Github user afs commented on a diff in the pull request:
https://github.com/apache/jena/pull/332#discussion_r158299025
--- Diff:
jena-arq/src/main/java/org/apache/jena/sparql/graph/GraphUnionRead.java ---
@@ -82,16 +94,29 @@ protected PrefixMapping createPrefixMapping() {
if ( graphs == null ) {
// This produces unique quads with the same graph node,
// hence the triples are distinct.
- Iterator<Quad> qIter = dataset.findNG(Quad.unionGraph,
m.getSubject(), m.getPredicate(), m.getObject());
- Iterator<Triple> tIter = Iter.map(qIter,
quad->quad.asTriple());
- return WrappedIterator.createNoRemove(tIter) ;
+ return quadsToTriples(dataset, Quad.unionGraph, m);
+ }
+ if ( graphs.isEmpty() )
+ return NullIterator.instance();
+ if ( graphName != null ) {
+ if ( ! dataset.containsGraph(graphName) )
+ // Avoid auto-creation.
+ return NullIterator.instance();
+ // Avoid needing distinct.
+ return dataset.getGraph(graphName).find(m);
}
// Only certain graphs.
IteratorConcat<Triple> iter = new IteratorConcat<>() ;
forEachGraph((g) -> iter.add(g.find(m))) ;
return WrappedIterator.createNoRemove(Iter.distinct(iter)) ;
}
+ private static ExtendedIterator<Triple> quadsToTriples(DatasetGraph
dsg, Node graphName, Triple m) {
--- End diff --
It avoids churning graph objects. Graphs will eventually have to call down
to core dataset-centric operations. Also this avoids an infinite recursion risk.
(This code fragment is out of sync.)
---
