Skip to content

chore(release): sync main back into development - #1348

Open
github-actions[bot] wants to merge 18 commits into
developmentfrom
sync/main-to-development-2.0.11
Open

chore(release): sync main back into development#1348
github-actions[bot] wants to merge 18 commits into
developmentfrom
sync/main-to-development-2.0.11

Conversation

@github-actions

@github-actions github-actions Bot commented Sep 1, 2026

Copy link
Copy Markdown

Opened by the release workflow after publishing 2.0.11 from main.

A release commits a version bump onto the branch it ran from. Without carrying
that back, development falls behind main, and since both branches then
changed the version file independently, the next development -> main
promotion conflicts on it — every time.

Version files are resolved to development's side, the higher line, so this
cannot move a version backwards. Any other conflict stops the workflow instead
of being resolved automatically.

Merge this, do not squash. A squash copies the files and discards the
relationship: the merge base would not move and the next promotion would
conflict exactly as before. A 0-file diff here is normal and is not a no-op —
recording the ancestry is the whole point.

WilcoLouwerse and others added 18 commits August 25, 2026 15:45
@WilcoLouwerse
feat(rbac): backfill publication+document read-rule two-rule shape (W…
ebc2a24 
…OO-536 Fase 4)

Robert Zondervan's WOO-536 architecture fix requires the public search
endpoint to filter depublished objects. The current seed shape has only
a single conditional rule (`publicatiedatum $lte $now`) — depublished
objects (past `depublicatiedatum`) remain visible under RBAC.

## Changes

- `lib/Settings/publication_register.json` — `publication` + `document`
  schemas updated to the two-rule read shape:
    read: [
      { group: public, match: { publicatiedatum $lte $now, depublicatiedatum $gte $now } },
      { group: public, match: { publicatiedatum $lte $now, depublicatiedatum $exists false } },
      "authenticated"
    ]
- `lib/Repair/WOO536RepairReadRules.php` — new IRepairStep for existing
  installations. Runs on `occ maintenance:repair`, detects the single-rule
  pre-fix shape, upgrades to two-rule, preserves "authenticated" element,
  admin-customised shapes are left alone. Idempotent.
- `appinfo/info.xml` — repair step registered.
- `openspec/config.yaml` — corrected `schema: conduction` → `spec-driven`
  (the `conduction` schema-file is not bundled in this repo, blocking
  all openspec commands). Fleet-wide alignment to `conduction` is a
  separate follow-up (5 apps still on `conduction`, mixed state).
- `openspec/changes/fix-fts-catalog-model-alignment/` — full OpenSpec
  change scaffold (proposal + design + specs + tasks). `openspec validate`
  green. depends_on: [rbac-as-public-toggle] (the OR precursor).

## Tests

`tests/Unit/Repair/WOO536RepairReadRulesTest.php` — 6 tests / 16
assertions covering: getName, skip-when-OR-absent, upgrade single-rule,
idempotency on two-rule, preserve admin-customised, skip-when-schema-absent.
All green.

Fase 5 (code refactor consuming `_rbac_as_public`) waits for OR PR #2855
merge before wiring the primitive.

@SPEC openspec/changes/fix-fts-catalog-model-alignment/specs/search/spec.md
@WilcoLouwerse
feat(search): refactor assemblePublicSearchResults per WOO-536 (Fase 5)
02d2f0d 
Rework the public /api/search assembler per Robert Zondervan's WOO-536
architectural fix. Ships Robert's Stap 1-5a + 7 on top of the OR precursor
primitive (openregister PR #2855, RBA-PUBLIC-001..006).

## What changed

Stap 1 — enable SQL RBAC with public-endpoint discipline:
  `searchObjectsPaginated(..., _rbac: true, _rbac_as_public: true)`.
  Visibility now enforced in SQL by OR's schema authorization, not by a
  PHP post-filter. Admin and anonymous callers see the SAME result set
  (SCH-PFTS-001).

