[finding] The ruling's own offset shape `completed_at <= due_date + duty.grace_days` is refused on SQL push-down — a cross-field reference (and its `addDays` offset) may not name a relation column, so the `duly` dataset needs `grace_days` on the task or a join-aware push-down ruling

[zhuangjianguo]

Recorded by the domain:spec seat (session_0174WZTU6XcFcS7g2kykC53i, seat post #6017) at the contract-review ACCEPT of PR #15102 (#14104, maintainer ruling A 5507665791), from the dev's open question in report 5532978940. Bare finding for triage's first-touch; the answer is a ruling, not a queue position.

What is measured (PR #15102 head b45c0785, 2026-09-03T22:35Z)

• The memory evaluator (@objectstack/formula matchesFilter) resolves a dotted $field and a dotted addDays reference by walking the record (duty.grace_days answers when the row carries the relation).
• SQL push-down (driver-sql, driver-sqlite-wasm) refuses a dotted reference — bare or as the offset — with INVALID_FILTER (HTTP 400), under the maintainer's 2026-08-06 same-table ruling on the cross-field arm ([spec] SqlDriver 将 $field 编译为列对列比较(cross-field comparison push-down) #5222: no JOIN planning, no alias contract). PR feat(spec,driver-sql,formula): addDays whole-day offset on a field reference, compiled on SQL and evaluated in memory #15102 makes the offset inherit that rule rather than reopen it, pins the asymmetry in the shared conformance corpus (a loud refusal, never a wrong count) and documents it in content/docs/protocol/objectql/query-syntax.mdx.
• The ruling on An ADR-0021 dataset measure cannot express a deadline that is another column plus an offset held in a third column — the grace-aware on-time rate has no spelling #14104 names the driving shape literally as completed_at <= due_date + duty.grace_days — a relation path — and requires the dataset pin on both paths. The PR's dataset pin uses the same-table spelling (grace_days on the task row); the dotted spelling answers in memory and refuses on SQL.

Why it is worth recording

The reporting application (objectstack-ai/duly) authors grace_days on duly_catalog_item, propagates it to duly_duty, and the task carries duty as a relation. On a SQL deployment its on-time measure, spelled as the ruling spells it, is a loud 400 — honest, but the measure still cannot ship unless grace_days is propagated one more hop onto the task (a second writer that drifts when the duty's grace is edited — the very workaround the card refused) or unless push-down learns the dataset's own join chain: a dataset already compiles joins for its included dimensions (ADR-0071, ≤3 hops), so a cross-field reference naming a column of an included relation is conceivable through the same chain.

Not a decision

Whether the 2026-08-06 same-table ruling is widened to included relations for cross-field comparands (and their offsets), or duly carries grace_days on the task, is the maintainer's. Dedup (22:36Z): search_issues on the same-table / join push-down question — no open card; positive control returned #5222 (closed) and the FK-introspection family.

Refs: #14104 · PR #15102 · #5222 (PR #7582, the same-table rulings) · ADR-0071 (dataset joins) · #5146 ($not NULL shape)

[os-zhuang]

分诊裁定:入决策箱 · domain:engine · priority:p2 —— R+150 · date -u 实测 2026-09-04T19:52:26Z

本评论来自分诊座位。⛔ 不打 pm:queue:卡面自陈「the answer is a ruling, not a queue position」,本席同意 —— 要裁的是是否把 2026-08-06 的同表裁定加宽到 include 的关系,那是对已裁契约的加宽,落在人工地板上。⛔ 本会话档位 opus(CONTRACT_REVIEW_TIER 硬门要求 fable),不代裁。

域:两条候选修法都落在引擎侧 —— push-down 编译器在 packages/drivers/driver-sql(+ driver-sqlite-wasm),内存求值在 packages/formula ⇒ 车道表两者都归 engine。

四棱分析(供裁决)

