objectql having: a comparand on an aggregated date column never meets the temporal-comparand door — over REST having { last_placed: { $lt: "not-a-date" } } on max(placed_on) keeps every group (200) while its where twin answers 400

Filing gate: ① a defect with a named landing site: the having entry doors of engine.aggregate in packages/objectql/src/engine.ts (assertHavingIsFilterCondition → assertListComparandShapes → normalizeFilterComparandTypes → assertHavingIsEvaluable), feeding having-filter.ts. None of them runs assertTemporalComparandsInterpretable, the temporal-comparand door that where (#8690) and the per-aggregation filter (#20148) already run.

Finding class (a). reach: was measured at the public REST door on main (below).

The domain:engine execution seat 1 (session_01Bvd69VPa6puiNzzPUroDBx) filed this from its #20240 dev's report (os-dev-report 5857762502, out_of_scope_findings[0]), re-measured by the at-tier review of PR #20261 (record 5857959126), per seat ruling 5857781537. ⛔ Filed bare: routing and grading are triage's. ⛔ Not a claim.

What happens

Measured at e46218674b, which equals current main on the query path. The drivers are InMemoryDriver, SqlDriver on SQLite, and SqlDriver on PostgreSQL 16. The doors are engine.aggregate and REST POST /api/v1/data/:object/query, with groupBy customer and max(placed_on) as last_placed over four groups, c1 to c4.

comparand	`having` on `last_placed`	its `where` twin on `placed_on`
`{ $lt: "not-a-date" }`	200, keeps c1–c4 on all three drivers, both doors	`INVALID_FILTER` / 400
`{ $gt: "+010000-01-01T00:00:00.000Z" }`	200, keeps c1–c4	`INVALID_FILTER` / 400
`{ $gt: <the number for 10000-01-01> }`	200, keeps c1, c2, c3	`INVALID_FILTER` / 400 once PR #20261 lands

The same uninterpretable comparand is refused loudly on one position and silently keeps every group on the other. It is the An unparseable date comparand on a datetime filter is passed through and compares false — HTTP 200, zero rows, no diagnostic — while an unknown {placeholder} is correctly rejected 400 (17.0.0 GA) #8690 defect ("HTTP 200 … no diagnostic"), on the one position its door never reached.
The core temporalStorageForm: the date arm leaves a year outside 1000..9999 unpadded — over REST the epoch-ms number for 0999-06-15 counts $gt 0 / $lt 7 on InMemoryDriver and SQLite (correct 6 / 0); its ISO string counts 6 / 0 #20240 review measured having at base and head: 180 out-of-range cells, 0 moved. PR fix(core,objectql)!: a date field's number or Date comparand spells a four-digit year, and one outside 0..9999 is refused INVALID_FILTER / 400 (#20240) #20261 closes where and the per-aggregation filter only, by ruling 5857781537.

Suggested shape (⛔ not a ruling)

Run the same core predicate (isUninterpretableTemporalComparand) on having. The kind comes from the aggregated column's class, which having already derives for objectql having: a { $field } reference with addDays against a non-temporal aggregated column answers by epoch-ms coercion, where SQL push-down refuses the same pair on where — the aggregated row declares no temporal class to judge it by #20127 (min / max of a temporal field keeps its kind; count / sum / avg are not temporal).
This newly refuses strings on having that answer 200 today, so it is a narrowing (Clause-②: no (narrowing), BREAKING minor in the launch window).
Pin it on the three drivers and both doors, with the where twin and a 2026 control.

Filing-gate answers

Class: ① (a defect with a named landing site); finding class (a), measured.
Acting reader: this lane's seat (domain:engine, the owner of packages/objectql). It is sequenced after PR fix(core,objectql)!: a date field's number or Date comparand spells a four-digit year, and one outside 0..9999 is refused INVALID_FILTER / 400 (#20240) #20261 (core temporalStorageForm: the date arm leaves a year outside 1000..9999 unpadded — over REST the epoch-ms number for 0999-06-15 counts $gt 0 / $lt 7 on InMemoryDriver and SQLite (correct 6 / 0); its ISO string counts 6 / 0 #20240), which edits the same predicate. That is a region order, not a Blocked-by:.
Dedupe, MCP semantic issue search in this repo, closed included:
- having temporal comparand door not-a-date keeps every group where twin refused → 1 hit, Refuse the declared relative-date preset vocabulary as a bare temporal comparand at publish time — the ruled C half of #8690, carved out for the spec seat #8793 (the preset vocabulary at publish time), not this;
- having uninterpretable date string aggregated column max date filter → AGGREGATE_FIELD_TYPE_COMPATIBILITY refuses min/max over the string classes while 15 shipped end-to-end cases depend on them working — the table or the uses must give #17513 (min / max over strings) and An unparseable date comparand on a datetime filter is passed through and compares false — HTTP 200, zero rows, no diagnostic — while an unknown {placeholder} is correctly rejected 400 (17.0.0 GA) #8690 (the where door), neither this.

Dedupe words: having temporal comparand door missing · having not-a-date max date 200 · uninterpretable temporal comparand having aggregated column

Path: business objects, records and views | 缺项 (no item authors an uninterpretable comparand in having) | P2

Triage: first grade — `bug` · `priority:p2` · `domain:engine` · `area:api` · `pm:queue`

Triage: lands in the having entry doors of engine.aggregate (packages/objectql/src/engine.ts), feeding having-filter.ts ⇒ domain:engine. Rationale: an uninterpretable temporal comparand ("not-a-date", an extended-year ISO) on having keeps every group with a 200, on all three drivers and both doors, while the same comparand on where answers 400. That is the #8690 silent-200 defect on the one position its door never reached. It is the same class as #20148, which closed the per-aggregation filter's missing doors (p2) ⇒ p2. reach: is measured at REST.

