pgrdf

pgrdf : RDF, SPARQL, SHACL, and OWL reasoning for PostgreSQL

Overview

Packages

Source

pig build pkg pgrdf;		# build rpm/deb

Install

Make sure PGDG and PIGSTY repo available:

pig repo add pgsql -u   # add both repo and update cache

Install this extension with pig:

pig install pgrdf;		# install via package name, for the active PG version

pig install pgrdf -v 18;   # install for PG 18
pig install pgrdf -v 17;   # install for PG 17
pig install pgrdf -v 16;   # install for PG 16
pig install pgrdf -v 15;   # install for PG 15
pig install pgrdf -v 14;   # install for PG 14

Config this extension to shared_preload_libraries:

shared_preload_libraries = 'pgrdf';

Create this extension with:

CREATE EXTENSION pgrdf;

Usage

Sources:

• pgRDF 0.6.20 README
• pgRDF 0.6.20 user guide
• pgRDF 0.6.20 changelog
• pgRDF 0.6.20 release

pgRDF stores RDF data inside PostgreSQL and exposes SQL-callable helpers for Turtle/TriG/N-Quads loading, SPARQL query/update, named graphs, SHACL validation, and RDFS/OWL 2 RL materialization.

CREATE EXTENSION pgrdf;
SELECT pgrdf.version();

Preload And PostgreSQL Version Caveat

pgRDF 0.6.20 supports PostgreSQL 14-18 and moves from pgrx 0.16.1 to 0.19.1. Upstream describes 0.6.20 as a build/runtime migration with no schema or query-surface change; PostgreSQL 19 remains a tracked follow-up.

pgrdf must be present in shared_preload_libraries before PostgreSQL starts. Without preload, upstream documents that the shared-memory dictionary and plan-cache atomics are not initialized and the first pgRDF call can fail.

shared_preload_libraries = 'pgrdf'

Restart PostgreSQL after changing this setting, then verify:

SHOW shared_preload_libraries;

SELECT pgrdf.parse_turtle(
  '@prefix ex: <http://example.org/> . ex:t a ex:T .',
  1::bigint,
  'http://example.org/'
);

Load RDF

Use parse_turtle for inline Turtle payloads and load_turtle for server-side files. Graph ids are bigint values; named graph helpers map ids to IRIs. Version 0.6.x adds a parallel bulk loader path via load_turtle(..., bulk_load => true).

SELECT pgrdf.add_graph(100::bigint, 'http://example.org/graph/main');

SELECT pgrdf.parse_turtle(
  '@prefix ex: <http://example.org/> .
   ex:alice ex:knows ex:bob .
   ex:alice ex:name "Alice" .',
  100::bigint,
  'http://example.org/graph/main'
);

SELECT pgrdf.load_turtle('/srv/rdf/foaf.ttl', 100::bigint);
SELECT pgrdf.load_turtle('/srv/rdf/bulk.ttl', 100::bigint, bulk_load => true);
SELECT pgrdf.count_quads(100::bigint);

Related ingest and graph-management functions documented upstream include parse_trig, parse_nquads, add_graph, drop, clear, copy, move_graph, graph_id, and graph_iri.

Carve Graph Slices

The 0.6.x series adds carve_graph overloads for copying a predicate-defined slice or a bounded neighbourhood into another graph without decoding and re-encoding the shared dictionary:

SELECT pgrdf.carve_graph(
  100::bigint,
  'http://example.org/type'::text,
  200::bigint
);

SELECT pgrdf.carve_graph(
  100::bigint,
  ARRAY['http://example.org/alice', 'http://example.org/bob']::text[],
  201::bigint,
  2
);

The neighbourhood form uses max_hops as a graph-distance boundary. Set pgrdf.on_path_truncation to warn or error when a truncated property-path walk must not be silently accepted.

Query With SPARQL

pgrdf.sparql(text) returns SPARQL results as SQL rows. The upstream v0.5 surface includes SELECT and ASK, filters, ordering, limits, OPTIONAL, UNION, MINUS, aggregates, VALUES, BIND, CONSTRUCT, DESCRIBE, named-graph GRAPH clauses, and property paths.

SELECT *
FROM pgrdf.sparql(
  'PREFIX ex: <http://example.org/>
   SELECT ?person ?name
   WHERE {
     ?person ex:name ?name .
     FILTER(REGEX(?name, "^A", "i"))
   }
   ORDER BY ?name
   LIMIT 20'
);

Named-graph queries can bind graph IRIs:

SELECT *
FROM pgrdf.sparql(
  'PREFIX ex: <http://example.org/>
   SELECT ?g (COUNT(*) AS ?n)
   WHERE { GRAPH ?g { ?s ex:name ?name } }
   GROUP BY ?g
   ORDER BY ?g'
);

Update Graphs

The upstream v0.5 surface includes SPARQL Update forms such as INSERT DATA, DELETE DATA, INSERT/DELETE WHERE, DELETE+INSERT WHERE, and graph lifecycle statements.

SELECT pgrdf.sparql(
  'PREFIX ex: <http://example.org/>
   INSERT DATA {
     GRAPH <http://example.org/graph/main> {
       ex:bob ex:name "Bob" .
     }
   }'
);

Reasoning And Validation

Use pgrdf.materialize(graph_id, profile) to write inferred triples for rdfs or owl-rl profiles. Materialization is intended to be repeatable; upstream documents that previous inferred rows are dropped before writing the new closure.

SELECT pgrdf.parse_turtle(
  '@prefix ex:   <http://example.com/> .
   @prefix rdf:  <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
   @prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
   ex:Engineer rdfs:subClassOf ex:Person .
   ex:Person   rdfs:subClassOf ex:Agent .
   ex:alice    rdf:type        ex:Engineer .',
  100::bigint
);

SELECT pgrdf.materialize(100::bigint, 'owl-rl');

SELECT *
FROM pgrdf.sparql(
  'PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
   PREFIX ex:  <http://example.com/>
   SELECT ?class WHERE { ex:alice rdf:type ?class }'
);

Use pgrdf.validate(data, shapes, mode) for SHACL validation; upstream documents JSONB sh:ValidationReport output and native SHACL Core support. SHACL-SPARQL constraint execution is documented upstream as gated by its RDF library dependency, so treat mode => 'sparql' as an advanced surface to verify against the exact installed version.

Operational Helpers

Useful introspection and cache-management helpers documented upstream include:

The pgrdf.path_max_depth setting guards property-path expansion depth, while pgrdf.on_path_truncation = count | warn | error controls how callers learn that the guard was reached.

Version Notes

The releases between 0.6.4 and 0.6.20 materially improve large RDF ingestion and query correctness: streaming/windowed bulk ingestion, a staged multi-backend loader, safe repeated loads into populated dictionaries, graph-carving helpers, dictionary inclusion in pg_dump, SPARQL expression/aggregate additions, and fail-closed path-truncation handling. The 0.6.20 release itself changes only the build/runtime layer to pgrx 0.19.1 and adds PostgreSQL 18 support.

For very large fresh N-Triples loads, upstream documents pgrdf.load_turtle_staged_run as the resumable, phase-oriented path. It commits parsing, dictionary, resolution, and index phases separately and is operationally different from the transactional load_turtle() call; validate staging tablespaces, disk headroom, and recovery procedures before using it for production-scale imports.