[Owlim-discussion] deleting a statement corrupts repository -- possible indexing problem

[Benno Blumenthal]

Dear Damyan, Barry & Ivan,

We are having problems with removing statements from a context. We are
using SwiftOWLIM-3.3 (owlim-3.3.jar & trree-3.3.jar) and
openrdf-sesame-2.3.1-onejar.jar.

In general we are having problems that the repository acts oddly after
we delete statements, i.e. query results are wrong.    In particular,
we have a clear test case which shows wrong behavior.  We are hoping
fixing this particular problem solves the more general issue.

Essentially what happens is that we remove one statement from a
context that contains multiple statements, after which the repository
is corrupted -- a trig dump shows all the triples in the proper
context, while a request for contexts claims that context is missing.
This state is preserved, i.e. we can reopen the repository and see the
same results.  If we read the dumped trig file into a new repository,
the repository is again correct.

Our guess is this is indexing, since presumably the trig file dump
does not need the indexing, and it has correct information.  But you
know better than us.


The test code is at

http://scm.opendap.org/svn/trunk/olfs/src/opendap/experiments/RepoTest.java


-- it reads in a trig file with four statements in a context, deletes
one, and shows the inconsistent state immediately as well as read back
in, as well as the correct state from reading the trig file.

More Particularly,

The  RDF file called test.trig  has a context
<http://iridl.ldeo.columbia.edu/ontologies/rdfcache.owl#startingPoints>

This context includes 4 statements:
<http://iridl.ldeo.columbia.edu/ontologies/rdfcache.owl#startingPoints> {
  <http://iridl.ldeo.columbia.edu/ontologies/rdfcache.owl> a
<http://iridl.ldeo.columbia.edu/ontologies/rdfcache.owl#StartingPoint>
.
    
<http://test.opendap.org:8090/opendap/ioos/200803061600_HFRadar_USEGC_6km_rtv_SIO.ncml.rdf>
a <http://iridl.ldeo.columbia.edu/ontologies/rdfcache.owl#StartingPoint>
.
    <http://iri.columbia.edu/~benno/opendaptest/daptestpreload.owl> a
<http://iridl.ldeo.columbia.edu/ontologies/rdfcache.owl#StartingPoint>
.
    <http://test.opendap.org:8090/opendap/ioos/ECMWF_ERA-40_subset.ncml.rdf>
a <http://iridl.ldeo.columbia.edu/ontologies/rdfcache.owl#StartingPoint>
.
}
After loading the test.trig in the repository, we read the repository
and got back the four statements in the context.

However, after dropping one statement, the context list is empty.

But the three not dropped statements do appear in the dumped trig
file, which also shows the correct context as being there.

<http://iridl.ldeo.columbia.edu/ontologies/rdfcache.owl#startingPoints> {
  <http://iridl.ldeo.columbia.edu/ontologies/rdfcache.owl> a
<http://iridl.ldeo.columbia.edu/ontologies/rdfcache.owl#StartingPoint>
.
    
<http://test.opendap.org:8090/opendap/ioos/200803061600_HFRadar_USEGC_6km_rtv_SIO.ncml.rdf>
a <http://iridl.ldeo.columbia.edu/ontologies/rdfcache.owl#StartingPoint>
.
    <http://iri.columbia.edu/~benno/opendaptest/daptestpreload.owl> a
<http://iridl.ldeo.columbia.edu/ontologies/rdfcache.owl#StartingPoint>
.
}

Again, this is our clear example -- we think queries/inferencing also
go astray in our real (production) code.

Benno Blumenthal

Haibo Liu
Nathan Potter

-- 
Dr. M. Benno Blumenthal          be...@iri.columbia.edu
International Research Institute for climate and society
The Earth Institute at Columbia University
Lamont Campus, Palisades NY 10964-8000   (845) 680-4450

RepoTest.log.gz
Description: GNU Zip compressed data

_______________________________________________
OWLIM-discussion mailing list
OWLIM-discussion@ontotext.com
http://ontotext.com/mailman/listinfo/owlim-discussion