On 09/12/15 10:38, Claus Stadler wrote:
Hi,
It appears that the FROM / FROM NAMED clauses of SPARQL queries are
ignored when executed over a Dataset.
For the general dataset, supplying the dataset directly to
QueryExecutionFactory overrides any query dataset description.
In the example below, I would expect the first result set to yield the
content of the file, whereas I expect the second one to be empty as the
specified named graph does not exist, yet, I get the exact opposite.
It is better to use GRAPH to access the named graph in a dataset.
Are there some magic switches or algebra transformations that can be
applied programmatically for changing the behavior?
Similar queries on Virtuoso on http://dbpedia.org/sparql work to my
expectation.
FWIW as does TDB.
https://jena.apache.org/documentation/tdb/dynamic_datasets.html
SPARQL says "there is a dataset description" - how it is interpreted is
not fixed. If you execute a query without dataset, it take the query
and use the FROM/FROM NAMED to load from the web.
The issue is what is the pool of graphs from which to resolve URIs.
A URL should go to the graph - it really does not name a graph inside
dataset if that graph is a copy of somewhere else.
Tested with jena-arq 2.13.0 and 3.0.0
A related issue might be this one, but it appears a fix was only
provided for Fuseki: https://issues.apache.org/jira/browse/JENA-1004
The commit shows the code that processes the query for this feature.
DynamicDatasets has the necessary code.
Dataset ds = ...
Query query = ...
// Get the description from the query
DatasetDescription desc = DatasetDescription.create(q) ;
// Create a new dataset using the original as the source of graphs.
Dataset ds1 = DynamicDatasets.dynamicDataset(desc, ds, false) ;
... QueryExecutionFactory.create(query, ds1);
but the proper way is to use GRAPH which is the SPARQL feature for
accessing a graph in a dataset.
Andy
Cheers,
Claus
Code:
public class TestDatasetQuery {
@Test
public void test() throws IOException {
Dataset ds = DatasetFactory.createMem();
RDFDataMgr.read(ds, new
ClassPathResource("test-person.nq").getInputStream(), Lang.NQUADS);
String graphName = ds.listNames().next();
Node s =
ds.getNamedModel(graphName).listSubjects().toSet().iterator().next().asNode();
System.out.println("Got subject: " + s + " in graph " +
graphName);
{
// Should yield some solutions - but actually doesn't
QueryExecution qe = QueryExecutionFactory.create("SELECT *
FROM <" + graphName + "> { ?s ?p ?o }", ds);
ResultSet rs = qe.execSelect();
System.out.println(ResultSetFormatter.asText(rs));
}
{
// Should not return anything, as the named graph does not
exist, yet, the original data is returned
QueryExecution qe = QueryExecutionFactory.create("SELECT *
FROM NAMED <http://foobar> { Graph ?g { ?s ?p ?o } }", ds);
ResultSet rs = qe.execSelect();
System.out.println(ResultSetFormatter.asText(rs));
}
}
}
File: test-person.nq
<http://ex.org/JohnDoe> <http://ex.org/label> "John
Doe"^^<http://www.w3.org/2001/XMLSchema#string> <http://ex.org/graph/> .
<http://ex.org/JohnDoe> <http://ex.org/age>
"20"^^<http://www.w3.org/2001/XMLSchema#int> <http://ex.org/graph/> .
Output:
Got subject: http://ex.org/JohnDoe in graph http://ex.org/graph/
-------------
| s | p | o |
=============
-------------
---------------------------------------------------------------------------------------------------------------------------
| s | p |
o | g |
===========================================================================================================================
| <http://ex.org/JohnDoe> | <http://ex.org/age> |
"20"^^<http://www.w3.org/2001/XMLSchema#int> | <http://ex.org/graph/> |
| <http://ex.org/JohnDoe> | <http://ex.org/label> | "John
Doe" | <http://ex.org/graph/> |
---------------------------------------------------------------------------------------------------------------------------
