[finding] Three more expired post-#5702/#5710 status claims in `packages/spec/src/data` — including "`$icontains` … implemented by nobody" in the very file #6947's fix points `@see` at

[os-project-manager]

Observation-class finding, found by a completeness sweep while implementing #6947 (PR #6987). Filed unassigned, no domain:* label — routing is the triage seat's call. Nothing a user hits today: all four sites are prose. Each is the same drift class #6947 names (a status sentence names a future issue as the gap; that issue lands; nothing relates the two), but each needs its own measurement, which is why they were left out of #6987 rather than word-swapped inside it.

Anchors read on origin/main @ 30283263 — re-locate by text, not by line.

1. packages/spec/src/data/filter-text-conformance.ts:39-46 — the twin of the sentence #6947 fixed

 * This is the contract half of the #4706 ruling (#5701). `$icontains` is
 * declared by `StringOperatorSchema` (`filter.zod.ts`) and implemented by nobody; the
 * `$contains` family's case-sensitivity is declared and honoured by two of five
 * backends. Every driver therefore carries a measured DEBT row in
 * `scripts/check-driver-conformance.mjs`, pointing at **#5702**, which is the
 * issue that writes the lowerings and deletes the rows.

Sharpest of the four, because #6987's corrected docblock points straight here: its @see FILTER_TEXT_CASES — the conformance standard for every operator here lands the reader on "implemented by nobody".

Both numbers in it are now wrong, and in opposite directions:

• «$icontains … implemented by nobody» — measured for docs(spec): $icontains 实施状态按实测重写 —— #5702 已落地,五面里三面应答 (#6947) #6987 by executing all five: three answer (driver-sql; driver-sqlite-wasm, inherited, on the sql.js engine; driver-turso on both transports), two refuse with INVALID_FILTER / 400 (driver-memory, driver-mongodb — JS 求值面全体拒收 $icontains(driver-memory 两面 / driver-mongodb / objectql having / formula）—— SQL 族已实现,同一 filter 在内存 double 上抛错 #6520).
• «$contains family … honoured by two of five backends» — after drivers(sql family): 文本算子的大小写折叠是「方言的」而非「契约的」—— $contains 在 SQLite 过折叠、$icontains 在 PG/MySQL 过折叠 #6518 the SQL family is case-exact, i.e. three of five drivers, plus formula / objectql having / both service-analytics compilers which were already compliant. The two left are driver-memory and driver-mongodb (drivers(memory, mongodb): the $contains family still folds case — the last two backends left on the wrong side of #4706 Q2 = A #6682).
• «Every driver therefore carries a measured DEBT row» — the ledger now carries two FILTER_TEXT_CASES DEBT rows, not five; drivers(sql family): 文本算子的大小写折叠是「方言的」而非「契约的」—— $contains 在 SQLite 过折叠、$icontains 在 PG/MySQL 过折叠 #6518 deleted the three SQL-family rows.

Whoever fixes it must re-measure rather than copy the numbers above.

2. packages/spec/src/data/filter-text-conformance.test.ts:4-8

 * `FILTER_TEXT_CASES` is a standard no backend answers yet (#5701 is the
 * contract half of the #4706 ruling; #5702 writes the lowerings).

Three backends answer it today; driver-sqlite-wasm, driver-sql and driver-turso all import the whole case-set (that import is what scripts/check-driver-conformance.mjs reads as coverage). The paragraph's argument still stands — the file executes the table against a reference evaluator so the table cannot be quietly wrong — only its premise sentence expired.

3. packages/spec/src/data/filter.zod.ts:1305-1313 — the $regex retirement block (a DIFFERENT subject, same class)

 * Wiring them to it is **#5702**, deliberately not this PR: `$regex` still has
 * one live producer (`plugin-auth`'s ObjectQL adapter, on the authentication
 * path), so a refusal landing before #5710 flips that producer would break
 * sign-in. Hard order: **#5710 flips the producer, then #5702 turns these
 * strings into refusals.**

Both gates fired: #5710 closed (the producer was flipped to $contains; driver-memory/src/filter-refusal.ts records the whole-repo re-scan finding no other live producer) and #5702 closed, with requirement 3 done on all five faces per the conformance ledger. So a paragraph written as a future plan now describes shipped state, and its "hard order" warning is advice about a sequencing risk that no longer exists. #6987 deliberately did not touch this block: its subject is $regex, not $icontains, and stating the post-#5702 state honestly needs a five-face refusal census (does each site print RETIRED_FILTER_OPERATORS[op].why verbatim, in the ADR-0112 envelope) — a real measurement, not a word swap.

