Reasoning profile selector

A reasoning-profile selector on pgrdf.materialize lets consumers choose between RDFS and OWL 2 RL per call. Shipped in v0.5.0.

The materialize UDF takes an optional profile argument:

The surface

pgrdf.materialize(
    graph_id BIGINT,
    profile  TEXT DEFAULT 'owl-rl'
) → JSONB

Profile value	Rule set
'owl-rl' (default)	OWL 2 RL — the full forward-chaining rule set documented in OWL 2 RL rule set. Unchanged default; existing callers keep this behaviour without passing the argument.
'rdfs'	RDFS closures — rdfs:subClassOf, rdfs:subPropertyOf, rdfs:domain, rdfs:range. Faster and cheaper than full OWL 2 RL.

-- Default: OWL 2 RL (back-compatible — no argument needed).
SELECT pgrdf.materialize(100);

-- Bound the per-graph cost to the RDFS closures only.
SELECT pgrdf.materialize(100, 'rdfs');

Both profiles are idempotent in the same way the default call is — re-running drops the previously inferred rows and replaces them. See Idempotence + operator safety.

Why it matters

Different workloads benefit from different rule sets:

• High-throughput ingest pipelines running thousands of graphs an hour can choose 'rdfs' to bound the per-graph materialization cost.
• Default consumers keep 'owl-rl', the W3C-aligned middle ground, with no code change.

rocket_launchv0.6-FUTURE — 'owl-rl-ext'

An extended 'owl-rl-ext' profile (OWL 2 RL plus selected DL extensions) is on the post-v0.5 backlog, tracked in SPEC.pgRDF.LLD.v0.5-FUTURE.md §3. It is not callable today; 'owl-rl' and 'rdfs' are the two shipped v0.5.0 profiles.