GRAPH — query named graphs

Every triple in pgRDF lives in a graph (see Managing graphs). A pattern without GRAPH matches across all of them, including the default graph. GRAPH narrows it down.

Form	Matches
GRAPH <iri> { … }	Only the graph with that IRI. An IRI that names no graph matches nothing.
GRAPH ?g { … }	Each named graph in turn, with ?g bound to its IRI. The default graph is not included.
No GRAPH	All graphs, including the default graph.

The default graph's IRI is urn:pgrdf:graph:0, so GRAPH <urn:pgrdf:graph:0> { … } reads only the default graph.

FROM and FROM NAMED are ignored

FROM <iri> and FROM NAMED <iri> are accepted but currently have no effect: the query still runs over all graphs. Use GRAPH to choose graphs.

Example data

Two small graphs, next to the people graph from the sample data:

SELECT pgrdf.add_graph('http://example.org/g1');
SELECT pgrdf.add_graph('http://example.org/g2');
SELECT pgrdf.parse_turtle('@prefix ex: <http://example.org/> . ex:a ex:p "in g1" .',
                          pgrdf.graph_id('http://example.org/g1'));
SELECT pgrdf.parse_turtle('@prefix ex: <http://example.org/> . ex:a ex:p "in g2" . ex:b ex:p "only g2" .',
                          pgrdf.graph_id('http://example.org/g2'));

Remove them afterwards with SELECT pgrdf.drop_graph('http://example.org/g1'); and the same for g2.

One graph

SELECT * FROM pgrdf.sparql($$
  SELECT ?s ?o
  WHERE { GRAPH <http://example.org/g1> { ?s ?p ?o } }
$$);
--  {"o": "in g1", "s": "http://example.org/a"}

Which graphs match

GRAPH ?g tells you where each answer came from:

SELECT * FROM pgrdf.sparql($$
  PREFIX ex: <http://example.org/>
  SELECT ?g ?o WHERE { GRAPH ?g { ex:a ex:p ?o } }
  ORDER BY ?g
$$);
--  {"g": "http://example.org/g1", "o": "in g1"}
--  {"g": "http://example.org/g2", "o": "in g2"}

Combined with an aggregate, it counts triples per graph:

SELECT * FROM pgrdf.sparql($$
  SELECT ?g (COUNT(*) AS ?n)
  WHERE { GRAPH ?g { ?s ?p ?o } }
  GROUP BY ?g ORDER BY ?g
$$);
--  {"g": "http://example.org/g1", "n": "1"}
--  {"g": "http://example.org/g2", "n": "2"}
--  {"g": "http://example.org/people", "n": "14"}

Combining with OPTIONAL, UNION and MINUS

OPTIONAL inside GRAPH

SELECT * FROM pgrdf.sparql($$
  PREFIX foaf: <http://xmlns.com/foaf/0.1/>
  SELECT ?g ?name ?mbox
  WHERE { GRAPH ?g { ?s foaf:name ?name
                     OPTIONAL { ?s foaf:mbox ?mbox } } }
  ORDER BY ?name
$$);
--  {"g": "http://example.org/people", "mbox": "mailto:alice@example.org", "name": "Alice"}
--  {"g": "http://example.org/people", "mbox": null, "name": "Bob"}
--  {"g": "http://example.org/people", "mbox": "mailto:carol@example.org", "name": "Carol"}

UNION across two graphs

SELECT * FROM pgrdf.sparql($$
  PREFIX ex: <http://example.org/>
  SELECT ?s ?o
  WHERE { { GRAPH <http://example.org/g1> { ?s ex:p ?o } }
          UNION
          { GRAPH <http://example.org/g2> { ?s ex:p ?o } } }
  ORDER BY ?o
$$);
--  {"o": "in g1", "s": "http://example.org/a"}
--  {"o": "in g2", "s": "http://example.org/a"}
--  {"o": "only g2", "s": "http://example.org/b"}

MINUS to subtract one graph from another

Subjects in g2 that g1 doesn't mention, the way you'd compare a graph with an earlier snapshot:

SELECT * FROM pgrdf.sparql($$
  PREFIX ex: <http://example.org/>
  SELECT ?s
  WHERE { GRAPH <http://example.org/g2> { ?s ex:p ?o }
          MINUS { GRAPH <http://example.org/g1> { ?s ex:p ?o2 } } }
$$);
--  {"s": "http://example.org/b"}

Choosing graphs by list

VALUES ?g { … } and FILTER(?g = …) on the variable of GRAPH ?g are refused with an error. Either name each graph in a UNION (as in UNION across two graphs), or query all graphs with GRAPH ?g and keep the ones you want in SQL:

SELECT sparql
  FROM pgrdf.sparql($$
    PREFIX ex: <http://example.org/>
    SELECT ?g ?o WHERE { GRAPH ?g { ex:a ex:p ?o } }
  $$)
 WHERE sparql->>'g' IN ('http://example.org/g1', 'http://example.org/g2')
 ORDER BY 1;
--  {"g": "http://example.org/g1", "o": "in g1"}
--  {"g": "http://example.org/g2", "o": "in g2"}

In updates

GRAPH <iri> { … } works in the templates and patterns of SPARQL UPDATE too, and WITH <iri> scopes a whole update to one graph. Check which graph an update writes to before running one.

See also

• Managing graphs: create, inspect, copy, lock and drop graphs.
• Named graphs: how graph IRIs map to ids.