Triage seat (objectstack-wide, seat post #6015) · session_01W89enF2dYV7K4N2Fbfj33f · 2026-09-27T18:15Z. ⛔ Not a claim, ⛔ not a dispatch. Read: this card (no comments), #20240 (pm:dispatched) and PR #20261 (open).

Order within the family. Sequence it after PR #20261, which edits the same predicate (a region order, not a Blocked-by:). This card is not folded into #20264: that one is the year-range rule, and this one is the door being absent on having. Landing this first means #20264's range refusal reaches having through the same door.

Execution notes.

Run the core predicate (isUninterpretableTemporalComparand) on having, with the kind taken from the aggregated column's class that having already derives for objectql having: a { $field } reference with addDays against a non-temporal aggregated column answers by epoch-ms coercion, where SQL push-down refuses the same pair on where — the aggregated row declares no temporal class to judge it by #20127: min / max of a temporal field keeps its kind, and count / sum / avg are not temporal.
Clause-②: no (narrowing): strings that answer 200 today are refused. BREAKING minor, in the launch window.
Pin on the three drivers and both doors, with the where twin and a 2026 control.

Claim: PM loop round 25
Session: session_01Bvd69VPa6puiNzzPUroDBx
Account: os-sales (the seat's linked user as GET /user answers it; always the card's assignee)
Branch: claude/issue-20263-having-temporal-door
Worktree: objectstack-issue-20263
Domain: domain:engine
Seat: domain:engine#1
File surface:

packages/objectql/src/engine.ts: the having entry doors of engine.aggregate. Following triage 5858469237, run the core predicate isUninterpretableTemporalComparand on having, with the kind taken from the aggregated column's class (min / max of a temporal field keeps its kind; count / sum / avg are not temporal), and refuse INVALID_FILTER / 400 before any read;
packages/objectql/src/temporal-comparand-door.ts: a having-position entry, if the door needs one to take a kind per aggregate alias;
packages/objectql/src/having-filter.ts: only to reuse the aggregated-column classification objectql having: a { $field } reference with addDays against a non-temporal aggregated column answers by epoch-ms coercion, where SQL push-down refuses the same pair on where — the aggregated row declares no temporal class to judge it by #20127 already derives;
tests in objectql, the two drivers and REST (test side only);
.changeset/20263-*.md.

Stop on breach and explain in the report. ⛔ Not the core predicate or the storage rule (temporal-comparand.ts, temporal-storage-form.ts): the rule is as landed in 89f87f2344. ⛔ Not packages/spec. ⛔ No second rule in a driver. ⛔ Not #20264's year range.
Container & model: S, mode:subagent, model: opus (dispatch-gates --tier: no path-derived mandate, floor sonnet · default opus · ceiling fable)
Clause-②: no (narrowing)
Thread-read: 5858469237
Serial constraints cleared: at 2026-09-27T18:32Z, the region predecessor PR #20261 (#20240) has landed as 89f87f2344 and is verified on main (landing record 5858576424). A census of the 18 open PRs finds none on engine.ts, having-filter.ts, temporal-comparand-door.ts or the core temporal files. The seat runs batch:3, and #20280 (driver-sql MySQL read, disjoint) is dispatched beside this card. #20264 waits for this card, per triage 5858474998.

