Ga naar hoofdinhoud

Publieke fulltext search — publicaties + documenten

Public, RBAC-filtered full-text search endpoint that returns publications and documents as mixed-typed rows in a flat envelope, discriminated by @self.schema. Anonymous-reachable — no login required.

Standards references​

  • WOO (Wet open overheid) — public discoverability of published documents.
  • GEMMA Catalogus / Data-catalogus — cross-organization publication discovery layer.
  • ADR-022 (Consume OpenRegister abstractions) — search fans out through OR's zoeken-filteren, does not re-implement query building.

Overview​

GET /apps/opencatalogi/api/search absorbs the previously admin-only search endpoint into a public surface. Rather than exposing a per-object-type endpoint (publications-only + documents-only), it returns a single flat result array containing both types. Frontends switch rendering on the @self.schema field per row.

Every document row includes an embedded publication: { id, slug, title } summary so a "document card" can link back to its parent publication without a second API roundtrip. Documents with no linked publication are suppressed from the anonymous result set.

Visibility is enforced server-side via isObjectPublic() post-scoring: publicatiedatum <= now (and depublicatiedatum in the future) — no drafts or depublished content leak into the response. Documents inherit their linked publication's visibility transitively.

Key capabilities​

  • Public endpoint — #[PublicPage] + #[NoCSRFRequired]; anonymous callers reach it, no session required (SCH-PFTS-001).
  • Mixed envelope with @self.schema discriminator — publications + documents share one flat results[] array; each row carries @self.schema set to its schema slug (SCH-PFTS-002).
  • Embedded publication summary on document rows — publication: { id, slug, title } on every document row for direct-linking (SCH-PFTS-003).
  • Post-scoring visibility filter — isObjectPublic() applied after search + merge; scoring decisions run against the full corpus before visibility is enforced (SCH-PFTS-004).
  • Transitive publication visibility — documents whose linked publication fails isObjectPublic() are dropped (SCH-PFTS-004).
  • Companion /publications endpoint unchanged — the existing publication-scoped endpoint keeps its exact contract (no field / admission / shape drift).
  • Config route enforces scope — the endpoint returns whatever schemas the catalog's registers + schemas array is configured for. Freshly-seeded catalogs contain only publication; admins can broaden by adding document (or other schemas) to the catalog config.
  • Opt-in document body-text search (WOO-517) — _content=true widens document matching to the extracted body text (PDF/DOCX/XLSX/EML/text) of attached files, not just metadata. Defaults to false — byte-identical to the base behaviour above when omitted. Content-matched document rows are indistinguishable in shape from metadata-matched rows, are deduplicated on @self.id against any metadata match on the same document, and pass through the same isObjectPublic() + transitive publication-visibility gate (SCH-PFTS-CONTENT-001/-002/-003). See ?_search=<term>&_content=true.

OpenSpec provenance​

Shipped by change add-public-fulltext-search (archived 2026-07-16). Requirements landed in canonical spec openspec/specs/search/spec.md as SCH-PFTS-001..007 (plus MODIFIED SCH-OR-003 for the anonymous-reachability + multi-schema delegation clauses).

The _content opt-in above is shipped by add-document-content-search (WOO-517), adding SCH-PFTS-CONTENT-001/-002/-003 and a MODIFIED SCH-PFTS-002 to the same canonical spec. Depends on OpenRegister's _content_search flag on ObjectService::searchObjectsPaginated(), shipped in openregister:expose-content-search-in-object-service (merged Conduction/openregister:development commit e9d4d787).

Jira parent: WOO-506. Follow-up: WOO-517 for document body-text content search.