Repository navigation
feat(drivers,spec)!: GroupByNode.alias is honoured by the three SQL faces (#6401) - #6849
Merged
Merged
Conversation
… faces (#6401) `GroupByNodeSchema.alias` was declared and only partially executed: the in-memory fallback projected `g.alias ?? g.field` while `SqlDriver.aggregate`, `driver-turso`'s remote transport and `driver-sqlite-wasm` (inheriting the former) all read `g.field` only. The same aggregate therefore came back keyed `closed_at` under pushdown and `qtr` under the fallback, and the choice between them is `engine.ts`'s `allStructuredSupported && !tzRequiresInMemory` — a driver capability bit and a timezone the caller cannot see. Resolved to ENFORCE under ADR-0049, chosen from a measurement rather than a preference: real non-test producers of the key number ZERO, but the capability is live on three consumers and is COMPELLED by the publish gate — `validate-react-page-props.ts` errors REACT_CHART_AXIS_UNKNOWN unless a chart's category axis is bound to `alias ?? field`. ADR-0049 removes a dangling promise and enforces a live one with a missing gate; this is the second. - driver-sql: both limbs of the structured groupBy branch project `alias ?? field`; `presentedOutput` re-keyed by the OUTPUT column, matching the aggregation branch beside it. - driver-turso REMOTE: `"field" AS "alias"`, with the alias held to `assertSafeIdentifier` like every other identifier. - driver-sqlite-wasm: inherits, covered by its own conformance suite. GROUP BY still keys on the FIELD everywhere — only the projection is renamed. `having` needed no change and now means one thing on every path. AGGREGATION_CASES gains a `groupByAlias` axis; `objectql`'s in-memory fallback is enrolled as a fourth face, answering #6409's open question 2. driver-memory already agreed and needed no alignment; driver-mongodb carries a measured DEBT row — it cannot take a structured GroupByNode at all. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01USxSgh1HUGZuh8vuFXKTGQ
|
The latest updates on your projects. Learn more about Vercel for GitHub. 1 Skipped Deployment
|
Contributor
📓 Docs Drift CheckThis PR changes 3 package(s): 116 hand-written doc(s) reference the affected code and may need an implementation-accuracy re-verification:
|
github-actions
Bot
added
documentation
…#6401) The changeset declares a breaking change (bang), so #6148's gate requires the ADR-0087 ledger question to be ANSWERED, not assumed. The answer is `not-required (no-migration-prescription)`: nothing is retired — the key keeps its declaration and starts being honoured — so there is no tombstone to write and no authored metadata a codemod could rewrite. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01USxSgh1HUGZuh8vuFXKTGQ
Contributor
|
ACCEPT — drivers lane, PM dispatch loop. This card's gate leads with a different question than an implementation card's, and it is the one that passes most clearly. 1. The measurement was done, and the leg follows from it. This was the pass/fail question: a leg chosen without the measurement would have failed the gate even with clean code. The scan was structural (balanced-bracket over every The reasoning on top of that count is the part worth keeping. Zero producers reads like "remove", and the report says so before explaining why it isn't: ADR-0049 splits on whether the feature exists, not on how many callers it has — a dangling promise is removed, a live one with a missing gate is enforced. The deciding fact is that Two things that make this more than a tidy narrative: the report noticed the gate's own error text claims rows are keyed by "the RAW FIELD NAMES", so the gate and the SQL faces contradicted each other inside one message; and it re-priced the remove leg as materially more expensive than the issue assumed — retracting 2. Verified independently against the diff, not off the report:
3. Frozen drivers handled by measurement, and the dedup is right. 4. The reverse verification is the best part of this card, because one prediction was wrong and the report says so. Direction C predicted that a harness reading 5. Conclusions gated on the JOB log. Verified from the check-runs API on head Open question ① is well placed: #6815 ( Marking ready and enabling auto-merge. The full suite runs again on the speculative merge result in the queue; per #4859 a red queue build gets diagnosed before any requeue. Generated by Claude Code |
os-zhuang
marked this pull request as ready for review
August 9, 2026 01:13
akarma-synetal
pushed a commit
to akarma-synetal/framework
that referenced
this pull request
Oct 7, 2026
…rds instead of a tracker number (stage 6) (objectstack-ai#21593) Part of objectstack-ai#20749 Clause-②: no Stage 6 of the `domain:spec` lane's share of the runtime-string burn-down (ruling `5902360492`, form D): class (c), the case notes and names of the eight shared conformance modules in `packages/spec/src`, with the two test seams that bind their text. Every rewritten note or name now states in words what the cited decision was, or drops a citation its sentence already explained. Text only. ## What changed - **58 messages / 68 tracker ids / 40 distinct cards** in eight modules: `contracts/metadata-service-roundtrip-conformance.ts` and, in `data/`, `aggregation-conformance.ts`, `filter-comparand-type-conformance.ts`, `filter-logic-conformance.ts`, `filter-text-conformance.ts`, `filter-text-operator-declared-type.ts`, `temporal-conformance.ts`, `value-roundtrip-conformance.ts`. Fifty-two are `note` / `why` texts and six are case names. - **The selector seam** (`packages/lint/src/validate-empty-combinators.test.ts:275`): selects the same four empty-combinator cases by a phrase the rewritten notes carry instead of `objectstack-ai#5322` (A3 below). - **The name-pin seam** (`packages/services/service-analytics/src/__tests__/icontains-dialect-sql.test.ts:369`): pins the renamed infix `icontains` case verbatim, the same strength (A4 below). - One `@objectstack/spec` **patch** changeset, `Clause-②: no`: the conformance tables ship in the package's `dist` (measured: the new note text is in `dist/data/index.mjs`; the two test files ship nowhere, since `@objectstack/lint` and `@objectstack/service-analytics` publish `dist` only and neither dist carries the seam text, with the package's own symbols as the positive control). - No generated artifact moves (`check:generated` green; the tables are values behind annotated types, so the API surface is unchanged). ## Census at the base (A1) Stage 3's instrument (`census.cjs`, byte-identical copy, md5 `e5fe562290fcd1ac392bd560cb86d861`) at base `cc645f2385`: class (c) is **58 messages / 68 ids in 8 files, 40 distinct cards**, exactly stage 3's count (the `[c-SELECTOR]` 4 / 7 and `[c-NAMEPIN]` 1 / 1 rows included). Under the ~60-card bar, so one stage. Lines are the base's. | file (under `packages/spec/src`) | line:id | messages / ids | field | consumer | |:--|:--|--:|:--|:--| | `contracts/metadata-service-roundtrip-conformance.ts` | 194:objectstack-ai#6725 · 202:objectstack-ai#6725 · 280:objectstack-ai#7378 · 296:objectstack-ai#7378 · 304:objectstack-ai#7378 | 5 / 5 | `why` | none prints it: both runners title a row by its `id`; the text is a reader's note in the shared table | | `data/aggregation-conformance.ts` | 342:objectstack-ai#6409 · 398:objectstack-ai#11065,objectstack-ai#11151 · 407:objectstack-ai#11065 · 417:objectstack-ai#11152 419:objectstack-ai#11249 · 535/550/561:objectstack-ai#15546 · 577/596:objectstack-ai#6401 | 9 / 11 | `note` | the assertion message in the driver-memory, driver-sql, driver-turso, driver-sqlite-wasm, objectql and mongodb-translation suites | | `data/filter-comparand-type-conformance.ts` | 153:objectstack-ai#7956 · 160:objectstack-ai#7872 · 272/282/289:objectstack-ai#19757 (names) · 178:objectstack-ai#5234 · 191:objectstack-ai#6050 (notes) | 7 / 7 | 5 `name`, 2 `note` | names are test titles in five driver suites and in `filter-save-door-face-parity.test.ts`; notes are assertion messages | | `data/filter-logic-conformance.ts` | 339/362:objectstack-ai#3774 · 420:objectstack-ai#5322 · 426:objectstack-ai#5322,objectstack-ai#5134 · 432/438:objectstack-ai#5322,objectstack-ai#5297 · 470:objectstack-ai#5298 · 476:objectstack-ai#5146 · 503/509:objectstack-ai#5298,objectstack-ai#13356 · 539:objectstack-ai#5299,objectstack-ai#5962 · 561/567/573/579/595:objectstack-ai#20444 | 16 / 22 | `note` | the assertion message in the eleven filter-logic harnesses; **the one selector**: `validate-empty-combinators.test.ts:275` picks the four `objectstack-ai#5322` notes | | `data/filter-text-conformance.ts` | 300:objectstack-ai#8934 (name) · 309:objectstack-ai#8934 · 349:objectstack-ai#6518,objectstack-ai#6682 · 428:objectstack-ai#4706 · 436:objectstack-ai#5710,objectstack-ai#6993 · 452:objectstack-ai#5240 | 6 / 8 | 1 `name`, 5 `note` | names are test titles in the text-conformance suites; **the one name pin**: `icontains-dialect-sql.test.ts:369`; notes are assertion messages | | `data/filter-text-operator-declared-type.ts` | 392:objectstack-ai#8296 · 553:objectstack-ai#8371 · 554:objectstack-ai#8296 | 3 / 3 | `note` | the assertion message in `engine-text-operator-declared-type-door.test.ts` | | `data/temporal-conformance.ts` | 211:objectstack-ai#3773 · 215/216:objectstack-ai#3777 · 220:objectstack-ai#20600 (row `why`) · 270:objectstack-ai#3777 · 393:objectstack-ai#3773 · 448/457:objectstack-ai#1874 · 549:objectstack-ai#3994 (case `note`) | 9 / 9 | 4 `why`, 5 `note` | a row `why` is seeded as DATA into each harness's `why` string column (never filtered on); a case `note` is an assertion message | | `data/value-roundtrip-conformance.ts` | 213:objectstack-ai#11535 · 225/226:objectstack-ai#11782 | 3 / 3 | `note` | printed in the TEST TITLE `round-trips NAME (NOTE)` in five driver suites | No non-test code reads any of these texts (every `.note` / `.why` read outside a test is another table's: `RETIRED_FILTER_OPERATORS`, the driver-conformance ledger). ## Delivered (A2): each site, the decision as read, the new words Each cited card was read with its body and every comment; the decision column names the record (comment id, or the landing record where the card carries none). Line = the base line of the edit anchor. "Old → new" shows only the words that changed. | file:line | ids | decision as read | old → new | |:--|:--|:--|:--| | `contracts/metadata-service-roundtrip-conformance.ts:194` | objectstack-ai#6725 | objectstack-ai#6725: card answers 404 here; landing record: changesets 1507ba3 / bbee302 (objectql and spec CHANGELOGs): MetadataFacade.register('object') wrote into a map none of its own object reads consulted | objectstack-ai#6725 would have failed → that catches an object write landing in a store none of the object reads consult — the hole a shipped implementation once fell through | | `contracts/metadata-service-roundtrip-conformance.ts:202` | objectstack-ai#6725 | objectstack-ai#6725: card answers 404 here; landing record: changesets 1507ba3 / bbee302 (objectql and spec CHANGELOGs): MetadataFacade.register('object') wrote into a map none of its own object reads consulted | produced objectstack-ai#6725 → let an object write land in the generic store while every object read looked elsewhere | | `contracts/metadata-service-roundtrip-conformance.ts:280` | objectstack-ai#7378 | objectstack-ai#7378: maintainer ruling 5261734580 (2026-08-12): row 1 refuse a name/data.name disagreement, row 2 one answer converged with check:meta-type-normalized, row 3 refuse non-object data | objectstack-ai#7378, maintainer 2026-08-12, row 1 → maintainer 2026-08-12, row 1 of the round-trip ruling | | `contracts/metadata-service-roundtrip-conformance.ts:296` | objectstack-ai#7378 | objectstack-ai#7378: maintainer ruling 5261734580 (2026-08-12): row 1 refuse a name/data.name disagreement, row 2 one answer converged with check:meta-type-normalized, row 3 refuse non-object data | objectstack-ai#7378, maintainer 2026-08-12, row 2 → maintainer 2026-08-12, row 2 of the round-trip ruling | | `contracts/metadata-service-roundtrip-conformance.ts:304` | objectstack-ai#7378 | objectstack-ai#7378: maintainer ruling 5261734580 (2026-08-12): row 1 refuse a name/data.name disagreement, row 2 one answer converged with check:meta-type-normalized, row 3 refuse non-object data | objectstack-ai#7378, maintainer 2026-08-12, row 