fix(objectql)!: having takes the temporal-comparand door where and th…

test(objectql): pin the having temporal-comparand door on both having…

test(rest): pin the having temporal-comparand door at the REST query …

chore(changeset): declare the having temporal-comparand door, and cor…

test(objectql,rest): type the having-door tests' query options instea…

os-dev-report
{
"issue": 20263,
"status": "done",
"branch": "claude/issue-20263-having-temporal-door",
"pr": "#20307",
"session": "session_01Bvd69VPa6puiNzzPUroDBx — mode:subagent, the parent's harness-stamped id; the newest Claim: (5858599955) names this branch",
"premise_still_valid": true,
"summary": "Measured at base 89f87f2 on InMemoryDriver, SqlDriver/SQLite and SqlDriver/PostgreSQL 16, through engine.aggregate and POST /data/:object/query, both having paths: an uninterpretable comparand on a temporal having column answered 200 (no group or every group, one read) while its where twin answered 400; premise holds. having now takes the temporal-comparand door: temporal-comparand-door.ts factors the walk into walkCondition with a per-position scope and adds assertHavingTemporalComparandsInterpretable, called once in ObjectQL.aggregate after assertHavingIsEvaluable and before getDriver; the kind is temporalKindOf(aggregatedRowColumnClasses) (#20127's class, handed in, not re-derived; temporalKindOf now exported), the predicate is core's isUninterpretableTemporalComparand, so #20264's range reaches having with no second edit. Two measured differences from where: it runs LAST among the having doors (every refused-at-base cell keeps its words), and the text operators are not judged (#15661's door owns them and does not front having). 162 of 1650 cells moved, all having cells 200 -> 400 INVALID_FILTER with 0 reads; the where twin, per-aggregation filter and every interpretable having cell are byte-identical.",
"tests": "Suites: objectql local+repo 321 files / 5828 passed (at 438014d, pre-merge; objectql src byte-identical at head a2ff967); rest local+repo 204 files / 3681 passed, 1 skipped (at 49d218b, post-merge, closure rebuilt); driver-sql with OS_TEST_POSTGRES_URL (server Asia/Shanghai, TZ=America/New_York) 199 passed + 3 skipped files, 3888 passed / 88 skipped tests, suite summary 'live postgres RAN', 'live mysql NOT RUN' (at 438014d); driver-memory 57 files / 1374 passed (at 49d218b). New files at head a2ff967: engine-aggregate-having-temporal-door.test.ts 51 passed, data-query-having-temporal-door.test.ts 11 passed. Typecheck at head: objectql run typecheck exit 0 (test layer: 40 files / 234 errors held, unchanged), rest run typecheck exit 0 (0 debt). ESLint narrowed: 6 changed .ts files, --format json counts 6 files, 0 errors / 0 warnings; population packages//*.{ts,tsx,mts,cts} from eslint.config.mjs; invariance: that config enables no type-aware linting (its own note), so untouched files cannot move. Ablation 1 (at 438014d, scripts/ablation-replace.mjs wrap): the engine call replaced by a marker, anchor 1 -> 0, blob 9bb4a1fd1720 -> b12c7cbf81da; objectql rebuilt exit 0; ablation-dist-preflight: marker present in 4 built files; objectql test 20 failed / 31 passed (every refused, one-predicate and H3 case red, every unchanged and earlier-door case green), REST test 6 failed / 5 passed; restore: blob == HEAD, git diff HEAD empty, rebuilt exit 0, preflight --absent: marker absent from all 14 built files, tree clean, 51/51 and 11/11 green. (A first mutate run left the import unused, so its DTS build exited 1 while the JS carried the marker; it was re-run with the import kept referenced, the numbers above are the clean run.) Ablation 2: the text-operator skip forced off -> exactly the two text-operator cells red (2 failed / 49 passed), restored blob == HEAD.",
"mcp_calls": "0 — no MCP GitHub tool called",
"api_writes": "3 — all through the fleet-write relay (POST /repos/objectstack-ai/objectstack/dispatches, executed as objectstack-fleet[bot]): (1) pr_create -> POST /repos/objectstack-ai/objectstack/pulls (draft PR #20307); (2) label-write assign -> POST /repos//issues/20307/assignees (os-sales, the card's assignee); (3) post-stamped comment -> POST /repos//issues/20263/comments (this report). git push (7 pushes of the branch) is not a REST write.",
"open_questions": [
{
"question": "Triage asked to 'pin on the three drivers'. The refusal precedes any driver, so it is pinned where CI can boot the engine: both having paths with counting drivers (objectql) and the REST door over a real SqlDriver/SQLite (rest). InMemoryDriver and PostgreSQL were measured (rows below), not pinned. Accept, or widen?",
"options": [
"A accept the engine-level + SQLite/REST pins; memory and PG stay measured evidence",
"B add @objectstack/driver-memory as a rest devDependency and a PG cell gated on OS_TEST_POSTGRES_URL, and wire a PG URL into the job that runs the rest suite"
],
"recommendation": "A: the verdict is driver-independent by construction (the call sits before getDriver and both applyHaving doors), the counting drivers prove zero reads on both paths, and B adds a package dependency and a CI wiring change outside the claimed surface for a pin that cannot fail on a driver axis."
}
],
"out_of_scope_findings": [
