ListViewBlock 仍带一份 dataSource 合成优先级的私有副本,与 #6953 落地的 ElementDataSourceGate 重复(漂移风险)

[os-zhuang]

Migrated from objectstack-ai/objectstack#7120 under the file-at-destination ruling (objectstack-ai/objectstack#7167, maintainer 2026-08-10). Originally filed 2026-08-09T17:46:26Z. The full prior thread — including triage rulings and hold/restart conditions — remains on the source issue and MUST be read before acting on this card.

---

发现于 objectstack#6953 的实施过程。观察类(observation-class):今天没有用户能碰到的错误行为 —— 两份实现当前语义一致,是未来漂移的风险面,故只打 finding、不入队。

事实

objectstack#5576 在 packages/plugin-list/src/ListViewBlock.tsx 内联写了「binding 覆盖组件键 / view 只作基线 / filter 走 and 合成 / 空 columns 视作未声明 / limit 落 pagination.pageSize / viewType 仅在未声明时取」这套优先级表。

objectstack#6953 为其余 8 个 block 接线时,把同一张表提到了 @object-ui/react 的共享件里(packages/react/src/element-data-source/ElementDataSourceGate.tsx 的 useElementDataSourceSchema + ElementDataSourceGate),因为一 block 一份就是「additional filter criteria」变成两种方言的路径。objectstack-ai/objectstack#6953 没有顺手改 ListViewBlock —— 那是 objectstack-ai/objectstack#5576 已合的落地代码,在一个接线 PR 里重构它是范围外的回归面。

于是现状是:同一张优先级表存在两份实现,list-view 用私有副本,其余 8 个 block 用共享件。

为什么是 finding 而不是 defect

两份当前语义等价(objectstack-ai/objectstack#6953 的共享件是照着 ListViewBlock 的规则写的,objectstack-ai/objectstack#5576 的整套测试在该 PR 下全绿:packages/plugin-list/ 60 文件 826 例通过)。用户今天碰不到任何差异。风险是下一次改语义的人只改一边 —— 这类漂移在本仓有前情(声明与读点各改一边)。

建议范围

把 ListViewBlock 收敛到共享件:

const LIST_VIEW_DATA_SOURCE = {
  columns: true, filter: true, sort: true,
  limit: 'pagination.pageSize', viewType: true,
};

配 ElementDataSourceErrorPanel / ElementDataSourceLoadingPanel(共享面板的 testId 前缀取 list-view 时,产出的 list-view-datasource-error / list-view-resolving-view 与 objectstack-ai/objectstack#5576 现有断言逐字符一致,面板文案用 errorTitle 传原句即可)。删掉 ListViewBlock 里的 useMemo 映射块(约 45 行)。

验收就是 objectstack-ai/objectstack#5576 那套测试不动一行仍全绿 —— 若有任何一例需要改断言,说明两份实现本来就有差异,那才是真 defect,应当在那时改判等级。

[os-zhuang]

Findings-triage: promoted — finding → pm:queue.

The hold recorded on the source thread (objectstack#7120, centralized pass 2026-08-10 00:15Z) listed as restart: "next card touching ListViewBlock, or the shared-sink-consolidation sweep together with #7148." The companion condition has fired: ui#4022 (ex-#7148) is promoted this round because its own gating PR (#4009) merged. Re-verified on origin/main @ 11c1e71: ListViewBlock.tsx still has zero ElementDataSourceGate / useElementDataSourceSchema references — the private priority-table copy is still in place, so the consolidation remains live work.

Acceptance criterion from the source ruling carries over: the existing #5576 test suite must stay green untouched — any assertion edit means the two implementations actually diverged, which is a defect to re-grade, not a refactor detail.

本评论来自分诊座位 Routine（#5474 试点），不构成认领。

---

Generated by Claude Code

[yinlianghui]

CLAIM (objectui whole-repo seat PM, session session_017Qqyix2QcnpUC9XeYVDzx3) — dispatching now.

• Branch claude/issue-4038-listview-shared-gate, worktree objectui-issue-4038 (S/M container)
• Scope = the card's suggested convergence exactly: ListViewBlock adopts ElementDataSourceGate/useElementDataSourceSchema with the card's LIST_VIEW_DATA_SOURCE mapping + the shared panels (list-view testId prefix keeping #5576's assertions byte-compatible), the private ~45-line useMemo priority table deleted. The acceptance criterion IS the card's own tripwire: #5576's existing suite passes UNTOUCHED — any assertion edit means the two implementations diverged, which is a defect to STOP and re-grade, not a refactor detail to absorb.

---

Generated by Claude Code

[yinlianghui]

ACCEPT (objectui seat PM, session session_017Qqyix2QcnpUC9XeYVDzx3, review of record) — PR #4267, CI fully green.

The convergence landed exactly as the card specified, and the two verdicts that mattered came back clean: the gate's mapping API was MEASURED against the card's shape before implementing (all five keys supported as-is — the shared component needed no change, additive or otherwise), and the tripwire acceptance held: #5576's suite passes UNTOUCHED, zero assertion edits — the two implementations had not diverged, so there is nothing to re-grade and the ~45-line private copy plus both hand-rolled panels are simply gone. The adapter bridge stays with its reason stated (ListView needs the adapter as a prop; the gate cannot supply it — a boundary, not an omission). Two touches of craft: the doc comment in packages/react that this change FALSIFIED ("still carries its own copy… belongs in both places") was corrected rather than left as a lie the PR itself created, and the reverse-verification note is honest about what CANNOT discriminate (the behavioral pins' green-through-revert is empirical confirmation of the equivalence premise, while the three seam pins are what catch a future re-fork). Flipping ready + auto-merge.

---

Generated by Claude Code