Stap 2 — catalog-derived scope (SCH-PFTS-CAT-001..003). New helper
  `resolveCatalogScope()` accepts `_catalog=<slug>` or `_catalogi[]=…`
  (Q4 naming convention: underscore-prefix, Dutch plural).
  Default = union of `listed:true` + published catalogs. The three
  `resolveConfiguredId()` calls (publication_register / publication_schema
  / document_schema) are gone — `resolveConfiguredId()` itself removed
  as dead code. Q7 Interpretation A preserved: `_schema` / `_registers`
  / `fq` / `catalogSlug` still stripped, plus the new `_catalog` /
  `_catalogi` are consumed by resolveCatalogScope before forward-to-OR.

Stap 3 — dynamic schema-discriminator via SchemaMapper. The hard-coded
  `$schemaSlugById` two-element map is gone; slugs resolved per-row via
  `$this->container->get('OCA\\OpenRegister\\Db\\SchemaMapper')->find()`,
  cached per-request. `@self.schema` carries the real slug for every
  schema in scope.

Stap 4 — remove PHP post-filter + restore totals/facets. `isObjectPublic()`
  post-filter loop deleted; anonymous total-under-count workaround gone;
  facets/facetable stripping for anon callers gone. OR's `total` +
  `facets` reflect visible-only counts by construction under
  `_rbac_as_public: true`.

Stap 5a — relations-based document→publication lookup.
  `resolveDocumentPublicationSummary()` rewritten to query the publication
  schema with `_relations_contains: <document_uuid>` (per-document
  refinement stays on the single-schema path). N4a (0 relations) drops
  the row; N4b (multiple relations) picks the OLDEST-by-`@self.created`
  as the most stable link.

Stap 7 — fixed broken docblock refs. `SCH-PFTS-002/003/004/005/006/007`
  refs replaced by inline descriptive text; `add-public-fulltext-search/*`
  and `add-document-content-search/*` refs replaced by
  `fix-fts-catalog-model-alignment/*` + canonical spec at
  `openspec/specs/search/spec.md`.

## Deferred (rationale in tasks.md)

- 3.5 (guard scope-schemas with missing read config) — edge case (admin
  responsibility to configure RBAC on added schemas); nice-to-have,
  not required for hotfix scope
- 6.2 (generalise document→publication to related→source) — future-work
  abstraction, current concrete impl covers WOO-536 DoD
- 8.1 (PublicationQueryServiceTest) — comprehensive unit tests
  deferred; validation via docker smoke tests in Fase 7

## Robert's DoD (unchanged)

`GET /apps/opencatalogi/api/search?_search=<term>` returns matches from
every schema in every catalog the caller may see, with documents linked
via relations, correct totals and facets, and no object visible that
its schema's `read` authorization would deny.

@SPEC openspec/changes/fix-fts-catalog-model-alignment/specs/search/spec.md
@WilcoLouwerse @claude
fix(search): address review #5041094649 findings (WOO-536)
bb21d16 
Applies 1 blocker + 5 concerns + 4 minor from the self-review on PR #1154.

B1 — RET-006 archived-visibility regression: re-introduce a targeted
`status: archived` guard in the row loop AND on the linked-publication
fast-path. SQL-first `$ne` on the schema read-rules would have silently
dropped legacy rows where `status` is NULL (SQL NULL semantics); PHP
guard is deterministic and safe for legacy data. Belt-and-braces above
the SQL RBAC.

C1 — Delete dead `isObjectPublic()` and `isAnonymous()` methods (zero
callers repo-wide, misleading docblocks). Drop the now-unused
`?IUserSession` constructor param.

C2 — Explicit `_catalog=<slug>` and `_catalogi[]` re-apply the
`published<=now` predicate via a new `isCatalogPubliclyAvailable()`
helper, so unpublished catalogs are indistinguishable from unknown
slugs on the public endpoint. `listListedPublishedCatalogs()` refactored
to use the same helper.

C3 — Drop the broken `@spec openspec/specs/search/spec.md#SCH-PFTS-001`
anchor; the SCH-PFTS-* IDs only exist in the change-delta until
archived. Keep the change-delta `@spec` line.

C4 — Log a warning when `catalog_register` / `catalog_schema` app-config
is empty (previously silent HTTP 200 empty).

C5 — Not viable: `ObjectService::searchObjects()` does not accept
`_rbacAsPublic` (only `searchObjectsPaginated` and `find()` do). Left
comment explaining why the `_rbac:false` + PHP filter stays on the
catalog-enumeration path.