3 → maintainer 2026-08-12, row 3 of the round-trip ruling | | `data/aggregation-conformance.ts:342` | objectstack-ai#6409 | objectstack-ai#6409: card body, executing the objectstack-ai#6188 split ruling: count_distinct stays declared and is enforced as distinct non-null values on the SQL family | objectstack-ai#6409: → count_distinct was kept and enforced on every face, as distinct non-null values: | | `data/aggregation-conformance.ts:398` | objectstack-ai#11065 objectstack-ai#11151 | objectstack-ai#11065: card answers 404 here; landing record: driver-memory changeset 2095040: avg/sum over a boolean answer the same numbers as every SQL face (a boolean is worth 1 or 0) · objectstack-ai#11151: PR objectstack-ai#12819 (triage 5446817173): driver-mongodb's boolean aggregand answers the ruled values too | objectstack-ai#11065/objectstack-ai#11151: an arithmetic accumulator that drops booleans answers its identity 0 here — a plausible number, which → A boolean aggregand is worth 1 or 0 on every face — driver-memory and then driver-mongodb were both moved onto that answer: an arithmetic accumulator that drops booleans answers its identity 0 here — a plausible number, which | | `data/aggregation-conformance.ts:407` | objectstack-ai#11065 | objectstack-ai#11065: card answers 404 here; landing record: driver-memory changeset 2095040: avg/sum over a boolean answer the same numbers as every SQL face (a boolean is worth 1 or 0) | objectstack-ai#11065: the rate-over-a-flag-column shape (an SLA-violation rate, a win rate). → The rate-over-a-flag-column shape (an SLA-violation rate, a win rate) that driver-memory answered null for until it counted a boolean as 1 or 0. | | `data/aggregation-conformance.ts:417` | objectstack-ai#11152 objectstack-ai#11249 | objectstack-ai#11152: maintainer ruling 2026-08-28, option A (changeset f6fa22c in the driver CHANGELOGs; decision request 5446817173): booleans aggregate as numbers, min/max answer 0/1 · objectstack-ai#11249: ruling 5386670755 (false/true for min/max over a boolean), superseded by objectstack-ai#11152's 2026-08-28 ruling | objectstack-ai#11152 ruling (2026-08-28): booleans aggregate as numbers with no per-aggregate exception, so the order statistics answer 0/1 in the same domain sum/avg answer in — not false/true (objectstack-ai#11249, superseded) → Ruled 2026-08-28: booleans aggregate as numbers with no per-aggregate exception, so the order statistics answer 0/1 in the same domain sum/avg answer in — not the false/true an earlier ruling had chosen, which this one superseded | | `data/aggregation-conformance.ts:535` | objectstack-ai#15546 | objectstack-ai#15546: ruling 5572006116 (option A): a non-empty group whose aggregand is NULL in every row sums to 0 on every face | objectstack-ai#15546: → Ruled: a non-empty group whose aggregand is NULL in every row sums to 0. | | `data/aggregation-conformance.ts:550` | objectstack-ai#15546 | objectstack-ai#15546: ruling 5572006116 (option A): a non-empty group whose aggregand is NULL in every row sums to 0 on every face | objectstack-ai#15546: the reachability control for the → The reachability control for the all-NULL-sums-to-0 | | `data/aggregation-conformance.ts:561` | objectstack-ai#15546 | objectstack-ai#15546: ruling 5572006116 (option A): a non-empty group whose aggregand is NULL in every row sums to 0 on every face | objectstack-ai#15546: the partial-null control → The partial-null control for the same ruling | | `data/aggregation-conformance.ts:577` | objectstack-ai#6401 | objectstack-ai#6401: dev report 5229051388 (PR objectstack-ai#6849): GroupByNode.alias is enforced, not removed; every SQL face projects the group under alias ?? field | objectstack-ai#6401: t → A structured group node's `alias` names the projected group column on every face — the SQL faces that ignored it were made to honour it. T | | `data/aggregation-conformance.ts:596` | objectstack-ai#6401 | objectstack-ai#6401: dev report 5229051388 (PR objectstack-ai#6849): GroupByNode.alias is enforced, not removed; every SQL face projects the group under alias ?? field | objectstack-ai#6401: the degenerate alias. Its twin above → The degenerate alias, under the same every-face `alias` rule. Its twin above | | `data/filter-comparand-type-conformance.ts:153` | objectstack-ai#7956 | objectstack-ai#7956: ACCEPT 5264821447: a measurement only; its control cell ({qty: {$eq: 100}}) returned the row on every driver, so the zeros were real answers | objectstack-ai#7956 control cell → control cell of the cross-driver comparand-type measurement | | `data/filter-comparand-type-conformance.ts:160` | objectstack-ai#7872 | objectstack-ai#7872: maintainer ruling 5265944890 and ACCEPT 5273103599 (PR objectstack-ai#8234): a shared comparand-type door; a safe-range bigint narrows to its exact number, beyond 2^53 is refused | (objectstack-ai#7872) → at the shared comparand door | | `data/filter-comparand-type-conformance.ts:178` | objectstack-ai#5234 | objectstack-ai#5234: ACCEPT 5217619370 (PR objectstack-ai#6296): a non-$field object member of $in/$nin and an object LIKE comparand are refused; each member is judged on its own | right (objectstack-ai#5234) → right, each judged like a scalar comparand | | `data/filter-comparand-type-conformance.ts:191` | objectstack-ai#6050 | objectstack-ai#6050: ruling B 5211349926: an undefined comparand is refused loudly; null keeps its meaning as the null predicate | (objectstack-ai#6050's untouched half) — → — the half left untouched when an undefined comparand was ruled a loud refusal — so | | `data/filter-comparand-type-conformance.ts:272` | objectstack-ai#19757 | objectstack-ai#19757: ruling 5793368540 (letter 乙): an array in the implicit-equality slot is refused at the shared face, for every driver at once | (objectstack-ai#19757) → at the shared face | | `data/filter-comparand-type-conformance.ts:282` | objectstack-ai#19757 | objectstack-ai#19757: ruling 5793368540 (letter 乙): an array in the implicit-equality slot is refused at the shared face, for every driver at once | (objectstack-ai#19757) → at the shared face | | `data/filter-comparand-type-conformance.ts:289` | objectstack-ai#19757 | objectstack-ai#19757: ruling 5793368540 (letter 乙): an array in the implicit-equality slot is refused at the shared face, for every driver at once | too (objectstack-ai#19757) → too, at the shared face | | `data/filter-logic-conformance.ts:339` | objectstack-ai#3774 | objectstack-ai#3774: defect: driver-sql OR-ed a $or branch's own keys and operators; rule: everything inside one filter object ANDs at every depth (card body) | objectstack-ai#3774: compiled → A $or combines its branches, never the keys inside one: a driver that OR-ed a branch's own keys compiled this | | `data/filter-logic-conformance.ts:362` | objectstack-ai#3774 | objectstack-ai#3774: defect: driver-sql OR-ed a $or branch's own keys and operators; rule: everything inside one filter object ANDs at every depth (card body) | objectstack-ai#3774: a single-key branch is miscompilable too — the operator map is looped with the same → A single-key branch is miscompilable too — that driver looped the operator map with the same OR | | `data/filter-logic-conformance.ts:420` | objectstack-ai#5322 | objectstack-ai#5322: maintainer ruling 5185589944: every face reduces an empty combinator to its boolean identity ($and:[] all rows, $or:[] none, {} TRUE, $not:{} none), whole tree | objectstack-ai#5322: a → Ruled: every face reduces an empty combinator to its boolean identity. A | | `data/filter-logic-conformance.ts:426` | objectstack-ai#5322 objectstack-ai#5134 | objectstack-ai#5322: maintainer ruling 5185589944: every face reduces an empty combinator to its boolean identity ($and:[] all rows, $or:[] none, {} TRUE, $not:{} none), whole tree · objectstack-ai#5134: defect repaired by PR objectstack-ai#5243 (ACCEPT 5178806144): driver-sql dropped an empty $and/$or group instead of applying the boolean identity; empty $or and empty $not now compile to FALSE | objectstack-ai#5322/objectstack-ai#5134: a disjunction of zero conditions matches nothing. Fail-closed for an RLS scope — a disjunct list that loops to zero items hides every row instead of exposing the table → Ruled: every face reduces an empty combinator to its boolean identity. A disjunction of zero conditions matches nothing. Fail-closed for an RLS scope — a disjunct list that loops to zero items hides every row instead of exposing the table, as a SQL lowering that dropped the empty group once did | | `data/filter-logic-conformance.ts:432` | objectstack-ai#5322 objectstack-ai#5297 | objectstack-ai#5322: maintainer ruling 5185589944: every face reduces an empty combinator to its boolean identity ($and:[] all rows, $or:[] none, {} TRUE, $not:{} none), whole tree · objectstack-ai#5297: dispatch 5184116664: read-scope-sql's { $not: {} } compiled to nothing (an RLS scope with no WHERE) and dropped a {} disjunct; both aligned to the boolean identity | objectstack-ai#5322: collapsing to the surviving branches instead compiles `a = x` — a silently NARROWED scope (objectstack-ai#5297) → Ruled: every face reduces an empty combinator to its boolean identity, so `{}` is a TRUE disjunct. Collapsing to the surviving branches instead compiles `a = x` — a silently NARROWED scope, the answer the RLS read-scope compiler gave until it was aligned | | `data/filter-logic-conformance.ts:438` | objectstack-ai#5322 objectstack-ai#5297 | objectstack-ai#5322: maintainer ruling 5185589944: every face reduces an empty combinator to its boolean identity ($and:[] all rows, $or:[] none, {} TRUE, $not:{} none), whole tree · objectstack-ai#5297: dispatch 5184116664: read-scope-sql's { $not: {} } compiled to nothing (an RLS scope with no WHERE) and dropped a {} disjunct; both aligned to the boolean identity | objectstack-ai#5322: emitting nothing for it runs the query UNSCOPED — on an RLS lowering that is a permission bypass (objectstack-ai#5297) → Ruled: every face reduces an empty combinator to its boolean identity, so NOT of `{}` is FALSE. Emitting nothing for it runs the query UNSCOPED — on an RLS lowering that is a permission bypass, which the read-scope compiler was until it compiled this to an always-false clause | | `data/filter-logic-conformance.ts:470` | objectstack-ai#5298 | objectstack-ai#5298: ruling 5202271174 (option A, confirmed 5204511921 item 4): $ne / $nin / $notContains are NULL-safe on the non-negated path, a no-value row is included | objectstack-ai#5298 → Ruled NULL-safe on every face | | `data/filter-logic-conformance.ts:476` | objectstack-ai#5146 | objectstack-ai#5146: maintainer ruling 5181102507: $not is NULL-safe on every driver; a row with no value does not satisfy the negated condition, so the negation returns it | objectstack-ai#5146: the same ruling reached through the combinator → The same NULL-safe ruling reached through the combinator, where it was first made for `$not` itself | | `data/filter-logic-conformance.ts:503` | objectstack-ai#5298 | objectstack-ai#5298: ruling 5202271174 (option A, confirmed 5204511921 item 4): $ne / $nin / $notContains are NULL-safe on the non-negated path, a no-value row is included | objectstack-ai#5298 option A → Ruled NULL-safe | | `data/filter-logic-conformance.ts:503` | objectstack-ai#13356 | objectstack-ai#13356: PR objectstack-ai#13356: driver-memory's reference matcher realigned so a no-value row satisfies $nin / $notContains | PR objectstack-ai#13356 → it was realigned to the ruling | | `data/filter-logic-conformance.ts:509` | objectstack-ai#5298 | objectstack-ai#5298: ruling 5202271174 (option A, confirmed 5204511921 item 4): $ne / $nin / $notContains are NULL-safe on the non-negated path, a no-value row is included | objectstack-ai#5298 option A → Ruled NULL-safe | | `data/filter-logic-conformance.ts:509` | objectstack-ai#13356 | objectstack-ai#13356: PR objectstack-ai#13356: driver-memory's reference matcher realigned so a no-value row satisfies $nin / $notContains | PR objectstack-ai#13356 → it was realigned to the ruling | | `data/filter-logic-conformance.ts:539` | objectstack-ai#5299 objectstack-ai#5962 | objectstack-ai#5299: ruling 5219858433 item 2, kept by 5238865560: $exists means has a value (non-null), never key-presence, since SQL cannot tell a missing key from null · objectstack-ai#5962: PR objectstack-ai#5962 (ACCEPT 5205011148 on objectstack-ai#5298): the NULL-safe operators and the $exists has-a-value reading shipped | objectstack-ai#5299 cell 2 / objectstack-ai#5962: `$exists` means HAS A VALUE, never key-presence → Ruled: `$exists` means HAS A VALUE, never key-presence, because SQL cannot tell a missing key from a stored null | | `data/filter-logic-conformance.ts:561` | objectstack-ai#20444 | objectstack-ai#20444: card body executing ruling A on objectstack-ai#20399 (5865693155): every compile surface answers $empty by the field's declared type; $empty: false is the exact complement | objectstack-ai#20444: null is empty on every row of the ruled table → Every face answers `$empty` by the field's declared type, and null is empty under every type's arm | | `data/filter-logic-conformance.ts:567` | objectstack-ai#20444 | objectstack-ai#20444: card body executing ruling A on objectstack-ai#20399 (5865693155): every compile surface answers $empty by the field's declared type; $empty: false is the exact complement | objectstack-ai#20444: the exact complement → `$empty: false` is the exact complement by ruling | | `data/filter-logic-conformance.ts:573` | objectstack-ai#20444 | objectstack-ai#20444: card body executing ruling A on objectstack-ai#20399 (5865693155): every compile surface answers $empty by the field's declared type; $empty: false is the exact complement | objectstack-ai#20444: `$empty` spells its NULL case out, so it → The ruled `$empty` arms spell their NULL case out, so `$empty` | | `data/filter-logic-conformance.ts:579` | objectstack-ai#20444 | objectstack-ai#20444: card body executing ruling A on objectstack-ai#20399 (5865693155): every compile surface answers $empty by the field's declared type; $empty: false is the exact complement | objectstack-ai#20444: → Under the same declared-type arms, | | `data/filter-logic-conformance.ts:595` | objectstack-ai#20444 | objectstack-ai#20444: card body executing ruling A on objectstack-ai#20399 (5865693155): every compile surface answers $empty by the field's declared type; $empty: false is the exact complement | objectstack-ai#20444: a face that lowers `$empty` → A face that lowers `$empty` to the field's declared-type arm | | `data/filter-text-conformance.ts:300` | objectstack-ai#8934 | objectstack-ai#8934: ruling A 5305708745: icontains joins the view and infix vocabularies; ilike/$ilike and icontains/$icontains are never aliased onto each other | objectstack-ai#8934 → ruled never an alias of ilike | | `data/filter-text-conformance.ts:309` | objectstack-ai#8934 | objectstack-ai#8934: ruling A 5305708745: icontains joins the view and infix vocabularies; ilike/$ilike and icontains/$icontains are never aliased onto each other | capability (objectstack-ai#8934) → capability, ruled when `icontains` joined the view and infix vocabularies | | `data/filter-text-conformance.ts:349` | objectstack-ai#6518 objectstack-ai#6682 | objectstack-ai#6518: ACCEPT 5226386725 (PR objectstack-ai#6706): the SQL family's $contains family is case-exact (GLOB on the SQLite dialects), per objectstack-ai#4706 Q2 = A · objectstack-ai#6682: ACCEPT 5251849449 and the memory half: the hardcoded case-folding flag came off driver-mongodb and driver-memory, the last two folding faces | match. SQLite's LIKE folds ASCII — the defect objectstack-ai#6518 replaced with GLOB on the SQLite dialects; a JS backend's equivalent is a RegExp carrying the `i` flag, which objectstack-ai#6682 took off → match: the `$contains` family is case-sensitive on every backend, by ruling. SQLite's LIKE folds ASCII — the defect the SQL family replaced with GLOB on the SQLite dialects; a JS backend's equivalent is a RegExp carrying the `i` flag, since taken off driver-memory and driver-mongodb, | | `data/filter-text-conformance.ts:428` | objectstack-ai#4706 | objectstack-ai#4706: ruling B 5199214776: $regex retired under ADR-0049 with a loud refusal naming the replacement, $icontains added | objectstack-ai#4706 retired the operator over → `$regex` was retired over, by ruling, with a loud refusal naming `$icontains` | | `data/filter-text-conformance.ts:436` | objectstack-ai#5710 objectstack-ai#6993 | objectstack-ai#5710: ACCEPT 5201171068 (PR objectstack-ai#5812): plugin-auth's adapter stopped emitting bare $regex for better-auth contains, unblocking the $regex retirement · objectstack-ai#6993: ACCEPT 5231388316 (PR objectstack-ai#7054): the expired status claims were re-measured by executing each face (2026-08) and rewritten | objectstack-ai#5710 flipped that producer before any backend enrolled this case (re-verified 2026-08, objectstack-ai#6993 → That producer was moved off `$regex` before any backend enrolled this case (re-verified 2026-08 by executing each face | | `data/filter-text-conformance.ts:452` | objectstack-ai#5240 | objectstack-ai#5240: maintainer ruling 5181107825: { field: {} } is refused loudly (INVALID_FILTER) on every backend, not read as TRUE or FALSE | objectstack-ai#5240 refused `{ field: {} }` over → for which a field with zero operators, `{ field: {} }`, is refused by ruling | | `data/filter-text-operator-declared-type.ts:392` | objectstack-ai#8296 | objectstack-ai#8296: standing ruling recorded in 5277439872: a where on a formula field is refused (INVALID_FIELD 400) because no driver materialises the column | objectstack-ai#8296 door refuses EVERY formula filter with INVALID_FIELD 400 whatever the `returnType` → unmaterializable-field door refuses EVERY formula filter with INVALID_FIELD 400 whatever the `returnType` (no driver stores a formula column) | | `data/filter-text-operator-declared-type.ts:553` | objectstack-ai#8371 | objectstack-ai#8371: ruling 5300373151 (option 2): a type-directed verdict on the dotted head segment; the structured/JSON head stays deliberately unjudged | unjudged there, objectstack-ai#8371 → left unjudged there, by ruling | | `data/filter-text-operator-declared-type.ts:554` | objectstack-ai#8296 | objectstack-ai#8296: standing ruling recorded in 5277439872: a where on a formula field is refused (INVALID_FIELD 400) because no driver materialises the column | objectstack-ai#8296 → the unmaterializable-field door | | `data/temporal-conformance.ts:211` | objectstack-ai#3773 | objectstack-ai#3773: defect: SQLite read an epoch-ms Field.datetime as a Julian day so date buckets were NULL; repaired by storage-aware bucketing (card body; cross-update 5099832227 on objectstack-ai#3777) | (objectstack-ai#3773) → as SQLite bucketing once did | | `data/temporal-conformance.ts:215` | objectstack-ai#3777 | objectstack-ai#3777: defect: a bare-day $lte on a datetime column stopped at midnight and dropped the rest of the final day; the bound keeps the whole day (card body; 5099832227) | by the objectstack-ai#3777 bug → when a bare-day upper bound stopped at midnight | | `data/temporal-conformance.ts:216` | objectstack-ai#3777 | objectstack-ai#3777: defect: a bare-day $lte on a datetime column stopped at midnight and dropped the rest of the final day; the bound keeps the whole day (card body; 5099832227) | by the objectstack-ai#3777 bug → when a bare-day upper bound stopped at midnight | | `data/temporal-conformance.ts:220` | objectstack-ai#20600 | objectstack-ai#20600: triage direction 5886142901, landed PR objectstack-ai#20643: the whole-day bound past 9999-12-31 is unbounded above and the drivers compile none | (objectstack-ai#20600) → and none is compiled | | `data/temporal-conformance.ts:270` | objectstack-ai#3777 | objectstack-ai#3777: defect: a bare-day $lte on a datetime column stopped at midnight and dropped the rest of the final day; the bound keeps the whole day (card body; 5099832227) | objectstack-ai#3777: the default dashboard window → The default dashboard window, whose bare-day $lte keeps the whole final day | | `data/temporal-conformance.ts:393` | objectstack-ai#3773 | objectstack-ai#3773: defect: SQLite read an epoch-ms Field.datetime as a Julian day so date buckets were NULL; repaired by storage-aware bucketing (card body; cross-update 5099832227 on objectstack-ai#3777) | objectstack-ai#3773 famil → family of the SQLite bucketing defect that read an epoch-ms datetime as a Julian da | | `data/temporal-conformance.ts:448` | objectstack-ai#1874 | objectstack-ai#1874: the templates' date-equality family that surfaced ADR-0053, whose Phase 1 stores Field.date as its calendar day (ADR-0053 Surfaced-by line; card body, no comments) | objectstack-ai#1874: `date == today` silently matched nothing while dates were stored as instants. E → `date == today` silently matched nothing while dates were stored as instants; ADR-0053 stores a date as its calendar day, so e | | `data/temporal-conformance.ts:457` | objectstack-ai#1874 | objectstack-ai#1874: the templates' date-equality family that surfaced ADR-0053, whose Phase 1 stores Field.date as its calendar day (ADR-0053 Surfaced-by line; card body, no comments) | from the objectstack-ai#1874 family → that surfaced ADR-0053 | | `data/temporal-conformance.ts:549` | objectstack-ai#3994 | objectstack-ai#3994: Field.time takes one canonical HH:MM:SS[.fff] text form on write, filter and read (ADR-0053 addendum D-C1..D-C3; driver-sql changeset 9774b78) | objectstack-ai#3994, measured → Measured before `Field.time` took one canonical `HH:MM:SS[.fff]` text form | | `data/value-roundtrip-conformance.ts:213` | objectstack-ai#11535 | objectstack-ai#11535: defect: a single-to-multi-value change kept the old text column on Postgres, so arrays came back as stringified literals; the drift detector now reports the base-type mismatch (claim 5395496220) | objectstack-ai#11535's exact shape → the exact shape a single-value column kept for a multi-value field corrupted | | `data/value-roundtrip-conformance.ts:225` | objectstack-ai#11782 | objectstack-ai#11782: landing 5406878917 (PR objectstack-ai#12019): a declared boolean answers JSON booleans on every read door and dialect; MySQL's tinyint 1/0 leaked before | objectstack-ai#11782 read this back as 1 on MySQL → MySQL read this back as 1 until every read door presented a declared boolean as true/false | | `data/value-roundtrip-conformance.ts:226` | objectstack-ai#11782 | objectstack-ai#11782: landing 5406878917 (PR objectstack-ai#12019): a declared boolean answers JSON booleans on every read door and dialect; MySQL's tinyint 1/0 leaked before | objectstack-ai#11782 → that MySQL read-back | | `lint: validate-empty-combinators.test.ts:275` | objectstack-ai#5322 | objectstack-ai#5322: maintainer ruling 5185589944: every face reduces an empty combinator to its boolean identity ($and:[] all rows, $or:[] none, {} TRUE, $not:{} none), whole tree | objectstack-ai#5322 → every face reduces an empty combinator to its boolean identity | | `analytics: icontains-dialect-sql.test.ts:369` | objectstack-ai#8934 | objectstack-ai#8934: ruling A 5305708745: icontains joins the view and infix vocabularies; ilike/$ilike and icontains/$icontains are never aliased onto each other | objectstack-ai#8934 → ruled never an alias of ilike | ## The selector seam (A3) The filter over `FILTER_LOGIC_CASES` now tests `(c.note ?? '').includes('every face reduces an empty combinator to its boolean identity')` where it tested `.includes('objectstack-ai#5322')`. The phrase is the objectstack-ai#5322 ruling in words, and the four rewritten notes open with it. Measured with an AST extractor over the table (`name` and `note` of all 36 cases, then the selector applied exactly as the test applies it): - base `cc645f2385` with `objectstack-ai#5322` → **4 of 36**: `$not of {} is FALSE — NOT TRUE`, `a {} branch is a TRUE disjunct and absorbs its $or`, `empty $and is TRUE — the AND identity`, `empty $or is FALSE — the OR identity`. - head with the phrase → **the same 4 of 36**, the same names. - cross: the old selector on the head selects 0 (so the move is required), the new phrase on the base selects 0. - control outside the set: the broader word `Ruled` selects 8, the four plus `$exists true selects exactly the valued rows`, `$ne …`, `$nin …` and `$notContains returns the rows with no value`; none of those four is picked by the phrase. - mutation control on a scratch copy (anchor hit 1, replacement present 1, anchor left 0): dropping the phrase from the empty-`$or` note leaves 3, which the test's own guard (`toEqual` on the four sorted names, `toBe(4)`) turns red. The