buildSeal

Make an imported graph query-ready by building its hexastore indexes. Seal is the second parallel verb — it finalizes the SPO / POS / OSP index permutations the query planner relies on.

What it is

Raw imported quads are queryable, but fast lookup in any direction — subject, predicate, or object first — needs the covering index permutations. Seal is the step that builds them. After Seal, a multi-pattern SPARQL join resolves against an index rather than a sequential scan.

Seal is a step, not a UDF

There is no pgrdf.seal(). Seal names the index-build step that the loaders perform. In the native staged loader it is the INDEX phase — the SPO / POS / OSP builds run concurrently, one background worker per CREATE INDEX. When you build a graph by other means, Seal is the equivalent CREATE INDEX step.

Where it sits in a chain

After Import, before everything else. A graph must be sealed before Query, Reason, or Validate run efficiently. In the scale-meets-hardware chain you seal twice: once on the full imported graph, then again on the carved slice.

Scaling class — parallel

Seal scales with the box. The staged loader's INDEX phase builds the hexastore permutations concurrently — on the 8.2 B flagship the index build ran in 31.9 min across the worker pool. See Scale & benchmarks.

See also

• Hexastore + dictionary — the SPO / POS / OSP index layout Seal builds.
• Native staged bulk loader — where Seal is the INDEX phase.
• Query — what a sealed graph unlocks.