M1 — Fast-path in `resolveDocumentPublicationSummary()`: if the document
row carries a resolvable `publication.id`, verify via a single `find()`
(with `_rbac:true, _rbacAsPublic:true`) before falling back to the
`_relations_contains` refinement. Eliminates the N+1 for well-formed
documents.

M2 — Add `SPDX-License-Identifier` + `SPDX-FileCopyrightText` to
`WOO536RepairReadRules` header (matches repo convention).

M3 — `resolvePublicationSchemaId()` now shares the outer
`$schemaSlugById` cache by reference, so the pre-loop schema scan
warms the per-row cache instead of duplicating mapper hits.

M4 — Multi-linked-document log downgraded from `info` to `debug` to
keep log volume sane on hot public deployments.

Also: replace deprecated `IAppManager::getInstalledApps()` with
`isEnabledForAnyone('openregister')` in the repair step, update its
test mocks accordingly.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@WilcoLouwerse @claude
chore(openspec): sync + archive fix-fts-catalog-model-alignment (WOO-…
9d4449c 
…536)

Sync SCH-PFTS-001 + SCH-PFTS-CAT-001..003 into specs/search/spec.md
(canonical requirement table + Requirement/Scenario blocks) and move
the change dir to `openspec/changes/archive/2026-08-28-*`.

Mirrors the discipline used on the OR precursor PR #2855:
sync + archive commits land in the same PR as the code, so the
canonical spec state after merge matches the implementation.

- `openspec validate search` -> green
- 3 tasks intentionally deferred (3.5 future scope-widening guard,
  6.2 helper generalisation, 8.1 unit test - tracked separately)

WOO-536.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
@WilcoLouwerse @claude
Merge branch 'main' into hotfix/woo-536-fts-catalog-model
9b41d66 
Bring the WOO-536 hotfix branch up to date with 1054 commits of
main-line development since fork. Three conflicts resolved:

1. `appinfo/info.xml` — took main's rewritten manifest (v2.0.6,
   EUPL-1.2, i18n, new repair steps, GitHub URL refresh) and
   added `WOO536RepairReadRules` after `RenameDutchPublicationColumns`
   so the two-rule read-block backfill sees English field names.

2. `lib/Service/PublicationQueryService.php` — kept the WOO-536
   Fase-5 refactor (`resolveCatalogScope`, `_rbac_as_public`,
   dynamic schema discriminator, `_relations_contains`-based
   document→publication lookup). Robert's WOO-536 diagnosis
   flagged main's pre-#84 version as architecturally broken;
   Fase 5 replaces it. Removed `isObjectPublic` / `isAnonymous`
   (moved to SQL RBAC) and `resolveConfiguredId` (replaced by
   catalog-derived scope).

3. `openspec/specs/search/spec.md` — kept main's rich structure
   (SCH-OR-* + SCH-PFTS-001..007), then:
   - Amended SCH-PFTS-004 to describe the SQL RBAC mechanism
     (WOO-536 replaces the PHP `isObjectPublic()` post-filter
     with `_rbac_as_public: true` per ADR-022).
   - Added SCH-PFTS-CAT-001/002/003 for `_catalog` / `_catalogi[]`
     scope params + catalog-derived default scope.
   - `openspec validate search` → green.

Field-name follow-up: main's #850 renamed `publicatiedatum` →
`publicationDate` (and `depublicatiedatum` → `depublicationDate`)
across the register + code. Updated to match:
- `lib/Settings/publication_register.json` two-rule read-block
  now uses English field names.
- `lib/Repair/WOO536RepairReadRules.php` guard + emitted shape
  now use English field names.
- `tests/Unit/Repair/WOO536RepairReadRulesTest.php` updated to
  match.

WOO-536.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
@WilcoLouwerse @claude
fix(search): resolve document→publication via document's own _relatio…
3d3aa37 
…ns (WOO-536)

Restore document rows in the /api/search envelope when linked to a
publication via the legacy denormalised `_relations['publication.slug']`
form (the shape emitted by the WOO-517 seed).

