Hi Rupert,
I do not understand your example. The "bNode" and "bNodeClone" are not
> the same. Even that your force them to return the same hashCode, they
> are in fact different BNode instances and therefore MUST NOT be
> treated as the same Resource in a Graph.
>
No, Two BNode are equals if the equals method returns true. This is
trivially the case when they are same instance but this doesn't need to be
the case. They can well internally have an ID mapping to an id of the
backend.
>
> AFAIK Clerezza does not support "ids" for BNodes, but relies on
> "identity" (instance equality - bnode1 == bnode2). So based on that
> "bode" and "bNodeClone" created by the provided test case are NOT
> equals, but in fact different resources.
>
No clerezza is defined by it's api, so any bnode subclass qualifies as
bnode. The above would be the case if and only is the bnode api would
specify that equals has to return true if and only if it's the same object
instance.
>
> What this means for the assertions in the provided test case is
> described in detail in the following statements.
>
> mGraph.add(new TripleImpl(bNode, uriRef1, uriRef2));
> mGraph.add(new TripleImpl(bNodeClone, uriRef2, uriRef3));
>
> bNode and bNodeClone are different resources
>
> NonLiteral bNodeBack = mGraph.filter(null, uriRef1,
> uriRef2).next().getSubject();
> Assert.assertEquals("The bnode we get back is not equals to
> the one we added", bNode, bNodeBack);
>
> "bNodeBack" is equals to "bNode"
>
> NonLiteral bNodeBack2 = mGraph.filter(null, uriRef2,
> uriRef3).next().getSubject();
>
> "bNodeBack2" is equals to "bNodeClone"
>
> Assert.assertEquals("The returnned bnodes are no longer
> equals", bNodeBack, bNodeBack2);
>
> but "bNodeBack" and "bNodeBack2" are not equals as "bNode" and
> "bNodeClone" are not.
>
> Assert.assertTrue("Not finding a triple when searching with
> equal bNode", mGraph.filter(bNodeBack, uriRef2, null).hasNext());
>
> also this is expected to be empty, because there is no triple such as
> "bNode", "uriRef2", "*"
>
>
> BTW: if you you change bNodeClone to "bNodeClone = bnode" the test
> runs fine (as expected)
>
> This does not mean that there is no issues related to bNodes. It just
> means that the provided example does not show the problem.
>
It does show the problem. And it's a very real problem because a clerezza
wrapper from some backend will not create inmenory mapping from orignal
bnodes to clerezza bnodes but simply wrap the bnodes and implement hashcode
and equals in a constistent fashion. So the test I added must pass for
conformant implementation.
Cheers,
Reto
>
> best
> Rupert
>
>
>
> On Fri, Jun 28, 2013 at 12:37 PM, Reto Bachmann-Gmür <r...@wymiwyg.com>
> wrote:
> > Hi Rupert,
> >
> > I've added a commented out failing test for
> > STANBOL-1130<https://issues.apache.org/jira/browse/STANBOL-1130>
> > .
> >
> > Cheers,
> > Reto
> >
> >
> > On Fri, Jun 28, 2013 at 12:23 PM, Rupert Westenthaler <
> > rupert.westentha...@gmail.com> wrote:
> >
> >> Hi Reto,
> >>
> >> I changed the comparator implementation of the IndexMGraph some weeks
> >> ago, because the old code was not working when different Resource
> >> implementations where mixed within the same graph. So this indicates
> >> indeed a bug.
> >>
> >> Can you add a failing unit test for this issue?
> >>
> >> best
> >> Rupert
> >>
> >>
> >> On Fri, Jun 28, 2013 at 12:07 PM, Reto Bachmann-Gmür <r...@apache.org>
> >> wrote:
> >> > Hello,
> >> >
> >> > Investigating why recipe-exmpansion in no longer working I found that
> it
> >> > seems to be an issue with the IndexedMGarph as things work when using
> >> > SimleMGraph instead. It used to work till a couple of week ago.
> >> >
> >> > Now it seems that a bnode is no longer identical to itself when
> occuring
> >> in
> >> > different triples. While I'm investigating further I though I ask if
> >> > someone else noticed problems like that.
> >> >
> >> > Cheers,
> >> > Reto
> >>
> >>
> >>
> >> --
> >> | Rupert Westenthaler rupert.westentha...@gmail.com
> >> | Bodenlehenstraße 11 ++43-699-11108907
> >> | A-5500 Bischofshofen
> >>
>
>
>
> --
> | Rupert Westenthaler rupert.westentha...@gmail.com
> | Bodenlehenstraße 11 ++43-699-11108907
> | A-5500 Bischofshofen
>