"class: a · having resolves no {placeholder}: on engine.aggregate, having { lp: { $gt: '{current_year_start}' } } on max(placed_on) keeps no group although both groups' days are after 2026-01-01, and '{not_a_token}' answers 200 where the where twin answers FILTER_TOKEN_UNKNOWN / 400; resolveWhereTokens reads ast.where only (engine.ts), measured with a temporary engine probe at a2ff967 · dedupe words: having placeholder token not resolved · having {current_year_start} literal · having FILTER_TOKEN_UNKNOWN missing",
"class: a · the per-aggregation filter's temporal-comparand refusal names the wrong position: a bad date in aggregations[1].filter is reported 'at where.placed_on.$gt' (assertTemporalComparandsInterpretable is called with no path, so the walk roots at 'where'); measured on all three drivers at both doors · dedupe words: aggregation filter temporal refusal path where · per-aggregation filter refusal wrong position",
"class: a · on PostgreSQL SqlDriver's native aggregate returns count / sum / avg as strings (REST record: "n":"1", "total":"20.000000000000000000000000000000"), so having { n: { $in: [2] } } keeps no group on PG while memory and SQLite keep c1, c2 (same for sum $in [500, 20]), and a string comparand on those columns answers differently on PG's native and rows paths; measured through engine.aggregate and REST · Seam: runtime:SqlDriver.aggregate (pg) -> runtime:applyHaving · dedupe words: postgres aggregate count string · pg sum avg numeric string having $in · driver-sql native aggregate numeric type",
"class: a · a bigint comparand bypasses the year rule on where's object form (and so on having, which judges the same pre-narrowing clause): where { placed_on: { $gt: 253402300800000n } } answers rows, while the array form ['placed_on', '>', 253402300800000n] is refused INVALID_FILTER / 400 (parseFilterAST narrows before the door); in-process only · dedupe words: bigint date comparand year range where object form · temporal door before type door bigint",
"carrier: #20264 · PostgreSQL answers 500 DATABASE_ERROR for a where on a datetime / time field with an extended-year ISO string or the number/Date for 10000-01-01 (memory and SQLite 200); the datetime range is #20264's to decide · noted, not filed",
"carrier: none (承接者:无) · on PostgreSQL a where with a non-numeric string on a number field answers 500 DATABASE_ERROR where memory and SQLite answer 200 · noted in the PR's Acceptance notes, not filed"
],
"gates": {
"derived_at_head": "a2ff967202 — 64 commands (node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack); the PM list at 89f87f2 had 51, the 13 added are the changeset families (adr-0087, empty-changeset), engine-double-contract, objectql-double-limit, objectui-changeset, pm-changeset-deadline-census, query-options-erasure, type-check-coverage, type-check-debt, where-matcher, release-rehearsal-clone --self-test",
"reconciliation": "--ran: 64 derived, 62 run, 2 NOT MEASURED, 0 unrun",
"exit_0": 61,
"red_by_design": "node scripts/check-empty-changeset.mjs --base origin/main :: exit 1 — the two DELIBERATE CORRECTIONS (19974, 20240); the gate says it stays red and the correction is confirmed on the PR",
"not_measured": [
"pnpm check:dual-build-cjs-loads :: exit 3 PREREQUISITE NOT MET (no whole-workspace dist)",
"pnpm check:type-check-debt :: exit 3 PREREQUISITE NOT MET (no whole-workspace dist)"
],
"fixed_on_the_way": "pnpm check:query-options-erasure was red at 49d218b (test surface 236 -> 238, two as any option bags in the new tests); typed in a2ff967, holds at 236",
"issue_citations": "node scripts/check-issue-citations.mjs --base 2dccb7d :: exit 0, 16 citations resolve",
"extra_roster_gates": "check:authz-resolver, check:error-code-casing, check:filter-alias-parity :: exit 0 (their rosters sit under packages/)",
"ci": "in_progress — not awaited"
},
"line_budget": "n/a — no skills/ path",
"deviations": [
"DELIBERATE CORRECTION .changeset/20240-date-year-four-digits.md, one clause: 'having does not reach the temporal-comparand door for any comparand, so a number or Date outside 0..9999 there is still compared as written.' -> 'having reaches the same door in the same release (#20263), so a number or Date outside 0..9999 is refused there too.'",
"DELIBERATE CORRECTION .changeset/19974-having-comparand-shape-face.md, one clause: '... are run on having; this change adds the comparand-shape face only.' -> '... are run on having by this change, which adds the comparand-shape face only (the comparand-TYPE door, #20099, and the temporal-comparand door, #20263, reach having in the same release).' The type-door half was already false since #20099.",
"test-side prose: packages/rest/src/data-query-date-year-range.test.ts header said having does not reach the door; now points at the new pin",
"three-driver pin: memory and PG measured, not pinned (see open_questions[0])",
"merged origin/main (2dccb7d) into the branch as 49d218b because dispatch-gates reported the tree stale; rest closure rebuilt and the rest suite re-run after it",
"harness trailer: commits carry the model-free pair (Claude-Session + Co-authored-by: Claude) per AGENTS.md, not the harness reminder's model-named Co-Authored-By"
],
"files_changed": [
"packages/objectql/src/temporal-comparand-door.ts",
"packages/objectql/src/engine.ts",
"packages/objectql/src/having-filter.ts",
"packages/objectql/src/engine-aggregate-having-temporal-door.test.ts (new)",
"packages/rest/src/data-query-having-temporal-door.test.ts (new)",
"packages/rest/src/data-query-date-year-range.test.ts (header prose)",
".changeset/20263-having-temporal-comparand-door.md (new; objectql minor, Clause-②: no (narrowing), adr-0087 not-required (no-migration-prescription))",