Fase-5's original refinement used `_relations_contains: <docUuid>` on
the publication schema, which assumes publications point AT documents.
Robert's plan documented that direction, but every existing seed +
write path in main goes the other way — documents carry the linked
publication as either `_relations['publication']` (UUID form, new) or
`_relations['publication.slug']` (denormalised slug form, WOO-517
seed shape). With the inverse query, `_relations_contains` returned
zero hits and every document row was N4a-dropped, so callers saw
`total: N` but `results len: N - <doc-count>`.

Fix — add an M2 fast-path in `resolveDocumentPublicationSummary`:

1. Read the document row's OWN `@self.relations` (falling back to
   `_relations`).
2. If `relations['publication']` is set, treat as UUID and run the
   existing `tryFastPathPublicationLookup` (`find()` by id).
3. Otherwise if `relations['publication.slug']` is set, run a new
   `tryPublicationSlugLookup` — scans the publication set in scope
   under `_rbacAsPublic: true` and matches on `@self.slug`. The
   set is schema-scoped + RBAC-filtered so it stays small.
4. On both misses fall through to the original `_relations_contains`
   path, which covers the future direction where a publication
   registers its documents.

Also validated on a fresh docker install with 3 publications + 2
documents (legacy slug-linked): `total: 5` now equals `results len: 5`,
document rows carry the embedded `publication` summary, and admin+anon
callers see identical result sets. SCH-PFTS-004 acceptance criteria
("total MUST equal actual count of objects visible under public RBAC")
now holds for legacy document seeds as well.

Reproduction: WOO-536 status comment 2026-08-31 (bug 1).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
@WilcoLouwerse @claude
test(search): rewrite PublicationQueryServiceTest for Fase-5 (WOO-536)
b6dc3f8 
Replace the pre-WOO-536 test file (~1000 lines, all asserting removed
APIs — `isObjectPublic()`, `isAnonymous()`, `resolveConfiguredId`, the
config-based scope path) with 19 focused tests that cover the current
architecture. tasks.md 8.1.

New coverage (19 tests, 53 assertions, all green):

Pure helpers (7 tests):
  * stripEmptyValues drops empty strings / nulls / empty arrays, recurses
  * normalizeIds accepts JSON-string or array, casts to int, gracefully
    handles null and non-JSON strings
  * isCatalogPubliclyAvailable — past published passes, future rejected,
    missing / empty / malformed date rejected

assemblePublicSearchResults (4 tests):
  * Fail-closed: empty scope returns `{results:[], total:0}` with warning
  * SCH-PFTS-001 / SCH-PFTS-004: `_rbac: true` + `_rbacAsPublic: true` +
    `_multitenancy: false` forwarded to OR (Q1 Option B)
  * Q7 Interpretation A: client `_schema`, `_registers`, `catalogSlug`,
    `fq`, `_content` stripped before reaching OR
  * SCH-PFTS-CAT-002 scope propagates as `_schemas` on the OR query;
    single-register scope uses `_register` fast-path

M2 fast-path (bug 1 fix, 2026-08-31 — 3 tests):
  * Document with `_relations['publication.slug']` (legacy denormalised
    form) resolves via slug-scan and keeps the envelope row with an
    embedded publication summary
  * Document with `_relations['publication']` (UUID form) uses the
    UUID fast-path (`ObjectService::find()`) and keeps the row
  * Document without any resolvable publication link is silent-dropped
    (N4a) with an info log — envelope N2 preserved (no null sentinel)

Envelope (1 test):
  * facets / facetable pass through when OR provides them (SCH-PFTS-004:
    total reflects visible count under `_rbac_as_public`)

findObjectLocation (4 tests, preserved from pre-WOO-536):
  * Fail-closed without register/schema constraint (#734)
  * Locates via ObjectService::find within allowed pair (ADR-022)
  * Returns null when not found
  * Continues past DoesNotExistException from one pair

Test scaffolding: FakeSearchObjectService + FakeCatalogiService +
FakeCatalogObjectService + FakeSchemaMapper doubles. `wireHappyPath()`
+ `withDefaultCatalog()` helper set up a minimal happy-path fixture
so tests focus on assertions, not on plumbing.