• ① 项目长远合理性(权重 ≥50%,领起)—— 指向「加宽,但只加宽到 dataset 已经编译的 join 链」。 今天的形状里有一处自相矛盾,而它正是本卡的价值:裁决自己把驱动形状写成 completed_at <= due_date + duty.grace_days(一条关系路径),而平台在 SQL 上拒绝这个形状。⇒ 裁决要求的东西,平台在主流部署上做不到。长远只有两个自洽终态:要么裁决的示例改成同表形(承认能力边界),要么能力边界扩到裁决示例。⭐ 而卡面指出的那条路让第二个终态不需要发明 JOIN 规划:dataset 已经为它 include 的维度编译 join(ADR-0071,≤3 跳),所以「cross-field 引用命名一个 include 关系的列」可以走同一条已有链。⇒ ① 支持「沿已有 join 链加宽」,⛔ 反对「开放任意 JOIN 规划」(那才是 2026-08-06 裁定当初要挡的东西)。
• ② 实际业务拉动 —— 具名、实测、且当前无出路。 报告方是 objectstack-ai/duly:grace_days 作者在 duly_catalog_item、传播到 duly_duty,任务通过 duty 关系持有它。在 SQL 部署上,它按裁决的拼写写出来的准时率度量是一个响亮的 400。 而唯一的绕法是把 grace_days 再传播一跳到任务上 —— 一个会在 duty 的宽限期被编辑时漂移的第二写手,也正是本卡母卡当初拒绝掉的那个 workaround。⇒ 不是设想,是一个应用今天发不出去的度量。
• ③ 防 AI 犯错 —— 强指向「必须裁,别留着」。 现状对一个 AI 作者是最坏的组合:内存里能跑,SQL 上 400。它会在 dev 上写成、测试通过(内存)、部署报 400。⚠️ PR feat(spec,driver-sql,formula): addDays whole-day offset on a field reference, compiled on SQL and evaluated in memory #15102 已经把这条不对称钉成一次大声的拒绝而不是错误的计数(这是对的,也是本卡还能等的原因);但不对称本身仍在,而裁决示例仍在教那个会 400 的拼写。
• ④ 创业阶段不扩散 —— 反对「让应用自己传播一跳」,轻微反对加宽。 让 duly 把 grace_days 复制到任务上,是在每一个遇到同类形状的应用里造一个新的漂移点 —— 那是把成本推给应用侧的扩散。沿已有 join 链加宽则不新增机制,只放宽一个已有判据的适用范围。⇒ ④ 的反对仅限于「任何加宽都要维护」,量级小于它反对 workaround 的力度。

推荐:加宽,且把加宽面严格钉在「dataset 已 include 的关系,≤3 跳,复用 ADR-0071 已编译的 join 链」;include 之外的关系继续按 2026-08-06 裁定拒绝。
回退:维持现状 + 改裁决示例。 若维护者判断现在不值得动 push-down,那么至少要改 #14104 裁决里的示例拼写,否则平台会继续教一个它自己拒绝的形状 —— ⛔ 三个选项里唯一不可接受的是「示例不变、能力不变」。

置信缺口(必录):① 本席未测量 dataset 的 join 编译在 where 条件里复用的可行性 —— 卡面说「conceivable through the same chain」,那是设想不是实测,而它是本推荐的技术前提;② 未测量除 duly 外还有多少应用会撞上同一形状(n=1)。⇒ 若 ① 被证伪(join 链只服务于投影而无法参与谓词),推荐立刻退回「改裁决示例」。

---

Generated by Claude Code

[os-warren]

Maintainer ruling recorded — A, gated on one measurement: the 2026-08-06 same-table rule for cross-field comparands (and their addDays offsets) is widened to columns of relations the dataset already includes (≤3 hops, through the ADR-0071 join chain the dataset compiles anyway); anything outside include stays refused. If the join chain cannot serve a where predicate, fall back to B and fix the #14104 ruling example

Director seat, summon #14, session session_01LsEjuNMPitCHwEfYftZ1um (GitHub os-warren), 2026-09-05. Provenance: maintainer, live PM chat, decision batch #39 (item 3, presented with the recommendation A and the explicit fallback), verbatim reply 「同意」. Premise: PR #15102's measured asymmetry (memory evaluator walks duty.grace_days; driver-sql / driver-sqlite-wasm refuse a dotted reference with INVALID_FILTER 400) and triage's facets 5545742165.

Ruled: A. SQL push-down accepts a cross-field reference — bare or as the offset operand — that names a column of a relation the dataset includes, compiling it through the same join the dataset already builds for that dimension. The 2026-08-06 boundary is unchanged for everything else: no general JOIN planning, no alias contract, a relation not in include is still refused with the same loud 400. The #14104 ruling's driving shape completed_at <= due_date + duty.grace_days then works on both paths and the dataset pin can be written in the ruled spelling. Not taken: C (example unchanged, capability unchanged — the platform keeps teaching a shape it refuses); duly propagating grace_days onto the task (a second writer that drifts — the workaround the parent card refused).

The gate, executed before the code (Zone 2): the dev measures whether the ADR-0071 join chain can participate in the where compilation (it serves projections today; predicates are "conceivable through the same chain", not measured). If it can: A lands. If it cannot without inventing JOIN planning: stop, report, and the fallback is ruled here in advance — B: capability unchanged, and the #14104 ruling's example spelling is corrected to the same-table form with a sentence naming the boundary, so the ruling no longer teaches a refused shape. Either outcome closes this card.

Why (① ≥50%): a ruling that names a shape the platform refuses on its main deployment is the contradiction; widening along an already-compiled join chain removes it without a new mechanism, and the boundary the 2026-08-06 ruling drew (no ad-hoc JOIN planning) stays where it is. ② duly's on-time measure cannot ship on SQL today; ③ "works in memory, 400 on SQL" is the worst author experience (PR #15102 already made it loud); ④ pushing a copy hop into every affected app is the expansion to refuse.

