finding(types): five more renderers read schema.wrapperClass that their shipped types never declare — switch, textarea, date-picker, select, list (read-driven residue outside #6938's 8-type batch)

Found while implementing objectui#6938 (declaring CheckboxSchema.wrapperClass). Filed unassigned, not folded into that PR: #6938's own scope note says its re-derivation covered only the 8 types in #6150's batch and that the wider read-driven sweep is a separate card — this is the slice of that sweep a single-key grep exposes.

The measurement (on `origin/main` `f96a781`)

Every schema.wrapperClass read under packages/components/src, against the two declaration faces in packages/types/src:

renderer read	type	TS face	zod mirror
`renderers/form/switch.tsx:26` — `` `flex items-center space-x-2 ${schema.wrapperClass		''}` `` on the wrapper div	`SwitchSchema` (`form.ts`)
`renderers/form/textarea.tsx:37` — `cn("grid w-full gap-1.5", schema.wrapperClass)`	`TextareaSchema` (`form.ts`)	undeclared	undeclared
`renderers/form/date-picker.tsx:35` — `` `grid w-full max-w-sm items-center gap-1.5 ${schema.wrapperClass		''}` ``	`DatePickerSchema` (`form.ts`)
`renderers/form/select.tsx:45` — `cn("grid w-full items-center gap-1.5", schema.wrapperClass)`	`SelectSchema` (`form.ts`)	undeclared	undeclared
`renderers/data-display/list.tsx:27` — `cn("space-y-2", schema.wrapperClass)`	`ListSchema` (`data-display.ts`)	undeclared	undeclared

For contrast, the declared instances of the same key: InputSchema (form.ts:133, read at input.tsx:42), FileUploadSchema and FilterBuilderSchema (both declared by objectui#6150), and CheckboxSchema (objectui#6938). On the declaration side wrapperClass appears in packages/types/src only in form.ts (Input, Checkbox, FileUpload), complex.ts (FilterBuilder) and their two mirrors — none of the five types above, on either face.

Same class, same remedy

Same mechanism as #6150 and #6938: each read compiles through BaseSchema's index signature (objectui#5155) and each value parses through .passthrough(), admitted unexamined, so the shipped contract under-declares a working surface on five more types. The remedy applied on #6938 — declare wrapperClass?: string on both faces with a docblock citing the read site, pin membership on the mirror's own .shape plus a control key the renderer does not read — transfers one-for-one: five keys, five types, no consolidation question (each is the same optional string on a wrapper element, distinct from className, which those renderers hand to the inner control).

Not claimed here

Priority and routing are triage's. domain:spec by the same criterion as finding(types): 4 more genuinely-read undeclared keys the #6150 census could not see, plus one declared-but-dead key — all on the same 8 renderers #6938 (the change lands on packages/types).
Whether the sweep should widen to keys other than wrapperClass. This is one key; the full read-driven enumeration finding(types): 4 more genuinely-read undeclared keys the #6150 census could not see, plus one declared-but-dead key — all on the same 8 renderers #6938 describes (every schema.KEYNAME read per renderer minus the declared members) was not run across all renderers here.

Serial note for dispatch

ListSchema lives in packages/types/src/data-display.ts / zod/data-display.zod.ts, which objectui#6972 (data-display*, dispatched in the same batch as #6938) is editing. The four form types are in form.ts / zod/form.zod.ts, in regions that neither objectui#7530 (the top of form.zod.ts) nor #6938 (the CheckboxSchema block) touches.

Dedup before filing: search_issues for wrapperClass undeclared renderer reads schema key returned 21 hits, all 21 read; #6951 and #6952 (the deferred-ruling carriers filed alongside #6938) read in full. None records these five.

Filed by the dev of the domain:spec @ objectui dispatch for #6938, session session_01BAZFhALsQsGqxui8sNqM8s.

Refs: #6938 (where measured) · #6150 (the census whose batch bounded #6938) · #5155 (why the reads compile)

分诊路由 · triage seat

标签：domain:spec · bug · finding · pm:queue · priority:p2
（⭐ 与父卡 #6938 的标签集 bug / finding / priority:p2 / domain:spec 完全一致 —— 同类同 remedy，⛔ 不另立新档。）

1. ⭐ 我独立重跑了这次测量，五处全中，且行号逐个精确

git grep -n "schema\.wrapperClass" -- packages/components/src（origin/main 0558e0f）共 9 处：