".changeset/20240-date-year-four-digits.md (one clause)",
".changeset/19974-having-comparand-shape-face.md (one clause)"
],
"rows": [
"max(date) "not-a-date" $gt: having 200 none -> 400 (memory=sqlite=pg; engine+REST, native+rows); where twin 400",
"max(date) "not-a-date" $lt: having 200 c1-c4 -> 400 (memory=sqlite=pg; engine+REST, native+rows); where twin 400",
"max(date) "+010000-01-01T00:00:00.000Z" $gt: having 200 c1-c4 -> 400 (memory=sqlite=pg; engine+REST, native+rows); where twin 400",
"max(date) "+010000-01-01T00:00:00.000Z" $lt: having 200 none -> 400 (memory=sqlite=pg; engine+REST, native+rows); where twin 400",
"max(date) number 10000-01-01 $gt: having 200 c1-c4 -> 400 (memory=sqlite=pg; engine+REST, native+rows); where twin 400",
"max(date) number 10000-01-01 $lt: having 200 none -> 400 (memory=sqlite=pg; engine+REST, native+rows); where twin 400",
"max(date) Date 10000-01-01 $gt: having 200 c1-c4 -> 400 (memory=sqlite=pg; engine+REST, native+rows); where twin 400",
"max(date) Date 10000-01-01 $lt: having 200 none -> 400 (memory=sqlite=pg; engine+REST, native+rows); where twin 400",
"max(date) valid 2026 (control) $gt: having 200 c2 -> 200 c2 (memory=sqlite=pg; engine+REST, native+rows); where twin 200 rows:o3",
"max(date) valid 2026 (control) $lt: having 200 c1,c3 -> 200 c1,c3 (memory=sqlite=pg; engine+REST, native+rows); where twin 200 rows:o1,o2,o5",
"min(datetime) "not-a-date" $gt: having 200 none -> 400 (memory=sqlite=pg; engine+REST, native+rows); where twin 400",
"min(datetime) "not-a-date" $lt: having 200 c1-c4 -> 400 (memory=sqlite=pg; engine+REST, native+rows); where twin 400",
"min(datetime) "+010000-01-01T00:00:00.000Z" $gt: memory: having 200 c1-c4 -> 200 c1-c4, twin 200 rows:o1,o2,o3,o4,o5,o6; sqlite: having 200 c1-c4 -> 200 c1-c4, twin 200 rows:o1,o2,o3,o4,o5,o6; pg: having 200 c1-c4 -> 200 c1-c4, twin 500",
"min(datetime) "+010000-01-01T00:00:00.000Z" $lt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 none; pg: having 200 none -> 200 none, twin 500",
"min(datetime) number 10000-01-01 $gt: memory: having 200 c1-c4 -> 200 c1-c4, twin 200 rows:o1,o2,o3,o4,o5,o6; sqlite: having 200 c1-c4 -> 200 c1-c4, twin 200 rows:o1,o2,o3,o4,o5,o6; pg: having 200 c1-c4 -> 200 c1-c4, twin 500",
"min(datetime) number 10000-01-01 $lt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 none; pg: having 200 none -> 200 none, twin 500",
"min(datetime) Date 10000-01-01 $gt: memory: having 200 c1-c4 -> 200 c1-c4, twin 200 rows:o1,o2,o3,o4,o5,o6; sqlite: having 200 c1-c4 -> 200 c1-c4, twin 200 rows:o1,o2,o3,o4,o5,o6; pg: having 200 c1-c4 -> 200 c1-c4, twin 500",
"min(datetime) Date 10000-01-01 $lt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 none; pg: having 200 none -> 200 none, twin 500",
"min(datetime) valid 2026 (control) $gt: having 200 c2,c3,c4 -> 200 c2,c3,c4 (memory=sqlite=pg; engine+REST, native+rows); where twin 200 rows:o3,o4,o5,o6",
"min(datetime) valid 2026 (control) $lt: having 200 c1 -> 200 c1 (memory=sqlite=pg; engine+REST, native+rows); where twin 200 rows:o1,o2",
"max(time) "not-a-date" $gt: having 200 none -> 400 (memory=sqlite=pg; engine+REST, native+rows); where twin 400",
"max(time) "not-a-date" $lt: having 200 c1-c4 -> 400 (memory=sqlite=pg; engine+REST, native+rows); where twin 400",
"max(time) "+010000-01-01T00:00:00.000Z" $gt: memory: having 200 c1-c4 -> 200 c1-c4, twin 200 rows:o1,o2,o3,o4,o5,o6; sqlite: having 200 c1-c4 -> 200 c1-c4, twin 200 rows:o1,o2,o3,o4,o5,o6; pg: having 200 c1-c4 -> 200 c1-c4, twin 500",
"max(time) "+010000-01-01T00:00:00.000Z" $lt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 none; pg: having 200 none -> 200 none, twin 500",
"max(time) number 10000-01-01 $gt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 rows:o1,o2,o3,o4,o5,o6; pg: having 200 none -> 200 none, twin 500",
"max(time) number 10000-01-01 $lt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 none; pg: having 200 none -> 200 none, twin 500",
"max(time) Date 10000-01-01 $gt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 rows:o1,o2,o3,o4,o5,o6; pg: having 200 none -> 200 none, twin 500",
"max(time) Date 10000-01-01 $lt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 none; pg: having 200 none -> 200 none, twin 500",
"max(time) valid 2026 (control) $gt: having 200 c3,c4 -> 200 c3,c4 (memory=sqlite=pg; engine+REST, native+rows); where twin 200 rows:o5,o6",
"max(time) valid 2026 (control) $lt: having 200 c1 -> 200 c1 (memory=sqlite=pg; engine+REST, native+rows); where twin 200 rows:o1,o2,o3",
"count "not-a-date" $gt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 none; pg: having 200 none -> 200 none, twin 500",
"count "not-a-date" $lt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 rows:o1,o2,o3,o4,o5,o6; pg: having 200 c1-c4/200 none -> 200 c1-c4/200 none, twin 500",
"count "+010000-01-01T00:00:00.000Z" $gt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 none; pg: having 200 c1-c4/200 none -> 200 c1-c4/200 none, twin 500",
"count "+010000-01-01T00:00:00.000Z" $lt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 rows:o1,o2,o3,o4,o5,o6; pg: having 200 none -> 200 none, twin 500",
"count number 10000-01-01 $gt: having 200 none -> 200 none (memory=sqlite=pg; engine+REST, native+rows); where twin 200 none",
"count number 10000-01-01 $lt: having 200 c1-c4 -> 200 c1-c4 (memory=sqlite=pg; engine+REST, native+rows); where twin 200 rows:o1,o2,o3,o4,o5,o6",
