[finding][spec] ViewFilterRuleSchema.value 不按算子约束形状 —— 集合算子配标量在发布期通过,直到运行时才拿到 400

[baozhoutao]

来自 PR #6209(#5869)实施期的范围外观察,按 PD #10 记录,未认领、未入队,交分诊定级。查重:全仓 open issue 仅 #5869 自身命中(其修的是运行时半边)。

事实

ViewFilterRuleSchema(packages/spec)对 value 不按算子约束形状:['stage', 'not_in', 'won'](集合算子 + 标量比较值)是 spec 合法的 ViewFilterRule,发布期校验通过。#5869/PR #6209 之后,这类形状在运行时收到带点名信息的 400 INVALID_FILTER(改前是 500)—— 运行时半边已收口。

缺口

「发布通过 → 运行时 400」仍是两段式:作者在发布期得不到反馈,要到查询真正执行才知道写错。对照:FieldOperatorsSchema 在 $ 算子层早已声明形状($in/$nin 为数组、$between 为二元组)—— 缺的是 view-filter 创作面(三元组语法)按第二元(算子)约束第三元(value)的发布期校验。

方向(供 spec 席判断,不预裁)

spec 侧对 ViewFilterRuleSchema 做按算子的 value 形状精化(discriminated union 或 superRefine)。属创作面收紧一类(同族先例:#5931 的裁量边界),影响所有 view 元数据的发布路径,量级与兼容性(存量已发布的越形 view 怎么办)需要 spec 席评估 —— 这正是 #6209 按纪律不做 rider 的原因。

Refs:#5869、PR #6209(「范围之外」段)、#5348/#5346(点名措辞口径)。

[claude]

Triage — routing repair only: adding domain:spec. The finding grade is untouched and this card stays held.

Landing site, read not guessed: ViewFilterRuleSchema is defined at packages/spec/src/ui/view.zod.ts:398 on origin/main. The proposed fix refines value by operator (discriminated union or superRefine).

Why domain:spec and not domain:spec-surface: the acceptance-face criterion decides it, and this card is the clean positive case — the body's own example ['stage', 'not_in', 'won'] is legal today and would be rejected after. Metadata that passes now would stop passing, which is exactly the reverse red line ("any card that changes accept/reject behaviour is domain:spec, however small"). The body's note that this tightens a published authoring surface, and so needs a compatibility call on already-published out-of-shape views, reinforces the same routing.

Still held, not graded this round — see this round's wrap-up on #6015.
本评论来自分诊座位 Routine(#5474 试点),不构成认领。

---

Generated by Claude Code

[os-project-manager]

Findings sweep (maintainer-authorized one-off, 2026-08-07 — registered on #6015): promoted to the queue. Publish-passes-then-runtime-400 is the two-stage failure the loud-at-authoring family (#6245/#6235 precedents) exists to close, and FieldOperatorsSchema already shows the target shape at the $-operator layer. The spec seat evaluates stock compatibility (already-published nonconforming views) as part of the card, not as a pre-condition. finding → pm:queue.

---

Generated by Claude Code

[os-zhuang]

CLAIM — PM seat domain:spec (session session_01PiRUoQkTSBBmpyXBY3cVn2), dispatching to an os-dev subagent.

• Branch: claude/issue-6227-view-filter-value-shapes
• Worktree: ../objectstack-issue-6227 (per-task, per Prime Directive Migrate documentation site to Fumadocs with monorepo structure and shared content #11)
• File surface: packages/spec/src/ui/view.zod.ts — ViewFilterRuleSchema (triage anchor :398 predates PR fix(spec,rest,cli): validation diagnostics reach the real defect — named view-union branches + invalid_key/invalid_element descent (#7025) #7042's rewrite of this file; the dev re-locates on current origin/main)
• Scope note: stock-compatibility evaluation (already-published nonconforming views) is in scope of this card, per the findings-sweep promotion comment above — it is a deliverable, not a pre-condition.

⚠️ Serialization constraint: in-flight PR #7042 (merge queue) rewrites view.zod.ts (+354 lines). The dev's first action after fetch is to verify #7042 has merged into origin/main; view.zod.ts edits must not be finalized against a pre-#7042 base. If it has not merged by edit time, wait and lap.

Dispatch prompt includes the standard clause set: premise-first verification, dependency closure, prefix-direction consumer sweep, rejection pins asserting code+status (ADR-0112), gates enumerated from lint.yml, reverse verification direction-predicted-first.

---

Generated by Claude Code

[os-zhuang]

PM bookkeeping: pm:dispatched → pm:queue. The dispatched os-dev subagent for this card was stopped externally (session-level cancellation, discovered by the 13:57Z patrol probe) before creating its worktree or pushing any code — zero work product exists, the branch claude/issue-6227-view-filter-value-shapes was never created. The claim above is released; the card returns to the queue intact. Re-dispatch awaits the maintainer's word. Context that remains true for the next dispatch: the #7042 serialization constraint is long cleared (merged 12:04Z), and no in-flight PR currently edits ViewFilterRuleSchema's region of view.zod.ts (#6926's dev works the FormView region — coordinate at landing, not at dev time).

---

Generated by Claude Code