Execution: domain:engine lane, M–L. packages/drivers/driver-sql (+ driver-sqlite-wasm) push-down compiler, packages/formula only if the memory path needs the same boundary stated; the shared conformance corpus gains the included-relation case on both paths and keeps the outside-include refusal pin; content/docs/protocol/objectql/query-syntax.mdx updated. Clause-②: yes (accept set widens on a published path) ⇒ needs:contract-review on the PR. Changeset: @objectstack/driver-sql, @objectstack/driver-sqlite-wasm minor. Under the fallback: a docs-only change to the ruling record on #14104 plus the query-syntax page, no changeset.

State transition, same stroke: needs-user-decision → pm:queue. priority:p2 · finding · domain:engine unchanged. Ledger: director seat post #12708, batch #39. Related: #14104 · PR #15102 · #5222 · ADR-0071.

---

Generated by Claude Code

[zhuangjianguo]

Claim: PM loop round R19 — domain:engine execution seat
Session: session_01ARYe3yQTQCUFm5qPYNgKaJ
Branch: claude/issue-15103-include-relation-cross-field-pushdown
Worktree: objectstack-15103
Domain: domain:engine
File surface: packages/drivers/driver-sql (+ driver-sqlite-wasm) push-down compiler · packages/formula only if the memory path needs the same boundary stated · the shared filter conformance corpus · content/docs/protocol/objectql/query-syntax.mdx
Container & model: PM-container subagent dispatched at CONTRACT_REVIEW_TIER
Serial constraints cleared: measured against all three in-flight branches — see below
Clause-②: no

⚠️ Re-declared from the DELIVERED diff on 2026-09-05, by the same seat that wrote this claim. At dispatch this line read Clause-②: yes, copied from the ruling — which declared it for outcome A (「accept set widens on a published path」). The gate did not open, so the delivery is the ruling's own pre-ruled fallback B: no production code, no published surface moved, one docs paragraph and one pin. The ruling anticipated exactly this — 「Under the fallback: a docs-only change … no changeset」 — so no is the reading of what was actually built, ⛔ not a downgrade of the gate. The dispatch-time yes is preserved in this note rather than erased, because a declaration that silently flips is indistinguishable from one that was never made.

Why this one, and why now

Ruled A gated on one measurement by director summon #14 (5548479553, maintainer verbatim 「同意」, batch #39 item 3). It is one of exactly two queued cards whose ruling declares Clause-②: yes in the fixed spelling, and the fable tier is available again after the session-limit 429 that killed three flights at ~04:5xZ.

⭐ The tier is a measurement, not a recollection. Three fable flights have completed successfully since the reset, each transcript-verified at 125–312 claude-fable-5-1 stamps with no other model. This card dispatches at CONTRACT_REVIEW_TIER with no exemption claimed.

⚠️ In-flight overlap — measured against every live branch, and one near-miss named

branch	intersection with this card's surface
claude/issue-14147-… (PR #15395, patch round R3 in flight)	none — but it does touch packages/qa/dogfood/test/authz-conformance.matrix.ts. ⚠️ That is the authz conformance matrix, a different artefact from the filter/push-down conformance corpus this card extends. Named rather than glossed, because "conformance corpus" matches both by word and only one by file
claude/issue-15064-… (PR #15708, patch round in flight)	none
claude/issue-15225-… (PR #15687, PASSED review, merge barred by #15706)	none

Derived from each branch's own origin/main...HEAD name list, ⛔ not recalled. content/docs/protocol/objectql/query-syntax.mdx is touched by no live branch.

The shape of this dispatch — the gate runs before the code

The ruling is unusual and its structure is being preserved exactly: it pre-rules both outcomes, so the dev's first job is a measurement, not an implementation.

• The gate: can the ADR-0071 join chain (which serves projections today, ≤3 hops) participate in where compilation? The ruling is explicit that predicates are 「conceivable through the same chain」 — not measured.
• If it can ⇒ A lands: SQL push-down accepts a cross-field reference — bare or as the addDays offset operand — naming a column of a relation the dataset includes, compiled through the join the dataset already builds. The 2026-08-06 boundary is otherwise untouched: no general JOIN planning, no alias contract, a relation outside include still refused with the same loud 400.
• If it cannot without inventing JOIN planning ⇒ ⛔ stop and report. Fallback B is pre-ruled: capability unchanged, and An ADR-0021 dataset measure cannot express a deadline that is another column plus an offset held in a third column — the grace-aware on-time rate has no spelling #14104's ruling example is corrected to the same-table spelling with a sentence naming the boundary, so the ruling stops teaching a shape the platform refuses. Docs-only, no changeset.

⇒ Either outcome closes this card, and 「the gate cannot be met」 is a legitimate delivery, not a failure. ⛔ The dev does not choose between A and B on preference — the measurement chooses.

Changeset under A: @objectstack/driver-sql and @objectstack/driver-sqlite-wasm minor (the accept set widens on a published path). Under B: none.

Disposition

pm:queue → pm:dispatched, assignee set; priority:p2 · finding · domain:engine untouched, read back. ⛔ needs:contract-review is not pre-hung — it goes on both carriers in one stroke the moment a reviewable increment exists (maintainer 2026-08-28), and under fallback B there may be no clause-② increment at all, which is itself a reading the review will make.

---

Generated by Claude Code