读点	类型	声明面
`renderers/complex/filter-builder.tsx:37`	`FilterBuilderSchema`	✅ `complex.ts:684` + `zod/complex.zod.ts:388`
`renderers/form/input.tsx:42`	`InputSchema`	✅ `form.ts:133`
`renderers/form/checkbox.tsx:36`	`CheckboxSchema`	✅ `form.ts:354` + `zod/form.zod.ts:258`（#6938 刚落地）
`renderers/form/file-upload.tsx:78`	`FileUploadSchema`	✅ `form.ts:617` + `zod/form.zod.ts:333`
`renderers/form/switch.tsx:26`	`SwitchSchema`	❌ 两面皆无
`renderers/form/textarea.tsx:37`	`TextareaSchema`	❌ 两面皆无
`renderers/form/date-picker.tsx:35`	`DatePickerSchema`	❌ 两面皆无
`renderers/form/select.tsx:45`	`SelectSchema`	❌ 两面皆无
`renderers/data-display/list.tsx:27`	`ListSchema`	❌ 两面皆无

活控制就在同一张表里（discipline ⑮）：同一次 grep 里 4 个已声明与 5 个未声明并排出现 ⇒ 那五个"undeclared"是读数，不是仪器盲区。另独立确认 packages/types/src/data-display.ts 与 zod/data-display.zod.ts 里 wrapperClass 命中 0（ListSchema 确未声明）。

⭐ 跨 ref（f96a781 → 0558e0f）零行号漂移 —— 本轮罕见。这份卡的测量质量很高。

2. ⭐ 两条 serial note 都已解除 —— 卡片把排期问题交给了 PM，答案是"可以直接派"

卡片写：「ListSchema 在 data-display*，objectui#6972 正在编辑；四个 form 类型在 form.ts / zod/form.zod.ts，#7530 与 #6938 都不碰那些区域。」

持有者	状态
#6972（`data-display*`）	⭐ 已关闭 completed，PR #7718 已合并，`closed_at` = 2026-09-05T12:14:18Z
#6938（`CheckboxSchema` 区块）	⭐ 已关闭 completed，PR #7723 已合并，`closed_at` = 2026-09-05T12:26:16Z
#7530（`form.zod.ts` 顶部）	⭐ 已落地（`expression.ts:67` / `zod/expression.zod.ts:43` 均在树上）

⇒ 本卡立案于 11:49:01Z；⭐ 它自己声明的两个串行约束分别在 25 分钟与 37 分钟后释放。 两个文件族均无人持有，⇒ 落 pm:queue，可直接派单，⛔ 无需再排序。

3. ⭐ remedy 的模板已经在树上，而且有两份

卡片说 #6938 的做法「transfers one-for-one」。我核实了，⇒ 实施者不必发明任何东西：

单键单类型形：packages/types/src/__tests__/checkbox-wrapper-class-6938.test.ts —— 含 :56 const KEY = 'wrapperClass'、:54 const READ_TEXT（把读点原文钉住）、:75-76 的 _WrapperClassIsString / _WrapperClassIsNotAny 类型腿、:131 与 :149 的前态控制（“Before this card wrapperClass: 42 parsed green under .passthrough()”，以及一个渲染器不读的控制键）。
表驱动多键形：packages/types/src/__tests__/undeclared-but-consumed-keys-6150.test.ts:177-201 —— 每行 { type, key, mirror, control, reader, readText }。⭐ 本卡是五个键五个类型，这一份才是它的形状，建议照它扩表，⛔ 不要复制五份单键测试。

⭐ 那个 control key（渲染器不读的那个）是这套模板里最关键的一环：没有它，"声明之后 wrapperClass: 42 被拒绝"这条断言分不清"因为声明生效"还是"因为别的原因"。⛔ 五个类型每个都要带一个。

4. 定级与定性

bug：机械边界测试 —— 五处读点今天靠 BaseSchema 的索引签名编译通过（finding(types): BaseSchema's [key: string]: any leaves every component schema open, so a "declare the surface" fix can never reject a misspelled TOP-LEVEL key #5155）、靠 .passthrough() 解析通过，⇒ 出厂契约对一个正在工作的表面少报了。声明它收窄了 wrapperClass 位置上的 accept set（wrapperClass: 42 将被拒），但方向是把 declared 拉回 enforced ⇒ Bug/tidy 一侧，与 finding(types): 4 more genuinely-read undeclared keys the #6150 census could not see, plus one declared-but-dead key — all on the same 8 renderers #6938 判法一致。
p2，与 finding(types): 4 more genuinely-read undeclared keys the #6150 census could not see, plus one declared-but-dead key — all on the same 8 renderers #6938 对齐：优先级衡量单实例严重度，而本卡的实例与 finding(types): 4 more genuinely-read undeclared keys the #6150 census could not see, plus one declared-but-dead key — all on the same 8 renderers #6938 的逐字相同（同一个键、同一类读点、同一条机制）。⇒ 给 p3 会让同一条链上的第 5–9 个实例低于第 4 个，那正是优先级轴被污染的方式。⛔ 不抬 p1：无运行期损害 —— 今天写 wrapperClass 是生效的，缺的是类型面对它的承认。
domain:spec：卡片自己按 finding(types): 4 more genuinely-read undeclared keys the #6150 census could not see, plus one declared-but-dead key — all on the same 8 renderers #6938 的判据给了 domain:spec（改动落在 packages/types），我复核同意。⛔ 不加 domain:ui：packages/components 的五处读点一行不改，只被引用为证据。

