Hi Boris,
The execution of the query plan is done by the ParallelEvaluationStrategy:
https://github.com/apache/incubator-rya/blob/master/sail/src/main/java/org/apache/rya/rdftriplestore/evaluation/ParallelEvaluationStrategyImpl.java
It looks like whoever did the rdf4j upgrade updated that class to take the
dataset object in its constructor, but it doesn't seem to be used anywhere
in query evaluation. That would be the class to modify, but I'm still not
clear on what the changes would look like since I'm not familiar with what
the contract is supposed to be.
On Tue, Dec 4, 2018 at 9:45 AM Boris Pelakh <boris.pel...@semanticarts.com>
wrote:
> So, a /sparql?query=SELECT...WHERE...&default-graph-uri=my-named-graph is
> logically equivalent to
>
> SELECT ...
> FROM <my-named-graph>
> WHERE ...
>
> (overriding whatever FROM/FROM NAMED clauses actually specified in the
> query) I looked at
> org.apache.rya.rdftriplestore.RdfCloudTripleStoreConnection.evaluateInternal(),
> and the dataset containing the overriding graph names is passed in, but
> does not seem to be honored, i.e if I have the following data:
>
> ns:g1 { ns:p1 ns:pred ns:o1 }
> ns:g2 { ns:p2 ns:pred ns:o2 }
>
> and I execute 'SELECT COUNT(?s) where { ?s ns:pred ?o }' with
> default-graph-uri set to ns:g1, I expect to get 1, but I get 2.
>
> I am assuming that there is someplace where the TupleExpr passed to
> evaluateInternal gets transformed into a scan pattern for Accumulo, after
> the right index (spo/ops/pso) is identified. If I remember correctly, S/P/O
> is used the the row ID, and the graph is the column name (or family?). So
> the dataset should modify the scan filter if applied correctly.
>
> Boris Pelakh
> Ontologist, Developer, Software Architect
> boris.pel...@semanticarts.com
> +1-321-243-3804
>
>
> -----Original Message-----
> From: Puja Valiyil <puja...@gmail.com>
> Sent: Tuesday, December 4, 2018 9:31 AM
> To: dev@rya.incubator.apache.org
> Subject: Re: Dataset not applied to queries
>
> Hi Boris,
> Unfortunately I'm not very familiar with that method -- can you provide
> some more details into exactly what its supposed to do?
> All of the connectors for implementing the sail api are in the rya.sail,
> and the main classes to look at are RdfCloudTripleStore, RyaSailRepository,
> and RdfCloudTripleStoreConnection. All of these classes all extend abstract
> implementations SAIL provides out of the box, so typically the way you
> extend them to do more stuff is to look at the Sail implementation and
> customize it for Rya's data structures.
> If you need any help or want to bounce any implementation ideas off of
> anyone, please send out an email over the dev list. Just let us know how we
> can help!
> Thanks,
> Puja
>
>
> On Tue, Dec 4, 2018 at 9:19 AM Boris Pelakh <boris.pel...@semanticarts.com
> >
> wrote:
>
> > I am in the process of implementing a SPARQL 1.1-compliant endpoint
> > for web.rya so that Rya can be used as a federated repository with
> > RDF4J workbench.
> >
> >
> >
> > One of requirements is to be able to pass default-graph-uri and
> > named-graph-uri filters external to the query to override the graph
> > filter in the query itself (see
> > https://www.w3.org/TR/sparql11-protocol/#query-operation).
> >
> >
> >
> > In the sail API this is accomplished by setting the dataset on the
> > query object (see
> > http://docs.rdf4j.org/javadoc/2.2/org/eclipse/rdf4j/query/Operation.ht
> > ml#setDataset-org.eclipse.rdf4j.query.Dataset-),
> > but that setting does not seem to be honored by the Rya SAIL
> > implementation. Any ideas on where I can start digging to correct this?
> >
> >
> >
> > (I have also tried wrapping the Rya repository connection in a
> > ContextAwareConnection and setting readContexts, but the effect is the
> > same).
> >
> >
> >
> > *Boris Pelakh*
> >
> > *Ontologist, Developer, Software Architect*
> >
> > *boris.pel...@semanticarts.com <boris.pel...@semanticarts.com>*
> >
> > *+1-321-243-3804*
> >
> > [image: SemanticArtsLogo]
> >
> >
> >
>
