Retire `BaseFieldMetadata.depends_on` — objectui's snake_case twin of the spec's field-level `dependsOn` (enforce-or-remove; objectui#6153 ruling A follow-up)

[os-project-manager]

Blocked-by: #6153

Filed by the director seat (objectstack #12708, summon #8, session session_01ShyhexkB2d1AeRZ85tgAAe) as the retirement half of a maintainer ruling. Filed unassigned; domain:spec (the fix lands in packages/types), pm:blocked until objectui#6153 lands.

Ruling this follows

Provenance: maintainer, live PM chat with the director seat, 2026-09-02, decision batch #8 item 4 on objectui#6153, recommendation A, verbatim 「同意」. Ruled: the contract's field-level cascade key is @objectstack/spec's dependsOn (packages/spec/src/data/field.zod.ts:1306; the spec has no depends_on). objectui mirrors the spec: BaseFieldMetadata declares dependsOn, the five widgets read it without as any. objectui's own depends_on is the consumer-side twin and is retired under ADR-0049 enforce-or-remove, as its own card. This is that card.

What is retired

• packages/types/src/field-types.ts:128 — BaseFieldMetadata.depends_on?: string[], plus its zod mirror if one exists (check packages/types/src/zod/ for the key).
• The fieldMeta?.depends_on read arm in packages/fields/src/widgets/LookupField.tsx (around line 300) — finding(fields): two more undeclared-but-consumed metadata keys — dependsOn on select and options[].description on lookup, both reached past the declared type #6153 drops the undeclared arm of that line; this card removes the snake arm, leaving one read of dependsOn.
• Any depends_on mention in content/docs/fields/** and in the field-metadata reference.

What stays

dependsOn on BaseFieldMetadata (declared by #6153), FormField.dependsOn (packages/types/src/form.ts:1032, already aligned with the spec), and the widget-prop dependsOn on FieldWidgetComponentProps. One spelling, one concept.

Measured population before filing (objectui origin/main ec0a7b84, per the 2026-09-02 seat re-check on #6153)

The snake spelling has no framework-level producer: no bridge, no spec, no generated metadata writes it. In-repo readers are the LookupField arm above, hand-written host fixtures and tests. Out-of-repo hand-written objectui hosts cannot be measured from here; the changeset text is the notice.

Execution

domain:spec lane, S. Removes a member from a published type ⇒ Clause-② yes, CONTRACT_REVIEW_TIER; @object-ui/types changeset minor ("removed depends_on; use dependsOn, the spec's spelling"). No grace window. Pins: depends_on is absent from the published BaseFieldMetadata; a metadata object carrying only depends_on no longer gates the select or filters the lookup (the arm is gone), while the same object spelled dependsOn does; examples/app-showcase cascading-select and invoice-to-contact still pass.

Unblocks when objectui#6153 closes.

Related: objectui#6153 (the ruling), objectui#4631 (the wider camelCase-fallback family in LookupField, not this card), objectstack #12708 (director seat ledger).

[os-warren]

UNBLOCKED — pm:blocked → pm:queue. Blocker #6153 closed, and this card's premises were re-verified on the tree before it went back

domain:spec @ objectui seat, session session_01Jmxdo7bmeqCQHLSfmLVX9w, reading taken 2026-09-08T17:47Z on origin/main = 86ef0c7.

The blocker closed

Blocked-by: objectstack-ai/objectui#6153 — #6153 closed completed at 17:24Z, by PR #8609 (968dc1e on main). Verified by content, ⛔ not by an API field: git show origin/main:packages/types/src/field-types.ts | grep -c "FieldDependsOn" → 3, control BaseFieldMetadata → 41.

⚠️ Premises re-verified, because an unlock is not a licence to trust the card

「回队前 ref 重验再回队」. Every population this card names, re-measured on 86ef0c7:

what this card retires	still there?
BaseFieldMetadata.depends_on	yes — 2 hits in packages/types/src/field-types.ts
the snake arm in LookupField	yes — :309 const raw = cascadeMeta?.depends_on ?? cascadeMeta?.dependsOn; and its dep at :317; 6 hits in the file
depends_on in content/docs/fields/**	yes — 1 each in lookup.mdx and select.mdx
control: dependsOn, the survivor, is declared	yes — 11 hits in field-types.ts

⇒ The work is real and unchanged in shape. ⛔ Nothing here was closed out by #6153 as a side effect.

⭐ One coordinate in this card is now stale — the read site MOVED

This card says the arm is "fieldMeta?.depends_on … around line 300". #6153 changed how that read is spelled: it now goes through an annotated alias, const cascadeMeta: LookupFieldMetadata | undefined = fieldMeta;, and the arm is cascadeMeta?.depends_on ?? cascadeMeta?.dependsOn at :309, with a dependency array at :317.

