3 · R2RML & Ontop — SPARQL becomes SQL

R2RML in one sentence

R2RML (W3C Recommendation) maps relational tables to RDF triples declaratively: each TriplesMap describes how rows in a logical table become subject IRIs, classes and predicate-object pairs.

Ontop reads the R2RML at startup, combines it with the ontology's TBox, and from that point on every incoming SPARQL query is rewritten — through unfolding and optimisation — into a single SQL query.

Why declarative matters

There is no code path between SPARQL and SQL in this design. No serialisers, no row-by-row transforms, no consumer to monitor. Wrong output ⇒ wrong mapping or wrong ontology, full stop. The blast radius of a bug is small enough to actually own.

The rewriting in action

A regulator asks (in SPARQL):

PREFIX : <https://atlas.example.com/onto/reporting/>
SELECT ?name WHERE {
  ?p a :HighRiskCounterparty ;
     :legalName ?name .
}

Ontop unfolds this against the R2RML and emits roughly:

SELECT legal_name AS name
FROM   dwh.party
WHERE  is_active = TRUE
  AND  risk_tier = 'HIGH'

One SQL query. The warehouse planner does what it always does. The regulator gets fresh numbers, computed at query time.