xxxxxxxxxxxx>]

> Yeah it's a tricky one... I've been in the this situation a few times
> now (and fwiw is where Kernow came from).  The task is, you have 1000s
> of input files and multiple steps, where the output of each step feeds
> the next.
>
> You need to be able to run each step individually, or one file in that
> step, and of course also run all the steps in one go.  Depending on
> the number of steps, a way to run from step x onwards is also helpful.
>
> So then its a case of run-to-the-end and log issues, or fail early and
> re-run individual steps.  Personally I much prefer the latter, as
> theres a much faster turnaround for fixing and testing issues, but it
> does vary depending on the project.
Certainly both modes of processing (stop at first error, or find as many 
errors as you can in one run) are appropriate, especially when running 
tests. But I think the right approach to continuing after an error is to 
use some kind of try/catch mechanism, not to suppress the error 
checking. In XSLT 2.0 that means either using extensions, or doing the 
try/catch at the level of an entire transformation.

FWIW, I'm considering putting try/catch back into the Saxon-CE code 
base, partly because it seems essential for testing, but also because 
error recovery seems particularly important for client-side applications.

Michael Kay
Saxonica