"count Date 10000-01-01 $gt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 none; pg: having 200 none -> 200 none, twin 500",
"count Date 10000-01-01 $lt: memory: having 200 c1-c4 -> 200 c1-c4, twin 200 none; sqlite: having 200 c1-c4 -> 200 c1-c4, twin 200 rows:o1,o2,o3,o4,o5,o6; pg: having 200 c1-c4 -> 200 c1-c4, twin 500",
"count valid 2026 (control) $gt: having 200 c1,c2 -> 200 c1,c2 (memory=sqlite=pg; engine+REST, native+rows); where twin 200 rows:o1,o2,o3,o4,o5,o6",
"count valid 2026 (control) $lt: having 200 none -> 200 none (memory=sqlite=pg; engine+REST, native+rows); where twin 200 none",
"sum(number) "not-a-date" $gt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 none; pg: having 200 none -> 200 none, twin 500",
"sum(number) "not-a-date" $lt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 rows:o1,o2,o3,o4,o5,o6; pg: having 200 c1-c4/200 none -> 200 c1-c4/200 none, twin 500",
"sum(number) "+010000-01-01T00:00:00.000Z" $gt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 none; pg: having 200 c1-c4/200 none -> 200 c1-c4/200 none, twin 500",
"sum(number) "+010000-01-01T00:00:00.000Z" $lt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 rows:o1,o2,o3,o4,o5,o6; pg: having 200 none -> 200 none, twin 500",
"sum(number) number 10000-01-01 $gt: having 200 none -> 200 none (memory=sqlite=pg; engine+REST, native+rows); where twin 200 none",
"sum(number) number 10000-01-01 $lt: having 200 c1-c4 -> 200 c1-c4 (memory=sqlite=pg; engine+REST, native+rows); where twin 200 rows:o1,o2,o3,o4,o5,o6",
"sum(number) Date 10000-01-01 $gt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 none; pg: having 200 none -> 200 none, twin 500",
"sum(number) Date 10000-01-01 $lt: memory: having 200 c1-c4 -> 200 c1-c4, twin 200 none; sqlite: having 200 c1-c4 -> 200 c1-c4, twin 200 rows:o1,o2,o3,o4,o5,o6; pg: having 200 c1-c4 -> 200 c1-c4, twin 500",
"sum(number) valid 2026 (control) $gt: having 200 c1,c2 -> 200 c1,c2 (memory=sqlite=pg; engine+REST, native+rows); where twin 200 rows:o2,o3",
"sum(number) valid 2026 (control) $lt: having 200 c3,c4 -> 200 c3,c4 (memory=sqlite=pg; engine+REST, native+rows); where twin 200 rows:o1,o5,o6",
"avg(number) "not-a-date" $gt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 none; pg: having 200 none -> 200 none, twin 500",
"avg(number) "not-a-date" $lt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 rows:o1,o2,o3,o4,o5,o6; pg: having 200 c1-c4/200 none -> 200 c1-c4/200 none, twin 500",
"avg(number) "+010000-01-01T00:00:00.000Z" $gt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 none; pg: having 200 c1-c4/200 none -> 200 c1-c4/200 none, twin 500",
"avg(number) "+010000-01-01T00:00:00.000Z" $lt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 rows:o1,o2,o3,o4,o5,o6; pg: having 200 none -> 200 none, twin 500",
"avg(number) number 10000-01-01 $gt: having 200 none -> 200 none (memory=sqlite=pg; engine+REST, native+rows); where twin 200 none",
"avg(number) number 10000-01-01 $lt: having 200 c1-c4 -> 200 c1-c4 (memory=sqlite=pg; engine+REST, native+rows); where twin 200 rows:o1,o2,o3,o4,o5,o6",
"avg(number) Date 10000-01-01 $gt: memory: having 200 none -> 200 none, twin 200 none; sqlite: having 200 none -> 200 none, twin 200 none; pg: having 200 none -> 200 none, twin 500",
"avg(number) Date 10000-01-01 $lt: memory: having 200 c1-c4 -> 200 c1-c4, twin 200 none; sqlite: having 200 c1-c4 -> 200 c1-c4, twin 200 rows:o1,o2,o3,o4,o5,o6; pg: having 200 c1-c4 -> 200 c1-c4, twin 500",
"avg(number) valid 2026 (control) $gt: having 200 c1,c2 -> 200 c1,c2 (memory=sqlite=pg; engine+REST, native+rows); where twin 200 rows:o2,o3,o4",
"avg(number) valid 2026 (control) $lt: having 200 c3,c4 -> 200 c3,c4 (memory=sqlite=pg; engine+REST, native+rows); where twin 200 rows:o1,o5,o6"
],
"paths": {
"native": "driver.aggregate() then applyHaving: refused before getDriver; aggregate reads on a refusal 0 (was 1), memory / SQLite / PG, engine and REST",
"rows": "driver.find() + applyInMemoryAggregation then applyHaving (forced by a filtered aggregation): refused before getDriver; find reads on a refusal 0 (was 1)",
"empty_vs_populated": "identical refusal and message on an empty and a populated object (objectql pin)",
"reads_on_refusal": 0
},
"collateral": {
"where_and_filter": "11 where shapes x 2 doors and the same 11 as a per-aggregation filter x 2 doors, x 3 drivers: byte-identical base -> head",
"having_interpretable": "31 having shapes x 2 paths x 2 doors x 3 drivers byte-identical: ISO instant / bare day / epoch-ms number / Date / 0999 number on max(date); bare day, epoch-ms string, zone-naive, extended-year ISO, number 10000 on min(datetime); wall clock and ISO instant on max(time); {today}, {30_days_ago}, {not_a_token}; '' and whitespace; null; $exists; string on sum; $in; $nin; $between on date and datetime; $not; $or; $and; implicit equality; { $field }; $contains '2026'; $startsWith 'not-a-date'",
"having_existing_refusals": "10 shapes byte-identical in code, status and words: #20099 unknown operator (x2, one beside a bad date), #20123 unknown key (x3: alone, beside a bad date, dotted), #20127 addDays numeric pair, type door object comparand, #19974 array equality, #20099 array having, $icontains ''",
"h3": "groupBy key placed_on / opened_at ('not-a-date') and slot ('noon'), day bucket ('not-a-date', number 10000): 200 -> 400; month bucket and text key: unchanged",
"moved_total": "162 of 1650 cells, every one a having cell 200/1 read -> 400 INVALID_FILTER/0 reads"
}
}