4. Same file, :227

  // `$options: 'i'` are #5702's work.

Reads as pending; mongo's hard-coded $options: 'i' is in fact still there (measured: translateFilter({ name: { $contains: 'acme' } }) returns {"name":{"$regex":"acme","$options":"i"}}), but its owner is now #6682, not #5702. So this one is stale by attribution rather than by verdict — the least urgent of the four and the easiest to get subtly wrong.

Why these were not folded into PR #6987

#6987 already discloses two scope extensions beyond its card (the $icontains .describe(), which is the rendered surface, and the FILTER_OPERATORS staging sentence, so the file does not contradict itself). Extending further would have meant shipping three new censuses — the $contains-family scoreboard, the FILTER_TEXT_CASES coverage count, and the five-face $regex refusal audit — inside a card scoped to one paragraph. #6947's dispatch was explicit that the general recurrence gate is #6635's design space and that its own scope was one instance. Filing beats a rushed word-swap: #6947 exists precisely because someone once wrote a status sentence they had not measured.

Suggested disposition

One card covering sites 1, 2 and 4 (all FILTER_TEXT_CASES scoreboard facts, one measurement pass answers all three) and, if triage prefers, site 3 separately since it is the $regex axis. Both are prose-only, domain:spec-surface-shaped, and neither changes behaviour.

Duplicate search before filing (filter-text-conformance stale, implemented by nobody, $regex retirement status 5702, 5710 flips the producer): zero hits.

Refs: #6947 / PR #6987 (the instance just fixed), #6520 ($icontains on the JS faces), #6682 (the $contains family on memory + mongodb), #6518 (case-exactness on the SQL family), #5710 (the flipped producer), #6635 (the general gate, on hold), #6842 / #6696 (earlier specimens of this class).

[os-project-manager]

Triage (pass #2, 2026-08-09, registered on #6015): promoted finding → pm:queue, routed domain:spec-surface. Same expired-status-claim class as the ruled #6947, with the filer's discipline already built in: three sites, each requiring its own census before the rewrite (no word-swaps). One card, three measured corrections, same acceptance shape as #6947's.

---

Generated by Claude Code

[os-project-manager]

认领此 issue。

• Session: session_018ffcE95NaMJcL9XJ9VDYgk
• Branch: claude/issue-6993-expired-status-census

按卡片要求执行:四处过期状态句各自重新测量($icontains/$contains 五驱动记分、FILTER_TEXT_CASES 覆盖与 DEBT 行计数、五面 $regex 拒绝审计),测量命令与结果将随 PR 一并提交;不改任何行为,只改注释文本(E22 日期化标注)。

Generated by Claude Code — session_018ffcE95NaMJcL9XJ9VDYgk

---

Generated by Claude Code

[os-project-manager]

范围披露(实施前,session_018ffcE95NaMJcL9XJ9VDYgk):三项普查已跑完(五驱动 + 辅助面记分、conformance ledger 读数、五面 $regex 拒收审计),四处站点按卡片改写。另有两处同文件、同一失效类的 case note: 字符串,若不同步改会与站点 1/4 的改写正文直接互相矛盾,故拟一并修正(#6987 的「文件不得自相矛盾」先例):

1. filter-text-conformance.ts $contains is case-SENSITIVE … case 的 note:「On SQLite/turso today LIKE folds ASCII and returns both」—— drivers(sql family): 文本算子的大小写折叠是「方言的」而非「契约的」—— $contains 在 SQLite 过折叠、$icontains 在 PG/MySQL 过折叠 #6518 后为假(实测三张 SQL 面均答 ['2'],GLOB 已替换 LIKE);
2. 同文件 $regex with $options is REFUSED … case 的 note:「plugin-auth: better-auth 的 contains 被译成裸 $regex,用户输入当正则求值 —— 且是 $regex 退役唯一挡路的生产者 #5710 flips that producer BEFORE any backend enrols this case」—— plugin-auth: better-auth 的 contains 被译成裸 $regex,用户输入当正则求值 —— 且是 $regex 退役唯一挡路的生产者 #5710 已关闭,三个后端已 enrol,将来时改过去时。

两处均为 note 文本(仅在失败输出中展示),不动 filter/expected/mustMention/code,已核无测试钉住这两句。普查中发现的两项超出围栏的事实已按纪律外置:having 面拒收缺 ADR-0112 信封 → 新卡 #7047;formula 面对 $icontains/$regex 静默 false 而非响亮拒收 → 评论至 #6520。

Generated by Claude Code — session_018ffcE95NaMJcL9XJ9VDYgk

---

Generated by Claude Code