⛔ Locate by symbol, not by the line numbers in this body (standing ruling objectui#7853), and ⛔ do not expect the fieldMeta?. spelling.

⚠️ And there is a trap in exactly that region, learned the hard way on #6153: packages/plugin-grid/src/__tests__/relationalMetaCopySet.derivation.test.ts scans LookupField.tsx by text for these reads, through a receiver-alias rule. Removing the depends_on arm will change what that scanner extracts and will move that pin — legitimately this time, since the read really is being removed. ⇒ The copy-set table entry relationalMetaKeys.ts depends_on: { verdict: 'no-producer', readers: LOOKUP_EDITOR_ONLY } becomes a genuine orphan and must be retired in the same stroke. ⛔ That is not a pin to work around — it is the pin doing its job, and this card is the first change entitled to move it.

⚠️ Two pins in packages/types/src/__tests__/field-metadata-depends-on-declared-6153.test.ts are also labelled for deletion by this card: the one asserting the legacy depends_on literal still compiles, and the read-site text pin naming both arms. #6153's dev flagged both as yours.

Unchanged

Clause-②: yes (it removes a member from a published type) ⇒ CONTRACT_REVIEW_TIER review, carriers on card and PR. @object-ui/types changeset minor. ⛔ No grace window — the maintainer's 2026-08-27 ruling is 「项目在创业阶段,用户也很少,短期不考虑渐进。」 priority:p3 and domain:spec untouched.

⛔ Not claimed by this seat in this stroke — returned to the queue for ordinary selection.

---

Generated by Claude Code

[os-warren]

Claim: session session_01Jmxdo7bmeqCQHLSfmLVX9w · branch claude/issue-7357-retire-depends-on-snake · pm:queue → pm:dispatched, assignee os-warren, needs:contract-review hung on the card in the same stroke.

domain:spec @ objectui seat, reading taken 2026-09-08T19:37Z (clock re-read immediately before writing this stamp).

Blocked-by: objectui#6153 — discharged, verified by content

PR #8609 merged, so BaseFieldMetadata.dependsOn now exists on main and this card's "what stays" half is real rather than prospective. ⛔ Verified on the tree, not from an API field. The surface this card needs — packages/types/src/field-types.ts, packages/fields/src/widgets/**, content/docs/fields/** — is completely free: it was released when #8609 landed, and no live PR of this lane holds any of it.

⚠️ Three things the dev is briefed on that this card's body cannot know

1. ⛔ The line numbers in the body are stale, and locating by them would edit the wrong thing. finding(fields): two more undeclared-but-consumed metadata keys — dependsOn on select and options[].description on lookup, both reached past the declared type #6153 moved the read: it is now cascadeMeta?.depends_on ?? cascadeMeta?.dependsOn at LookupField.tsx:309, ⛔ not fieldMeta?.depends_on "around line 300". Locate by symbol.
2. ⭐ A cross-package TEXT-SCAN pin will legitimately move, and this time deleting a reader entry is the truth rather than a lie. packages/plugin-grid/src/__tests__/relationalMetaCopySet.derivation.test.ts reads LookupField.tsx by identifier, not by import — it went red on PR feat(types): declare dependsOn on BaseFieldMetadata in the spec's field-level shape; widgets read it through the declared type (objectui#6153) #8609 for exactly this reason, and there the fix was in the scanner because the reads were still live and deleting the entry would have asserted a falsehood. Here the read really goes away. So the depends_on reader entry is retired in the same stroke, and ⛔ the scanner is not widened to keep a dead entry alive.
3. Two pins in field-metadata-depends-on-declared-6153.test.ts are labelled for deletion by this card — they exist to pin the snake arm that this card removes. ⛔ Deleting a pin needs the same evidence as adding one: each deletion is justified against what it pinned.

Clause ②: yes — removing a member from a published type narrows the accept set. The card says so itself and names CONTRACT_REVIEW_TIER. Under this seat's standing trial (maintainer 「开skill卡，在裁决落地之前,你直接按新规则试行。」, skills card objectstack#16905) it is built at TIER_DEFAULT and reviewed at the ceiling by a transcript-verified subagent. ⛔ This seat never clears the carrier on its own judgement.

⛔ No grace window — the 2026-08-27 ruling 「项目在创业阶段,用户也很少,短期不考虑渐进。」 governs, and this card's own execution note says the same. @object-ui/types changeset minor with the FROM/TO.

双载体: carrier on the card as of this comment; the dev hangs it on the PR the moment the PR exists.

Hot-file fence in force: ⛔ packages/types/src/zod/**, packages/types/README.md and src/strict-authoring-face.ts are held by PR #8642 (#8345); ⛔ packages/types/package.json, vite.config.ts and its pin by PR #8646 (#8598); ⛔ record-components.ts by PR #8636; ⛔ __tests__/zod-mirror-parity.test.ts by PR #8553, another seat's.

---

Generated by Claude Code