test asserts exactly what it asserted before (the names pin, the count, and the row-set wording per case); only the selector moved. `validate-empty-combinators.test.ts` ran green: 21 of 21, the three identity cases included. ## The name pin (A4) `'icontains (the infix/view spelling, objectstack-ai#8934) lowers to $icontains — % stays a LITERAL through that door too'` becomes `'icontains (the infix/view spelling, ruled never an alias of ilike) lowers to $icontains — % stays a LITERAL through that door too'` in `FILTER_TEXT_CASES`, and the `toEqual` pin in `icontains-dialect-sql.test.ts:369` moves to the new string verbatim. No other consumer keys on that name (searched the whole tree for the old name and its fragments: the source and the pin only); the other consumers use it as a test title. The file ran green: 16 of 16. The five renamed `FILTER_COMPARAND_TYPE_CASES` names have no pin anywhere; they are test titles only. ## Text only (A5) Stage 3's `skeleton.cjs` (one line changed: the TypeScript load path to this worktree), TypeScript 6.0.3: leg 1 an AST skeleton with every string's text masked (a `+` chain's adjacent string operands read as one string), leg 2 the text of every string group, each changed group required to carry a tracker id before and none after, every other group byte-identical. Base copies vs the committed files, **10 of 10 SAME on both legs, exit 0**: | file | tokens | string groups | changed | |:--|--:|--:|--:| | `metadata-service-roundtrip-conformance.ts` | 1352 | 157 | 5 | | `aggregation-conformance.ts` | 1268 | 133 | 9 | | `filter-comparand-type-conformance.ts` | 1309 | 107 | 7 | | `filter-logic-conformance.ts` | 1682 | 223 | 16 | | `filter-text-conformance.ts` | 1093 | 137 | 6 | | `filter-text-operator-declared-type.ts` | 1821 | 114 | 3 | | `temporal-conformance.ts` | 2041 | 374 | 9 | | `value-roundtrip-conformance.ts` | 1197 | 175 | 3 | | `validate-empty-combinators.test.ts` | 2684 | 129 | 1 | | `icontains-dialect-sql.test.ts` | 4345 | 217 | 1 | Parse diagnostics 0 / 0 throughout. 58 changed groups in the eight modules, one per seam. No case's filter, input, expected rows, verdict, code, operator, dialect, order or count moves. Edits were applied by a script whose 62 anchors each had to hit exactly once before any write, and were read back from disk after it (each anchor gone, each replacement present once). Census after the edit: class (c) **0 / 0**; non-test 118 → 60 messages, 297 → 229 ids; test strings unchanged (1804 / 1920). ## Pins, titles and quotes (A6) - Pins moved: the two seams above, nothing else. Every old note and name was searched across the tree in 32-character windows: no test asserts a changed phrase; the remaining hits are code comments and other modules' own prose that share a phrase with an unchanged part of a note. - Titles that follow the text: the comparand-type and text-case names (test titles in the driver suites), and the `VALUE_ROUNDTRIP_CASES` notes (`round-trips NAME (NOTE)`). No skip list, ledger or snapshot names any of those titles. - Quotes: no `content/docs/**` page and no `skills/**` file quotes a changed note or name. - `TEMPORAL_ROWS[].why` is seeded into a `string` column (a `varchar` on Postgres and MySQL), so the four rewritten `why` texts stay short (164 characters at most). ## Gates Head `1427993cc3` (the merge of `origin/main` `901e7cf13a`; that merge touched no file under `packages/spec`, `packages/lint` or `packages/services/service-analytics`, 0 diff lines there). - `node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --commands`: 11 paths vs merge base `901e7cf13`, 153 changed lines; **88 derived commands, 88 exit 0** on `1427993cc3`, each exit code written from `$?` before any pipe; `--ran`: "✓ dispatch-gates --ran: 88 derived famil(ies) accounted for — 88 run, 0 NOT-MEASURED". - `pnpm --filter @objectstack/spec build` exit 0 (the dependency closure is empty: no workspace dependencies); `check:generated`: "✓ All 15 generated artifacts are up to date". - `pnpm --filter @objectstack/spec test`: "Test Files 606 passed (606) / Tests 17952 passed | 1 todo (17953)"; `typecheck` exit 0 ("check:test-typecheck: OK — … 52 file(s) / 246 error(s) / 135 pinned signature(s) held"). - `pnpm --filter @objectstack/lint test`: "Test Files 119 passed (119) / Tests 5620 passed (5620)"; `typecheck` exit 0. - `pnpm --filter @objectstack/service-analytics test`: "Test Files 175 passed (175) / Tests 4152 passed | 253 skipped (4405)"; `typecheck` exit 0; `icontains-dialect-sql.test.ts` alone 16 of 16. - `pnpm check:doc-authoring` (self-test, then the run): "✓ doc authoring guard: 17337 customer-facing string(s) across 1251 spec sources clean" and "sibling-package prose ids hold the baseline — 0 pinned site(s) … no growth, no burn-down unrecorded". `pnpm check:nul-bytes`: "check-nul-bytes: OK (scanned 9947 text file(s) …; no raw ASCII control bytes)". - Changeset gates: `check-changeset-no-major.mjs`, `check-empty-changeset.mjs`, `check-adr-0087-registration.mjs` (each `--base origin/main`) exit 0; with this body as the `--event` payload, see the report. - ESLint, a proven narrowing: `eslint --no-inline-config --format json` over the 10 changed TS files reads 10 files, 0 errors, 0 warnings; the population is ESLint's own (`calculateConfigForFile` returns a config for each, `isPathIgnored` false); invariance: `eslint.config.mjs` enables no type-aware linting (`parserOptions.project` / `projectService` null for all 10), so a string-text edit cannot move an untouched file's verdict. Repo-wide `pnpm lint` is CI's. - Spec `test:repo` was not run locally: none of its 51 files names a changed module or table. ## Acceptance notes - **Two cited cards answer 404** in this repository (objectstack-ai#6725 and objectstack-ai#11065; the same numbers in objectui are unrelated PRs). Their decisions were read from the shipped landing records instead: the objectql and spec CHANGELOG entries `1507ba3` / `bbee302` (`MetadataFacade.register('object')` wrote into a map none of its own object reads consulted) and the driver-memory entry `2095040` (a boolean aggregand counts as 1 or 0, as on every SQL face). Neither decision read as