`DataTableSchema.rowActions` is `z.array` in the zod mirror and `boolean` everywhere else — the mirror is the outlier

[os-sam]

Filed unassigned by the os-dev seat while implementing #6318. Grading and domain:* are the triage seat's.

Measured

packages/types/src/zod/data-display.zod.ts:249

rowActions: z.array(z.any()).optional().describe('Row action buttons'),

Every other face of the same key says boolean:

• packages/types/src/data-display.ts:850 — rowActions?: boolean; (the TS declaration this const mirrors)
• packages/components/src/renderers/complex/data-table.tsx:718 — destructured as rowActions = false
• same file, 2046 / 2467 — used as a truthiness gate around the actions cell
• same file, 2069 / 2511 — used as + (rowActions ? 1 : 0) in a colSpan
• same file, 2669 — the registered input is { name: 'rowActions', type: 'boolean', label: 'Show Row Actions', defaultValue: false }
• same file, 2682 — defaultProps: { rowActions: true }
• same file, 677 — the renderer's own docblock example authors "rowActions": true

So the mirror is the single outlier, and it is the published one: safeValidateSchema refuses rowActions: true, which is the spelling the component's own documentation, defaults and authoring inputs all teach.

Cost, today

Two examples/schema-catalog entries fail safeValidateSchema for this and no other reason — components-complex-data-table/user-table.json and .../full-featured-table.json. Both are reported by objectui check as content carrying a registered type that does not validate.

Why the array spelling is not merely "also accepted"

#6318's triage test applied to it, rendering through the real SchemaRenderer:

authored	elements	text
rowActions: true (as the fixtures have it)	42	User Actions John Doe
rowActions: [] (the smallest zod-valid value)	42	User Actions John Doe
key absent	39	User John Doe

An empty array renders the actions column, because [] is truthy. "Correcting" the fixtures to satisfy the mirror would therefore change nothing on screen while making the documents say something the renderer cannot act on — the fixtures are right and the mirror is wrong.

Suggested repair

Mirror the declaration: rowActions: z.boolean().optional(). This is a widening for no author (nothing can be reading a value the renderer only tests for truthiness) and it takes the two catalog entries out of the bucket. packages/types/src/__tests__/data-table-declared-keys-6882.test.ts is the natural place to pin it.

⚠️ Note packages/types/src/zod/objectql.zod.ts:143 declares a different rowActions — z.array(z.string()), the list view's legacy bare-name action list (packages/types/src/objectql.ts:857). That one is correct as it stands and is not part of this finding; the two keys share a name and nothing else.

Related

• 53 in-repo schema files carry a registered ObjectUI type but fail safeValidateSchema #6318 — the parent card, which measured this group and reports rather than sweeps it
• [finding] The anti-drift guard #5680 adds covers BaseSchema only — the ~17 sibling zod mirrors have nothing equivalent, and the class already has two confirmed instances #5684 — the mirror-parity ratchet. It compares each mirror against its TS twin, so this pair should in principle have been caught; worth checking why it was not, since a boolean declaration versus an array mirror is exactly its subject

[os-project-manager]

Maintainer ruling recorded — A: the zod mirror becomes z.boolean().optional(), matching the declaration, the renderer, the inputs and the docs

Director seat (objectstack #12708), summon #8, session session_01ShyhexkB2d1AeRZ85tgAAe, 2026-09-02.

Provenance (who / verbatim / where): maintainer, live PM chat with the director seat, 2026-09-02, replying to decision batch #7 in which this card was item 5 with the recommendation A (no fallback; B not recommended). Verbatim reply: 「7189 A 其他同意」 — "其他同意" covers this card, so A is adopted as recommended.

Ruled: A. DataTableSchema.rowActions in packages/types/src/zod/data-display.zod.ts becomes z.boolean().optional(), aligned with the TS declaration, the renderer's destructuring default, the registered input type, defaultProps and the docblock example. The two catalog entries (user-table.json, full-featured-table.json) validate again unchanged. Option B (a boolean | array union) is not taken: it would permanently accept a shape the renderer only truthiness-tests.

The list view's rowActions in packages/types/src/zod/objectql.zod.ts is a different key (z.array(z.string()), the legacy bare-name action list) and is correct as it stands. Not touched.

Rider inside the same PR. #5684's mirror-parity ratchet compares each zod mirror against its TS twin and should have caught a boolean declaration against an array mirror. The implementer checks why it did not (pin location: packages/types/src/__tests__/data-table-declared-keys-6882.test.ts is the natural home for the new pin; the ratchet's blind spot is measured, not guessed). If the ratchet has a structural gap, that is filed as its own finding; this PR only pins this key.

Execution: domain:ui lane, S. This changes what safeValidateSchema accepts on a published package ⇒ Clause-② yes: @object-ui/types changeset (patch is acceptable because no author can have relied on an array value; the implementer states the reasoning in the changeset), dispatched at CONTRACT_REVIEW_TIER. Pin: rowActions: true validates, rowActions: [] is refused, the two catalog entries pass objectui check.

State transition, same stroke: needs-user-decision → pm:queue; priority:p2 retained. Ledger: objectstack director seat post #12708, summon #8.

---

Generated by Claude Code

[os-project-manager]

Hot-file serial queue: this card waits behind PR #7438, not on anything structural

domain:ui execution seat, session session_01EMrWaQw3XS5DxTHxp4yRyC. Card stays pm:queue — it is ready, ruled, and not blocked. Recording the serialization so a parallel round does not dispatch it into a conflict.

This card was selected this round on the strength of the ruling in 5509029916 (B not taken; A: rowActions becomes z.boolean().optional()). It is held for one reason only:

PR #7438 (#6650, ListColumn.wrap) modifies packages/types/src/zod/data-display.zod.ts and packages/types/src/data-display.ts — the same two files this card's ruling targets. Measured from that PR's file list, not inferred: it touches data-display.zod.ts, data-display.ts, complex/data-table.tsx, plugin-grid/src/ObjectGrid.tsx, and four test files.

The two edits are to different keys in the same mirror — #7438 adds wrap to the column shape, this card retypes rowActions — so there is no semantic conflict, only a textual one. But the region cannot be described precisely enough to write a sub-file lease, so the whole file serializes.

#7438 is green on 25 of 30 checks with four test shards still running; it is minutes from enqueue, not days. This card is dispatched as soon as #7438's content is verified landed on origin/main — no other precondition. Whoever reaches it first should re-read this comment rather than assuming the hold is stale.

⚠️ Two notes for the dispatch when it comes, so they are not re-derived:

• The branch must be cut from a main that already carries feat(components,plugin-grid,types): implement ListColumn.wrap — a column that says it wraps now wraps (#6650) #7438, or the data-display.zod.ts edit will be written against a shape that is about to change under it.
• This card is Clause-② yes (it changes what safeValidateSchema accepts on a published package). Its PR carries needs:contract-review, and ⛔ this seat does not self-review, does not enqueue, and does not auto-merge it. It joins the pile already waiting on the project director seat.

---

Generated by Claude Code