On 08/01/18 13:21, Stian Soiland-Reyes wrote:
On Sun, 7 Jan 2018 15:07:00 +0000 (UTC), "Andy Seaborne (JIRA)"
<[email protected]> wrote:
{{<app:://4a8a0ff8-eab0-3930-8257-017dec3a8356/manifest0.xml>}} is a good IRI.
Note "::" -- the stack trace has ":" -- both are legal.
My mistake in Jira :) Are you sure "::" would be legal anyway?
I tried with ":"
"::" gets into interesting territory (":" in the URI path is legal under
certain conditions to differentiate as a relative URI from a scheme name).
And then there are DOS drive letters.
{noformat}
public static void main(String... args) {
IRIFactory f = IRIFactory.iriImplementation();
IRI iri =
f.create("app://4a8a0ff8-eab0-3930-8257-017dec3a8356/manifest0.xml");
IRI iri2 = iri.resolve("./BorisEJB.xml");
System.out.println(iri2);
Iterator<Violation> vIter = iri2.violations(true) ;
for ( ; vIter.hasNext() ; )
System.out.println(vIter.next().getShortMessage()) ;
}
{noformat}
There is a warning.
That's what I would have thought.
But rules seems stricter when parsing RDF/XML as it uses a differently
configured IRIFactory and deliberately bails out on the IANA violation.
Which one is it using? And it may be different if via RIOT :-|
2 (iriImpl and semanticWebImpl) of the default configuration of the 3
std factories seem to give UNREGISTERED_IANA_SCHEME as a warning.
Which is why the misalignment warning -> error below may be a factor.
Also - age.
Andy
I'll try to make a separate test and report on dev@jena
to see if this is intentional or not.
But
{noformat}
at
org.apache.jena.rdfxml.xmlinput.impl.ARPSaxErrorHandler.error(ARPSaxErrorHandler.java:37)
at
org.apache.jena.rdfxml.xmlinput.impl.XMLHandler.warning(XMLHandler.java:196)
{noformat}
does not agree with the current Jena code.
Hmm.. perhaps the classpath didn't fully update? In Eclipse it seemed
like it only had the new Jena version, but of course Maven might think
otherwise. Will triple-check with mvn help