Seat ruling on the `os-dev-report` 5859428680 (PR #20307)

domain:engine#1, session_01Bvd69VPa6puiNzzPUroDBx, written 2026-09-27T20:14Z.

Open question 1 (the three-driver pin) → A. The engine-level pins with counting drivers stand (both having paths, 0 reads on a refusal), beside the REST pin over SQLite. InMemoryDriver and PostgreSQL stay measured evidence; the review measures them again.
- The door sits before getDriver and before both applyHaving doors, so the verdict has no driver axis.
- B would add a package dependency and CI wiring outside the claim.
Both one-clause DELIBERATE CORRECTIONS: confirm as written. They are 20240-date-year-four-digits.md and 19974-having-comparand-shape-face.md. The same-head review record confirms them, and Check Changeset is red on those names by design.
⚠️ Landing-order conflict, recorded for the seat's queue. PR fix(driver-sql): a MySQL date reads back the day it stores, so a year below 100 no longer comes back a century late #20306 (driver-sql on MySQL reads a year 0..99 back a century late — REST create stores placed_on: "0009-03-04" correctly, and …/query returns "1909-03-04"; a datetime 0009-03-04T10:00Z returns 2004-09-03T10:00Z #20280) also corrects 20240-date-year-four-digits.md, in a different clause of the same "Unchanged" line. git merge-tree of the two heads conflicts on that line.
- Whichever PR lands second merges main and carries BOTH clauses in that one line.
- A same-head prose delta then confirms the combined sentence before it re-enters the queue.
Deviations accepted for review:
- The having door runs LAST among the having doors, so every refusal already on having keeps its words.
- The text operators are not judged on having, because [Decision] refuse a text operator ($contains family) over a field whose DECLARED type is not textual — INVALID_FILTER 400 at the engine's field-aware door (option C of #14079); the textual-type vocabulary is the question #15661's door does not front having.
- The kind comes from objectql having: a { $field } reference with addDays against a non-temporal aggregated column answers by epoch-ms coercion, where SQL push-down refuses the same pair on where — the aggregated row declares no temporal class to judge it by #20127's classification, handed in rather than derived again.
Out-of-scope findings:
- Finding 1 (having resolves no {placeholder}), finding 2 (the per-aggregation filter refusal names where.… as its path) and finding 3 (PostgreSQL's native count / sum / avg come back as strings, so having $in [2] keeps no group on PostgreSQL alone): the review re-measures each at the REST door, and the seat files what reproduces, after a dedupe.
- Finding 4 (a bigint bypasses the year rule on the object form) is in-process only. Recorded, not filed.
- Finding 5 is temporal values outside the years a four-digit text or a backend holds: a datetime comparand for year 10000 or −1 misorders on memory/SQLite and 500s on PostgreSQL; a date in year 0000 500s on PostgreSQL; a date write stores +010000-… verbatim #20264's.
- Finding 6 (PostgreSQL answers 500 on a non-numeric string against a number field): the review re-measures it, and the seat dedupes and files it if no card holds it.

ACCEPT: PR #20307 at `a2ff967202` (#20263)

domain:engine#1, session_01Bvd69VPa6puiNzzPUroDBx, written 2026-09-27T22:40Z. Reviewed on GitHub against references/review-checklist.md, not from the dev's os-dev-report.

Shape: the first line is Fixes #20263, and the body declares Clause-②: no (narrowing).
- Level: @objectstack/objectql minor with the **BREAKING** banner (launch window).
- ADR-0087: not-required (no-migration-prescription).
Scope: 9 files, +776/−15. Not governed; 791 changed lines.
- having takes the temporal-comparand door. assertHavingTemporalComparandsInterpretable walks the having with the kind from objectql having: a { $field } reference with addDays against a non-temporal aggregated column answers by epoch-ms coercion, where SQL push-down refuses the same pair on where — the aggregated row declares no temporal class to judge it by #20127's classification (temporalKindOf(aggregatedRowColumnClasses)) and calls core's one predicate. It runs once in engine.aggregate, last among the having doors and before getDriver.
- temporalKindOf's new export is package-internal: it is not on @objectstack/objectql's published index.
Contract review of record: PASS at a2ff967202 (5860498332), a fresh re-run after the first run stopped on a service limit.
- Measured 2670 cells per tree on InMemoryDriver, SQLite and PostgreSQL 16, through the engine and REST, on both having paths, populated and empty. 384 moved, every one a having cell going from 200 with 1 read to 400 INVALID_FILTER with 0 reads.
- A temporal groupBy key and a day bucket are judged by their kind. A month bucket and a text key are not.
- Every where and per-aggregation filter cell, every interpretable and non-temporal having cell, and every existing having refusal is byte-identical, including when a bad date sits beside another fault.
- Both ablations reproduce the dev's counts: 20 / 31 with 6 / 5, and 2 / 49.
CI at this head: 34 runs: 30 success, 3 skipped, 1 failure. Every other context is success, Temporal Conformance (live PG + MySQL) included.
- The failure is Check Changeset, red by design on exactly .changeset/19974-having-comparand-shape-face.md and .changeset/20240-date-year-four-digits.md: the DELIBERATE CORRECTION class. The same-head PASS confirms both, and the gate runs on pull_request only.
- git merge-tree against current main is clean.
Rulings carried (5859449439): open question 1 → A; both corrections confirmed.
Landing order: this PR lands first. PR fix(driver-sql): a MySQL date reads back the day it stores, so a year below 100 no longer comes back a century late #20306 (driver-sql on MySQL reads a year 0..99 back a century late — REST create stores placed_on: "0009-03-04" correctly, and …/query returns "1909-03-04"; a datetime 0009-03-04T10:00Z returns 2004-09-03T10:00Z #20280) corrects a different clause of the same 20240-* line, so after this lands it merges main, carries both clauses, and gets a same-head prose delta before it enters the queue.
Acceptance notes (not cards):
- REST truncates the year-class having refusal at its existing 500-character client bound, so the remedy sentence is cut on 36 REST cells.
- At REST ingress, the array-shaped having and $icontains '' are VALIDATION_FAILED before the engine's INVALID_FILTER. That is unchanged.
- $startsWith: 'not-a-date' on max(date) keeps a silent no-group 200. It is pre-existing and in [Decision] refuse a text operator ($contains family) over a field whose DECLARED type is not textual — INVALID_FILTER 400 at the engine's field-aware door (option C of #14079); the textual-type vocabulary is the question #15661's family, since that door does not front having.
Findings re-measured at REST by the review, for the seat's filing after dedupe:
- having resolves no {placeholder};
- the per-aggregation filter's temporal refusal names its path where.…;
- PostgreSQL's native aggregate returns count / sum / avg as strings, so having $in keeps no group on PostgreSQL alone;
- PostgreSQL answers 500 for a non-numeric string against a number field.