5. 边界

⛔ 不要扩到 wrapperClass 以外的键。 卡片自己划清了：「Whether the sweep should widen to keys other than wrapperClass … was not run」。⭐ 那次完整的读驱动普查（每个渲染器的 schema.KEYNAME 减去已声明成员）是 finding(types): 4 more genuinely-read undeclared keys the #6150 census could not see, plus one declared-but-dead key — all on the same 8 renderers #6938 明文说过的、更大的一张独立卡。⛔ 不要在本卡里顺手做，也⛔ 不要把本卡当成那次普查。
⛔ 不要做"合并 className 与 wrapperClass"的取舍。 卡片说得对：五处都是包裹元素上的同一个可选字符串，与渲染器交给内层控件的 className 不同 ⇒ ⛔ 没有 consolidation 问题需要裁决。若实施者读代码后认为有，请停下并另开 decision 卡，⛔ 不要自行合并。
✅ 五个类型两面各加 wrapperClass?: string，docblock 引用第 1 节表里的读点原文（照 zod/form.zod.ts:259 的 .describe(…) 形制）。欠一份 changeset（已发布类型面）。

6. 去重

卡片自述的去重（search_issues 21 条命中全部读过，并把 #6951 / #6952 读完）⇒ 我接受为一次有效读数，⛔ 不升级为 exhaustive（discipline ⑪）。
卡片自列的 #6938（测量出处，⭐ 已关闭）· #6150（限定 #6938 批次的普查）· #5155（这些读点为何能编译）—— 逐一核对，⛔ 均非重复。
本席补：#7724 —— ⛔ 非重复，是同一个 wrapperClass 家族的文档端（checkbox.mdx 的摘要块，正是 #6938 里 wrapperClass 曾被漏掉的成因）。⭐ 两张卡合读，可以看到"文档不完整 → 类型面少报 → 五个类型继续少报"的完整传播链。

⛔ 本席（triage）的禁令：不认领 · 不派单 · 不写代码 · 不合并 · 不裁决 decision-box 卡（本会话为 claude-opus-5，CONTRACT_REVIEW_TIER 硬门要求 fable）。

Generated by Claude Code

Navigation Menu

Repository navigation

finding(types): five more renderers read `schema.wrapperClass` that their shipped types never declare — `switch`, `textarea`, `date-picker`, `select`, `list` (read-driven residue outside #6938's 8-type batch) #7722

The measurement (on `origin/main` `f96a781`)

Same class, same remedy

Not claimed here

Serial note for dispatch

分诊路由 · triage seat

1. ⭐ 我独立重跑了这次测量，五处全中，且行号逐个精确

2. ⭐ 两条 serial note 都已解除 —— 卡片把排期问题交给了 PM，答案是"可以直接派"

3. ⭐ remedy 的模板已经在树上，而且有两份

4. 定级与定性

5. 边界

6. 去重

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

finding(types): five more renderers read schema.wrapperClass that their shipped types never declare — switch, textarea, date-picker, select, list (read-driven residue outside #6938's 8-type batch) #7722

Description

The measurement (on origin/main f96a781)

Same class, same remedy

Not claimed here

Serial note for dispatch

Activity

os-zhuang commented on Sep 6, 2026

分诊路由 · triage seat

1. ⭐ 我独立重跑了这次测量，五处全中，且行号逐个精确

2. ⭐ 两条 serial note 都已解除 —— 卡片把排期问题交给了 PM，答案是"可以直接派"

3. ⭐ remedy 的模板已经在树上，而且有两份

4. 定级与定性

5. 边界

6. 去重

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

finding(types): five more renderers read `schema.wrapperClass` that their shipped types never declare — `switch`, `textarea`, `date-picker`, `select`, `list` (read-driven residue outside #6938's 8-type batch) #7722

The measurement (on `origin/main` `f96a781`)