unclear. - Two decisions live in shipped records rather than card comments: objectstack-ai#11152's 2026-08-28 ruling (option A, booleans aggregate as numbers, superseding objectstack-ai#11249) is recorded in changeset `f6fa22c` and matches the source comment above those cases; objectstack-ai#3994's in ADR-0053's D-C addendum. - Left for later stages, untouched here: the seam files' own test titles (`(objectstack-ai#5322/objectstack-ai#5659)`, `[objectstack-ai#16028]`) and every other test string; code comments (including the section comments beside these cases); class (f) in `api/error-code-ledger.zod.ts` and `kernel/public-auth-features.ts`; `migrations/registry.ts`. - Not governed: no `.claude/**`, `skills/**`, ADR, NORTH-STAR or AGENTS.md path. ## Line budget 153 changed lines (+90 / -63) over 11 files vs merge base `901e7cf13` (dispatch-gates), under the 5000 human-merge threshold: the eight modules, the two test files (one line each) and the changeset. No generated file, no governed surface. --- _Generated by [Claude Code](https://claude.ai/code/session_01YDt3PzwfrkuFzUBF89WPmM)_ Co-authored-by: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #6401
The fact, re-verified on
e120a5aGroupByNodeSchemadeclaresalias("Alias for the projected group value",defaulting to
field). One execution path read it:alias?objectqlin-memory fallbackg.alias ?? g.fielddriver-memorynormalizeGroupBy,memory-driver.ts:1068— not in the issue's tablechartAggregateCategoryKey,alias ?? field— not in the issue's tabledriver-sqlg.fielddriver-tursolocal + remotedriver-sqlite-wasmdriver-mongodbSo one aggregate returned rows keyed
closed_atunder pushdown andqtrunderthe fallback, decided by
engine.ts'sallStructuredSupported && !tzRequiresInMemory— a driver capability bit and atimezone, neither visible to the caller.Step one was a measurement, and it chose the leg
Real non-test producers of
GroupByNode.alias: ZERO. Nothing inexamples/(showcase, CRM, todo — the showcase'sObjectChartuses the barestring form), nothing in
apps/, nothing in non-testpackages/. Writers are4 test sites plus 4 declaring surfaces (
query.zod.ts,chart.zod.ts,mcp-http-tools.ts, and the chartstrictObjectalias map that steersas/label→alias).Zero producers reads like "remove" — but ADR-0049's table does not split on
producers, it splits on whether the feature exists: a dangling promise is
removed, a live one with a missing gate is enforced.
aliasis live on threeconsumers, and decisively it is compelled by the publish gate —
validate-react-page-props.ts:491errorsREACT_CHART_AXIS_UNKNOWNunless achart's category axis is bound to
alias ?? field, and tells the author in somany words to "bind it to" the alias. A key the build gate makes you write is
not dangling. (Its error text meanwhile claims rows are "keyed by the RAW FIELD
NAMES" — the gate and the SQL faces contradicted each other.)
The zero-producer count is what makes enforcing safe, not what argues
against it: no shipped payload changes its result keys.
What changed — every face, or none
driver-sql— both limbs projectalias ?? field;?? as ??only whenthe name actually moves.
presentedOutputre-keyed by the OUTPUT column,matching the aggregation branch beside it (an aliased group value went
unpresented before).
driver-tursoREMOTE —"field" AS "alias"; the alias reaches thestatement as a quoted identifier and is therefore held to
assertSafeIdentifier, likefield.driver-sqlite-wasm— inherits, covered by its own suite.GROUP BY still keys on the FIELD on every face — pinned, because SQLite
resolves output names in
GROUP BYand a face that grouped by the alias wouldlook correct on SQLite and diverge elsewhere.
havingneeded no change:applyHavingfilters on the aggregated row's own keys, so it follows theprojection and now means one thing on every path.
Conformance
AGGREGATION_CASESgains agroupByAliasaxis + 2 cases whose VALUES are anexisting case verbatim — they can fail only on the KEY.
objectql's in-memoryfallback is enrolled as a fourth face (#6409 open question ②): it is the
face the SQL three were converged onto, and
applyInMemoryAggregationturnedout to be a pure function of rows, so no engine was needed.
Reverse verification — predicted before running
alias === fieldgreendriver-sql1/18,driver-turso3/34,sqlite-wasm1/15,objectql0driver-sql18/18 and wasm 15/15 green; turso fails onGROUP BY "bucket"c.groupByTwo results recorded rather than tidied away:
alias fails loudly. The false green needs both mistakes together, measured
directly: face reverted + harness reading the field name = 18/18 green.
driver-sqlite-wasm's suite consumesdriver-sql's builtdist, not itssrc, so direction A showed a spurious 15/15 green untildriver-sqlwasrebuilt. That suite checks its inheritance at BUILD granularity.
Frozen drivers (#5499) — measured, never flipped
driver-memory: agrees. It already returned{ field, alias: node.alias ?? node.field }— it had reached the enforceanswer independently, so no mechanical alignment was needed.
driver-mongodb: DEBT, and wider than alias.buildAggregationPipelinetypes
groupByasstring[]and buildsgroupId[field] = '$' + field, so astructured node — aliased or not — becomes the literal key
"[object Object]".The alias is unreachable rather than ignored.
mongodb-driver.ts:512passes(query as any).groupBy, which is whytscnever saw it. Home: [drivers] driver-memory / driver-mongodb disagree with the SQL family oncount_distinct— the two AGGREGATION_CASES cells #6409 left open #6814.Tests & gates
objectql2685,driver-sql1097,driver-turso906,driver-sqlite-wasm304 — all green.spec8877,lint1730,driver-memory540,driver-mongodb220 — green.lint.ymlgates: 59/59, run step by step with exit statuses readdirectly. That includes the three workspace-wide
turbo run build/turbo run typechecksteps that are notcheck:*scripts — the omissionthat took drivers: implement
count_distinctin the SQL family — enforcement half of the #6188 ruling #6409 red in CI — plusexamples/downstream-contracttypechecks,and the
check:i18n*pair run against a built workspace.Changeset included; no
content/docs/releases/edits.🤖 Generated with Claude Code
https://claude.ai/code/session_01USxSgh1HUGZuh8vuFXKTGQ
Generated by Claude Code