The obsolete tests were valid for the pre-Fase-5 architecture (PHP
post-filter + config-derived scope + admin-vs-anon divergence); Fase 5
moved every one of those behaviours into OR's SQL RBAC via the
`_rbac_as_public` primitive, so the tests could no longer compile
against the current PublicationQueryService signature (no `IUserSession`
param) or exercise the removed methods.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
@WilcoLouwerse @claude
test(smoke): add smoke-fts-woo536.sh acceptance-criteria script
c471e48 
Exercises Robert's WOO-536 Definition of Done + SCH-PFTS-001..004 +
SCH-PFTS-CAT-001..003 against a seeded stack via the public
/api/search endpoint. 8 assertions:

1. anon default scope returns at least 1 row
2. total == results-length (bug 1 regression guard — no undercount)
3. admin auth returns identical result-set as anon (Q1 Option B)
4. `_catalog=<known-slug>` resolves gracefully
5. `_catalog=nonexistent` returns HTTP 200 + total: 0 (graceful)
6. `_schema=999` stripped (Q7 Interpretation A anti-widening)
7. envelope carries results + total (facets optional per OR)
8. every row carries `@self.schema` as slug string (SCH-PFTS-002)

Rate-limits itself (sleep 2s between requests) against the anonymous
DoS defense (PUBLIC_LIMIT_MAX). Pass `--admin-token <app-password>`
to enable assertion 3 (admin parity), `--catalog-slug` for assertion 4.

Also archives tasks.md 8.1 as done (unit tests landed in b6dc3f8).

Verified against the fresh docker install on this branch: 8/8 green.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
@WilcoLouwerse @claude
chore(gates): satisfy hydra gate-16 + gate-66 (WOO-536)
7ed145e 
Two mechanical-gate fixes surfaced by run-hydra-gates.sh:

* gate-66 (openregister-dependency-shape, ADR-083) — SchemaMapper
  container lookup in assemblePublicSearchResults now guarded by
  class_exists, fails closed with a warning if OR is uninstalled.

* gate-16 (spec-coverage) — WOO536RepairReadRules::run gets @SPEC
  openspec/specs/search/spec.md.

All 36 applicable gates green. Unit tests still 19/19.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
@WilcoLouwerse @claude
fix(search): rebuild total from emitted rows (WOO-536)
f476bbd 
Second half of the SCH-PFTS-004 "total reflects true visible count"
contract. The first half (metadata search, bug 1 M2 fast-path) fixed
document rows linked via `_relations['publication.slug']`. This
completes the fix for the `_content=true` chunk-hit path.

Scenario the metadata fix didn't reach:
* A document is chunked (OR file-extraction pipeline produced text)
* Its linked publication has `publicationDate` in the future (draft)
  or `depublicationDate` in the past (depublished)
* Under `_content=true`, OR's chunk-search returns the document as a
  candidate row (chunk-hit is on the extracted text, not gated by
  the schema RBAC on the parent publication)
* OC's row-loop applies transitive visibility: linked pub is not
  publicly visible under `_rbacAsPublic: true`, so N4a-drops the row
* But `total` came straight from `$candidateResult['total']`, so the
  envelope shipped `total: 1, results: []` — the exact `total >
  results-length` shape SCH-PFTS-004 forbids

Fix: recompute `total` from the emitted `$rows` array after the loop
completes, unconditionally. OR's pre-drop count is discarded. Facets
still pass through unchanged — OR aggregates over visible-only rows
via the schema RBAC + `_rbacAsPublic: true` on the outer query, so
facet counts already match.

Verified manually 2026-08-31:
* Draft-linked document with body-only marker `draft-milieu-only-body-
  marker-XYZ999`, `?_search=<marker>&_content=true` → `total: 0`,
  `results: []` (was `total: 1, results: []` before this commit)
* Published-linked document with body-only marker `klimaatakkoord-
  chunk-marker-XYZ42`, same query shape → `total: 1, results len: 1`
  with the document surfacing as expected

