The four pillars at a glance

pgRDF is one Postgres extension exposing four pillars under the pgrdf.* schema. They compose — the same graph_id you load Turtle into is the one you SPARQL-query, the one you materialize, and the one you validate.

Pillar	What it gives you	Entry-point UDFs
1 · Semantic storage	RDF triples land in dictionary-encoded, partitioned Postgres tables you can SELECT from with vanilla SQL. Turtle in, quads out. v0.6 adds the native staged bulk loader.	pgrdf.load_turtle, pgrdf.load_turtle_staged_run, pgrdf.parse_turtle, pgrdf.add_graph, pgrdf.count_quads
2 · Semantic query	Full SPARQL 1.1 — SELECT / ASK / CONSTRUCT / DESCRIBE / UPDATE over those triples: multi-pattern joins, FILTER, OPTIONAL, UNION, MINUS, aggregates (incl. over UNION), BIND, VALUES, type-aware ORDER BY, GRAPH, property paths. Returns JSONB rows you can join with regular SQL.	pgrdf.sparql, pgrdf.construct, pgrdf.describe, pgrdf.sparql_parse
3 · Semantic materialization	OWL 2 RL and RDFS forward-chaining inference (per-call profile selector). Implicit consequences (subclass, subproperty, equivalence, inverse, transitive) are written back into the same tables as queryable rows.	pgrdf.materialize
4 · Semantic validation	Native SHACL Core constraint checking — W3C SHACL Core 25/25. A graph + a shapes graph produce a W3C-shape sh:ValidationReport JSONB you can persist, alert on, or gate ingestion with.	pgrdf.validate

The operating model

The four pillars split cleanly along one axis: how they scale.

• Ingest is parallel and scales with the box. Storage — and especially the staged bulk loader — fans across a background-worker pool. Add cores and the load goes faster: the staged loader takes the full 8.2-billion-triple Wikidata truthy graph end to end (see Scale).
• Reasoning is single-threaded and sized to fit.Materialization and validation run on one backend. The pattern, then, is: ingest the full graph in parallel, carve out a right-sized slice, and reason over that slice on the box you have. See Processes & flows for the composable verb chains, and Roadmap for the carving surface landing across the v0.6.n line.

Anatomy of a single workflow

A typical pgRDF session uses all four pillars in sequence:

-- 1. Storage — load the data + shapes graphs.
SELECT pgrdf.load_turtle('/data/orders-2026-Q1.ttl', 100);
SELECT pgrdf.load_turtle('/shapes/order-shape.ttl',  200);

-- 2. Inference — materialize OWL 2 RL closures on the data graph.
SELECT pgrdf.materialize(100);

-- 3. Validation — check the data graph (now including inferences)
--    against the shapes graph.
SELECT pgrdf.validate(100, 200);

-- 4. Query — pull whatever subset of the graph your app needs.
SELECT * FROM pgrdf.sparql(
  'PREFIX ex: <http://example.com/>
   SELECT ?order ?customer ?total
     WHERE { ?order a ex:Order ;
                    ex:placedBy ?customer ;
                    ex:total    ?total
             FILTER(?total > 1000) }');

Continue — Pillar 1 · Semantic storage →