Greetings readers,
just to add my take on this..
I use ant extensively and with saxon too. My first attempts used the ant <xlst>
task, but that fell over when my stylesheets needed to use xslt2.0.
So I switched to <saxon-xslt>, except that then I could not switch on
line-numbers and tracing. After a discussion on the Saxon list Mike Kay updated
Saxon so that the remaining gotchas from the <xslt> task were covered. This
rendered the <saxon-xslt> task as deprecated. So I have now switched back to
the <xslt> task. Whereupon I also found validation did not work. My validation
task calls java direct.. Just at the moment it is not a satisfactory solution
as the java task does not fail even though the validation failed. I think I
remember reading something about that ages ago... Note that my task is actually
a macro which does other validation too..
<java fork="true" classname="com.saxonica.Validate" failonerror="true"
resultproperty="rcodeSaxonValidate">
<classpath>
<pathelement location="${SaxonJar}"/>
</classpath>
<arg value="-s:@{in}"/>
<arg value="-xsd:@{xsd}"/>
<arg value="-val:@{validationlevel}"/>
</java>
I include here Mike Kay's offline assessment of using saxon validate inside ant.
>It looks as if this task relies on the validation engine implementing
>XMLReader - that is, it's assuming validation is something that parsers do.
>I would have expected something that expects to work with any validation
>engine to use javax.xml.vaidation.Validator, or ValidationHandler, both of
>which Saxon does implement. I guess it's easy enough to provide a wrapper
>that implements XMLReader if people need it.
>
>It's just a shame that ValidatorHandler is a concrete class rather than an
>interface. Otherwise I could have made ValidatorHandlerImpl extend XMLFilter,
>>and you would get an XMLReader for free.
>
>Regards,
>
>Mike
>>Hi Mike, this looks like a similar problem that I have faced (and run
>>away from).
>>
>> I reverted in the end to calling saxon directly from the command
>>line for validation..
Hope this helps
Cheers
Graham
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
