insert from two named graphs?

Question

I'm looking for an easy way to insert triples from two or more named graphs (but not the entire unnamed default graph) into another named graph. I'm using GraphDB.

I guess this could be done by writing out the same query multiple times in the WHERE section, wrapped in multiple GRAPH specifications, and then unioning them together, but my WHEREs are long and I'd prefer not to write them out multiple times.

Let's say I have loaded some data like this:

INSERT DATA {
  GRAPH <http://example.com/ngA> {
    <http://example.com/person1> <http://example.com/name> "Arthur" .
  }
  GRAPH <http://example.com/ngB> {
    <http://example.com/person1> <http://example.com/name> "Brian" .
  }
  GRAPH <http://example.com/ngC> {
    <http://example.com/person1> <http://example.com/name> "Charlie" .
  }
}

I can copy all of the triples of a certain pattern from the default unnamed graph into a new named graph with something like this:

INSERT {
  GRAPH <http://example.com/ngZ> {
    ?s <http://example.com/moniker> ?o .
  }
}
WHERE
  { ?s  <http://example.com/name>  ?o }

An easy way to SELECT for triples of a given pattern from two or more (but not all) named graphs is

SELECT  *
FROM <http://example.com/ngA>
FROM <http://example.com/ngB>
WHERE
  { ?s  <http://example.com/name>  ?o }

What if I want to copy those triples, from those specified graphs, into another graph?

I'm getting an error from GraphDB 8.3 (and from the sparql.org validator) when I try to

INSERT {
  GRAPH <http://example.com/ngZ> {
    ?s <http://example.com/moniker> ?o .
  }
}
WHERE
  { SELECT  *
FROM <http://example.com/ngA>
FROM <http://example.com/ngB>
WHERE
  { ?s  <http://example.com/name>  ?o } }

Smithers

Answer 1

Try this query:

PREFIX ex: <http://example.com/>

INSERT {
  GRAPH ex:ngZ { ?s ex:moniker ?o }
}
WHERE {
  GRAPH ?g { ?s ex:name ?o }  
  FILTER (?g IN ( ex:ngA, ex:ngB ) )
}

And then:

PREFIX ex: <http://example.com/>

SELECT * 
FROM NAMED ex:ngZ
WHERE { 
    GRAPH ?g { ?s ?p ?o }
} LIMIT 100 

Is it what you need?

By the way, there exist COPY (use with caution!) and ADD.

Unmistakable

Answer 2

SPARQL Update provides USING and USING NAMED analogous to FROM and FROM NAMED in queries:

The USING and USING NAMED clauses affect the RDF Dataset used while evaluating the WHERE clause. This describes a dataset in the same way as FROM and FROM NAMED clauses

You can express the requirement as an UPDATE like so:

INSERT {
  GRAPH <http://example.com/ngZ> {
    ?s <http://example.com/moniker> ?o .
  }
}
USING <http://example.com/ngA>
USING <http://example.com/ngB>
WHERE
  {  ?s  <http://example.com/name>  ?o }

Also note that, according to the SPARQL query grammar, a subquery does not admit a dataset clause. This is why the SPARQL parsers are rejecting your query.

Gillies

Answer 3

Thanks, @Stanislav Kralin

Come to think of it, this also works:

PREFIX ex: <http://example.com/>
INSERT {
    GRAPH ex:ngZ {
        ?s ex:moniker ?o 
    }
}
WHERE {
    values ?g {
        ex:ngA ex:ngB 
    }
    GRAPH ?g {
        ?s ex:name ?o 
    }  
}

Smithers