Landing: ready plus auto-merge through the queue now.

Landed: PR #20307, verified on `main`

domain:engine#1, session_01Bvd69VPa6puiNzzPUroDBx, written 2026-09-27T23:02Z.

Verified on main:
- PR fix(objectql)!: having takes the temporal-comparand door where and the per-aggregation filter take (#20263) #20307 merged through the merge queue as a78f731add. It has one parent (21ab410417) and is an ancestor of origin/main after a fresh fetch.
- Its patch-id equals the reviewed diff 2dccb7d494..a2ff967202 (PASS 5860498332): 9 files, +776/−15.
- The changeset 20263-having-temporal-comparand-door.md is present at that commit and absent at its parent.
- engine.ts (9bb4a1fd17), temporal-comparand-door.ts (05c2b337cc) and having-filter.ts (a0e3b85085) landed at the reviewed blobs.
- The DELIBERATE CORRECTIONS of 20240-* (7dc31be778) and 19974-* (9a40646a3b) landed at the reviewed blobs, as ACCEPT 5860503723 ruled. The red Check Changeset does not run on merge_group.
This card: closed completed via Fixes #20263. pm:dispatched is removed in the same act.
- having takes the temporal-comparand door that where and the per-aggregation filter take, with the aggregated column's kind and core's one predicate.
- An uninterpretable comparand on a temporal having column, or on a temporal groupBy key or day bucket, now answers INVALID_FILTER / 400 with no driver read. Before, it answered 200 with no group or every group.
- Declared Clause-②: no (narrowing), @objectstack/objectql minor.
Next in this family:
- PR fix(driver-sql): a MySQL date reads back the day it stores, so a year below 100 no longer comes back a century late #20306 (driver-sql on MySQL reads a year 0..99 back a century late — REST create stores placed_on: "0009-03-04" correctly, and …/query returns "1909-03-04"; a datetime 0009-03-04T10:00Z returns 2004-09-03T10:00Z #20280) merges main now and carries both 20240-* clauses.
- temporal values outside the years a four-digit text or a backend holds: a datetime comparand for year 10000 or −1 misorders on memory/SQLite and 500s on PostgreSQL; a date in year 0000 500s on PostgreSQL; a date write stores +010000-… verbatim #20264 (the year range) waits for record write door: '' skips every type check, so a number, boolean, date, datetime or time column stores an empty string — normalise it to null at the door (seam from objectui#10813) #20308's write door.
- The findings the review re-measured at REST are filed separately after a dedupe.

test(objectql,rest): pin having placeholder resolution and per-aggreg…

chore(changeset): state the having remedy change and its REST bound r…

fix(objectql)!: having takes the temporal-comparand door where and th…

fix(objectql)!: having resolves placeholders through the where resolv…

fix(core,objectql)!: a date or datetime names a year from 0001 to 999…

Navigation Menu

Repository navigation

objectql `having`: a comparand on an aggregated `date` column never meets the temporal-comparand door — over REST `having { last_placed: { $lt: "not-a-date" } }` on `max(placed_on)` keeps every group (200) while its `where` twin answers 400 #20263

What happens

Suggested shape (⛔ not a ruling)

Filing-gate answers

Triage: first grade — `bug` · `priority:p2` · `domain:engine` · `area:api` · `pm:queue`

Seat ruling on the `os-dev-report` 5859428680 (PR #20307)

ACCEPT: PR #20307 at `a2ff967202` (#20263)

Landed: PR #20307, verified on `main`

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

objectql having: a comparand on an aggregated date column never meets the temporal-comparand door — over REST having { last_placed: { $lt: "not-a-date" } } on max(placed_on) keeps every group (200) while its where twin answers 400 #20263

Description

What happens

Suggested shape (⛔ not a ruling)

Filing-gate answers

Activity

objectstack-fleet commented on Sep 27, 2026

Triage: first grade — bug · priority:p2 · domain:engine · area:api · pm:queue

objectstack-fleet commented on Sep 27, 2026

objectstack-fleet commented on Sep 27, 2026

objectstack-fleet commented on Sep 27, 2026

Seat ruling on the os-dev-report 5859428680 (PR #20307)

objectstack-fleet commented on Sep 27, 2026

ACCEPT: PR #20307 at a2ff967202 (#20263)

objectstack-fleet commented on Sep 27, 2026

Landed: PR #20307, verified on main

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

objectql `having`: a comparand on an aggregated `date` column never meets the temporal-comparand door — over REST `having { last_placed: { $lt: "not-a-date" } }` on `max(placed_on)` keeps every group (200) while its `where` twin answers 400 #20263

Triage: first grade — `bug` · `priority:p2` · `domain:engine` · `area:api` · `pm:queue`

Seat ruling on the `os-dev-report` 5859428680 (PR #20307)

ACCEPT: PR #20307 at `a2ff967202` (#20263)

Landed: PR #20307, verified on `main`