feat(quality): the quality assessment is @self metadata, not object data

[rubenvdlinde]

Why

A per-object quality score is an assessment of the object's data, not a fact about the thing the object describes. It had nowhere to live except the object body, which forced every schema wanting a score to declare qualityScore and qualityStatus as ordinary properties. Three things followed, and all three are wrong:

• The properties appear on every form the schema drives. A case handler filing a case in dossiq was shown a "Quality score" number field to fill in, for a value the platform overwrites on save. That is what surfaced this.
• Removing the declaration silently deletes the values, because the store strips what the schema does not declare. So the mistake could not be undone without data loss.
• Two schemas scoring the same way had to agree on property names by convention, with nothing to enforce it.

What

_quality becomes a metadata column of the same kind as _validation and _retention, surfaced in the @self envelope as quality:

"@self": { "quality": { "score": 0.5, "status": "fair", "scoredAt": "2026-09-03T…" } }

The change is additive on purpose. The listener still writes the body properties, but only where the schema actually declares them. Dropping that outright would freeze the stored value of every schema that has one, and a score that silently stops updating reads exactly like a score that is simply good. A schema that has migrated declares neither property, so nothing is written to its objects and the body stays clean.

That gives every app a migration path it can take on its own schedule: read @self.quality, then drop the declarations.

The migration

It sweeps every per-schema object table, adding the column where missing and skipping it where present, so re-running is a no-op. It identifies an object table by the _uuid column rather than by the name prefix alone, so a table that merely shares the prefix is never altered.

I verified that signature against a live database rather than only against the source: every oc_openregister_table_% carries _uuid and _validation, and none yet carries _quality.

Verification

• 1886 unit tests pass across tests/Unit/Db and tests/Unit/Listener. The pristine tree runs 1881; the 5 added here are the difference, and the 3 pre-existing vendor warnings and 5 skips are unchanged
• The listener had no test at all before this; there are now 5, covering the @self write, the schema that declares nothing, the schema that still declares both, a statusField naming a property the schema never declared, and a schema with no annotation
• PHPCS (repo config, lib scope) and PHPStan clean on every changed file
• Psalm reports one UnusedBaselineEntry on ObjectEntity, which reproduces identically on the pristine tree. It is an artifact of analysing single files, not something this branch introduces

Reviewer note

This touches the storage layer of the foundation repo, so the blast radius is every app. The two things worth your eyes are the migration's table-identification guard and the decision to keep writing declared body properties rather than cutting over hard.

🤖 Generated with Claude Code

[github-actions]

Quality Report — ConductionNL/openregister @ b79012a

Check	PHP	Vue	Security	License	Tests
lint	✅
phpcs	✅
phpmd	✅
psalm	✅
phpstan	✅
phpmetrics	✅
eslint		✅
stylelint		✅
build		✅
check-specs		✅
test-l10n		✅
test-l10n-parity		✅
format		✅
check-schema-l10n		✅
check-l10n-js		✅
composer			✅	✅ 174/174
npm			✅	✅ 543/543
app:check-code	⏭️
info.xml	✅
REUSE	❌
PHPUnit					✅
Newman					✅
Playwright					⏭️ deferred — runs on the promotion into beta/main, not on a pull request into development
Hydra gates					✅

---

Quality workflow — 2026-09-03 06:06 UTC

Download the full PDF report from the workflow artifacts.