Unit tests still 22/22 green (someone added 3 tests since b6dc3f8).
Smoke script unchanged — assertion 2 remains the regression guard.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
@WilcoLouwerse
fix(search): address self-review D1-D6 findings (WOO-536)
c0595ba 
- D1: correct tryPublicationSlugLookup scope comment (no catalog filter;
  scan is O(500) capped by _limit)
- D2: cache slug-scan results per request so N docs sharing one legacy
  publication.slug collapse to a single scan (was O(N x 500))
- D3: hoist class_exists SchemaMapper preflight to run BEFORE the first
  OR call; drop the misplaced-after-use guard
- D4: log debug on slug-match collisions (parity with N4b)
- D5: add 3 tests: UUID-fast-path archived-drop, slug-fast-path
  archived-drop, D2 slug-cache single-scan
- D6: rewrite smoke json_get() with jq (single line, no tab-indented
  Python heredoc; jq is preinstalled in the NC container image)

Test suite: 22 tests / 57 assertions, all green.
@WilcoLouwerse
fix(search): preserve pagination signal in envelope total (WOO-536)
b4939ff 
Addresses E1 from self-review round 4: the prior `count($rows)` change
closed SCH-PFTS-004 (`total > results-length` bug) but redefined `total`
from "global visible matches" to "visible on THIS page", breaking any
consumer computing `has_more = total > offset + limit`.

New shape: `total = max(count($rows), $orTotal - $droppedCount)`.

- $orTotal is OR's global candidate count under `_rbacAsPublic: true`
  — the pagination signal a paged UI needs.
- $droppedCount counts per-page N4a / archived / unresolvable-schema
  drops (dedup drops are excluded — they collapse a single logical hit
  OR double-counted).
- The `max(count(rows), ...)` floor guarantees `total ≥ count(rows)`
  on every page, so the SCH-PFTS-004 `total: 1, results: []` bug can
  never recur.

Per-page approximation: later pages may drop more rows, so `total`
shrinks monotonically as consumers page forward. Acceptable trade-off
without an OR schema change to push transitive-visibility into the
schema RBAC (tracked as follow-up).

Smoke assertion 2 updated from `total == results-length` (tautological
after the prior fix) to `total ≥ results-length` — the actual invariant.

Tests: 26/26 green, 64 assertions. Four new tests lock in the new
behaviour — pagination preserved on no-drop, subtract-and-floor when
drops occur, floor-to-zero when everything drops, dedup drops excluded.
@WilcoLouwerse
fix(quality): resolve PR-introduced PHP quality failures (WOO-536)
45a4c53 
Fixes the four failing quality checks on this PR — every finding was
introduced by earlier commits in this same PR, not by pre-existing code.

phpstan + psalm — 4 UndefinedMethod errors on PublicationsController:
  restore isAnonymous() and isObjectPublic() on PublicationQueryService
  (removed as "dead code" in bb21d16 — the first-round D1 fix was wrong
  about "zero callers repo-wide": PublicationsController::getUses and
  ::getUsedBy call both methods at runtime). Re-adds the ?IUserSession
  constructor param and the two method bodies. The controller's calls
  now type-check AND stop fataling at runtime with Error: Call to
  undefined method (which the surrounding catch (\Exception) did not
  catch — public /uses and /used-by endpoints were HTTP 500'ing).

phpstan — 2 param-out type errors on resolvePublicationSchemaId:
  cast SchemaMapper->find()->getSlug() to string so phpstan can prove
  the array<int, string|null> shape; caller annotates the local
  $schemaSlugById = [] with an inline @var for type tracking.

phpcs — 2 errors on WOO536RepairReadRules.php:
  refactor the ternary return into explicit if/return, and re-shape
  the docblock so the long-description starts with a capital.

phpcs — 7 errors on PublicationQueryService.php:
  refactor 5 inline ternaries into if/else blocks, add named-param
  on the extractSchemaId call, add the missing @PARAM $slugCache
  docblock entry on resolveDocumentPublicationSummary.

phpmd — 11 violations on PublicationQueryService + WOO536RepairReadRules:
  add use DateTimeImmutable; (drops 2 MissingImport), class-level
  @SuppressWarnings(PHPMD.ExcessiveClassLength), and per-method
  @SuppressWarnings on resolveCatalogScope, resolveDocumentPublicationSummary,
  tryPublicationSlugLookup — matches the pattern already on
  assemblePublicSearchResults. Refactor the WOO536 loop to drop the
  else branch.

