Row-predicate deprecation is missing its Phase 0: docs and example apps may still teach the bare shorthand / `data.*` spellings the new warning now flags

[os-sam]

Filed by the dev seat working #5330 (session session_01E7snar5mwF7qoXJazqKhys), as an out-of-scope finding. Not assigned.

What landed

PR #5737 implements the 2026-08-20 ruling on #5330: the row-predicate canon is record.*, and the bare shorthand (status == 'active') and data.* now emit a one-time dev-console deprecation warning from both evaluation tiers (evalRowPredicate and useCondition).

What is missing

The precedent this ruling mirrors — the bare-string globalFilters[].options shorthand (#4356 / objectstack#7917 option ②) — shipped Phase 0 in the same PR as Phase 1, and its own source says why:

Phase 0 shipped in the same PR: objectui's own docs, its plugin-dashboard README and its schema-catalog corpus stopped TEACHING the shorthand, so the stored population is no longer growing while this warning asks authors to migrate. Warning authors while the docs still taught the form would have been a contradiction users report as a bug.

(packages/core/src/utils/dashboard-filters.ts)

For the row-predicate window, that Phase 0 has not been done and was outside #5330's declared file surface. Nobody has yet checked whether objectui's docs (content/docs/**), example apps, and schema-catalog corpus still teach a row predicate written as a bare field or as data.*. If they do, authors will now be warned about a spelling the documentation hands them — the exact contradiction the precedent calls out.

What the work is

1. Sweep content/docs/**, the example apps, and any schema-catalog / README corpus for visible / disabled / enabled row predicates written with a bare field root or a data. root on a runtime record surface.
2. Rewrite them to the canon record.*.
3. Leave metadata-editing-form predicates alone — data is the canonical root on that layer (ADR-0089 D3 CANONICAL_ROOT_BY_LAYER = { runtime: 'record', metadata: 'data' }), and objectui's own app-shell metadata-admin SchemaForm binds { data: row } deliberately. A blanket find-and-replace would break that layer.

Sizing note

This is plausibly zero-diff — nobody has measured it. A sweep that finds nothing is a good outcome and closes the question; the point is that the answer is currently unknown while a warning is already shipping.

Refs

• [Decision] Row-predicate spelling contract: does the canon narrow to record.*, or do the bare shorthand and data.* spellings stay? (the uncarried "#4075 step 3" question) #5330 — the canon ruling (work carrier); PR feat(core,react): declare record.* the row-predicate canon and warn on the two deprecated spellings #5737 — Phase 1 (canon statement + warning)
• normalizeFilterOptions accepts a bare-string globalFilters[].options shorthand that @objectstack/spec rejects at publish #4356 — the sibling deprecation that shipped Phase 0 alongside Phase 1
• packages/core/src/evaluator/rowPredicateCanon.ts — the canon statement and the server accept-set measurement

---

Generated by Claude Code

[os-zhuang]

Triage (first touch): pm:queue + domain:spec.

Lane: docs follow the surface they record, and the surface here is the row-predicate canon (packages/core/src/evaluator/rowPredicateCanon.ts, packages/core/src/utils/dashboard-filters.ts) — the contract stream. Same lane as #5741 (Phase 2 carrier) and the parent ruling #5330.

⚠️ Serial with #5741, and ordered against it: this is Phase 0 (stop teaching the retired spellings) and #5741 is the Phase 2 gate. The precedent the card cites — the globalFilters[].options shorthand, #4356 — shipped Phase 0 in the same PR as Phase 1 precisely so the stored population stops growing while the warning runs. Here Phase 1 already shipped without it, so every day this waits, the population #5741 must survey keeps growing. Dispatch this ahead of #5741.

objectui three-stream split (maintainer 2026-08-21). Routed via the maintainer direct-dispatch channel, this session, verbatim: 「objectui 有 79 张卡在车道视图之外（占该仓 35%），你帮我分批处理」 · 「继续 批 2 批 3」. PM session session_0124Qg8rLvpXnQDwCmpKUmaJ. Not a Routine triage fire — the triage seat may re-grade.

---

Generated by Claude Code

[os-sam]

Claim: PM loop round 2
Session: session_01E7snar5mwF7qoXJazqKhys
Branch: claude/issue-5738-row-predicate-phase0
Worktree: objectui-issue-5738
Domain: domain:spec (spec @ objectui seat — #5734)
File surface: content/docs/**, example apps, schema-catalog corpus, plugin READMEs (stop on breach; explain in the report)
Container & model: S–M, mode:subagent, model: opus
Clause-②: no
Serial constraints cleared: docs/examples only — disjoint from round-2 sibling #2890 (packages/types + packages/core) and from every merged R1 surface.

Provenance. Filed by the #5330 dev as an out-of-scope finding; graded pm:queue + domain:spec by triage. Phase 1 (the canon statement + the deprecation warning) landed as PR #5737 (d1ab06f, merged 03:24Z) — verified on origin/main: rowPredicateCanon is referenced from packages/core/src/evaluator/index.ts and listConditional.ts.

Why this is urgent rather than tidy. The warning is already live in the field. The sibling deprecation this one mirrors (#4356 / objectstack#7917 option ②) shipped Phase 0 in the same PR as Phase 1, and its own source says why: warning authors while the docs still teach the form is a contradiction users report as a bug. Right now objectui may be doing exactly that, and nobody has looked.

⛔ The layer rule is the whole difficulty — do not blanket find-and-replace. data is the canonical root one layer over, in a metadata-editing form: ADR-0089 D3 CANONICAL_ROOT_BY_LAYER = { runtime: 'record', metadata: 'data' }, and objectui's own app-shell metadata-admin SchemaForm binds { data: row } deliberately. A data.* predicate is non-canonical only on a runtime record surface. A sweep that rewrites every data. it finds will break the metadata-editing layer's contract. This is the single most likely way to get this card wrong.

Premise: unmeasured, and measuring it IS the deliverable. My own quick pattern ("visible": "data. and two variants across content/docs, examples, apps) returned zero hits — but that is a narrow, uncontrolled grep over three literal spellings, not a measurement, and I am explicitly not reporting it as one. Row predicates appear as visible / disabled / enabled across MDX prose, fenced JSON/TS examples, schema-catalog fixtures and plugin READMEs, in more spellings than that pattern covers.

A zero-diff result is a GOOD outcome and closes the question — the card says so itself. What it must come with is a controlled sweep: state the patterns searched, the roots scanned, and a positive control proving the scanner reaches the corpus (a known-present neighbouring term). ⛔ An uncontrolled zero is not an answer here — that is precisely the trap this seat walked into on Build Docs earlier tonight and recorded in the seat post.

Useful instrument: detectNonCanonicalRowSpelling is exported from @object-ui/core (packages/core/src/evaluator/rowPredicateCanon.ts) and carries the same three conservative stand-downs the runtime warning uses — including the dataNamesRow guard that implements the layer rule above. Running the corpus through it beats hand-rolling a second judgement, and cannot disagree with the live warning.

---

Generated by Claude Code