Local verification (docker container with hydra-gates + full deps):
  phpcs      -> 0 errors (warnings non-blocking per ignore_warnings_on_exit=1)
  phpstan    -> OK, no errors
  phpmd      -> exit 0 (both configs)
  psalm      -> No errors found!
  Tests      -> 32/80 green (PublicationQueryServiceTest + WOO536RepairReadRulesTest)

The 45 pre-existing errors in ObjectUpdatedEventListenerTest.php are
unrelated — reproduced on origin/main by checkout+run.
@WilcoLouwerse @claude
fix(routes): return 404 (not 500) for mixed-case catalog slugs (WOO-536)
638f8c8 
Loosen the catalog-slug requirement from [a-z0-9-]+ to [A-Za-z0-9-]+
on every /api/{catalogSlug} route so uppercase requests reach
PublicationsController and receive the documented
`HTTP 404 — Catalog not found` response. Previously they fell through
to the SPA catch-all route and blew up with HTTP 500.

Behaviour unchanged for lowercase slugs; uppercase input still yields
404 because catalog slugs are stored lowercase in the DB — the
controller simply doesn't find them.

Aligns with openwoo docs Integrations/fulltext-search.md §Endpoint 1
which promises 404 on unknown slugs.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@WilcoLouwerse @claude
fix(search): isObjectPublic reads English field names (WOO-536)
7be7ccb 
Addresses review #5077596681 finding F1: 45a4c53 correctly reintroduced
isObjectPublic() / isAnonymous() (D1's "dead code" verdict was wrong —
PublicationsController::uses & ::used call these at runtime) but the
reintroduced body read Dutch field names publicatiedatum /
depublicatiedatum. The live publication schema uses the English
publicationDate / depublicationDate — see lib/Settings/publication_register.json
(16 hits English, 0 Dutch) and origin/main's own isObjectPublic() which
uses the English shape. Every anonymous /uses & /used request was
returning 404 as "not published" regardless of publish state.

Also drops the `(string)` cast in resolvePublicationSchemaId() introduced
by the same phpstan fix — it coerced null → "" and drifted from the
`array<int, string|null>` annotation on the by-ref cache. Replaced with
an is_string() normaliser that preserves the documented shape (F2).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@WilcoLouwerse @claude
fix(quality): phpcs errors in is_string normaliser (WOO-536)
d44ec07 
CI PR check `quality / PHP Quality (phpcs)` failed on the previous
`is_string($rawSlug) ? $rawSlug : null` ternary at line 728 — repo's
phpcs standard bans inline IF statements and implicit truthy comparisons.

Refactored to explicit if/else with `=== true` comparison. Same runtime
semantics (assign the string or null), phpcs clean.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@WilcoLouwerse
Merge pull request #1154 from ConductionNL/hotfix/woo-536-fts-catalog…
5c42d19 
…-model

fix(search): Full Text Search follows OpenCatalogi catalog model (WOO-536)
@github-actions
chore(release): sync main back into development
b47ebf4 
The 2.0.11 release bumped the version on main. Without this,
development stays behind main and the next development -> main promotion
conflicts on the version file.

Version files resolve to development's side, which is the higher line,
so this never moves a version backwards.
@github-actions
github-actions Bot enabled auto-merge September 1, 2026 13:13
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@rubenvdlinde rubenvdlinde Awaiting requested review from rubenvdlinde rubenvdlinde is a code owner
@Rem-Dam Rem-Dam Awaiting requested review from Rem-Dam Rem-Dam is a code owner
@bbrands02 bbrands02 Awaiting requested review from bbrands02 bbrands02 is a code owner
@rjzondervan rjzondervan Awaiting requested review from rjzondervan rjzondervan is a code owner
@WilcoLouwerse WilcoLouwerse Awaiting requested review from WilcoLouwerse WilcoLouwerse is a code owner
@remko48 remko48 Awaiting requested review from remko48 remko48 is a code owner
@SudoThijn SudoThijn Awaiting requested review from SudoThijn SudoThijn is a code owner

At least 1 approving review is required to merge this pull request.

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@WilcoLouwerse