[maintainer] refactor(driver-sql): split src/sql-driver.ts (22,096 lines, +2,533 in 30 days) along its SqlDriver seams, one move-only slice per PR

Filing gate ③: a task the maintainer dispatched, quoted below. Filed by domain:services seat 1 (seat post #6021) · os-sales · session session_016g7nYP9wdNx1GDa5XDr7Uy, as the second engine-surface file of #24586's restart act. #25467 (protocol.ts) went first as the ruling orders, and its cut 1 landed in PR #25494. ⛔ Not a claim.

The maintainer's words, verbatim (#24586, recorded in 6270097840): 「24586 同意A，是否建议这波分拆等 v18 发布之后再处理。」 The restart act calls for "one [maintainer] card per file over 10,000 lines into its owning lane … Each card starts with a fresh seam measurement taken at that time, in #23050's shape". Triage read the restart condition as true in 6291735609.

Reader who acts: the domain:server lane, engine surface. One dispatch per slice. The first act is the seam table, sections by contiguous member ranges with their 30-day lines, measured fresh at claim.

Measured (on `origin/main` `842a0aef9d`, 2026-10-23)

Size and churn:
- 22,096 lines, up from 19,563 thirty days ago (dabf8d795e), a gain of +2,533 (+12.9%).
- 17 commits in 7 days and 67 in 30 days.
- git blame --since=2026-09-23T15:00Z attributes 3,752 lines (17.0%) to the last 30 days. Of the three engine-surface files, this is the coldest.
Structure:
- One SqlDriver class of 16,593 lines with 264 methods: 210 private or protected, 54 public.
- Only 1 public method has an inferred return type (supports, 25 lines). So the .d.ts limit measured on [maintainer] refactor(metadata-protocol): split src/protocol.ts (33,065 lines, +10,169 in 30 days) along its ObjectStackProtocolImplementation seams, one move-only slice per PR #25467 barely bites here.
- The rest is the module preamble and the helpers.
What reads it by path:
- the ADR anchor file scripts/adr-anchors/packages__drivers__driver-sql__src__sql-driver.ts.json;
- 9 files under scripts/;
- 15 files under docs/ and content/, 10 of them ADRs.
- src/index.ts re-exports named members from ./sql-driver.js (resolveSqliteAbsentFileTarget, withheldFilterDiagnosticOf, and others). Each slice checks that list against the .d.ts proof.
In flight: none of the 13 open PRs touch the file.

The ask (from #24586's ruling and #23050's shape, as on #25467)

Split along the existing seams, one area per PR. Each slice is move-only: no logic changes.
Shared state stays explicit. Moved code reaches the class's shared members through one named host object, as in [maintainer] refactor(rest): split packages/rest/src/rest-server.ts (13,701 lines, 70 commits in 30 days) along its register*Endpoints seams, one move-only slice per PR, /meta after #22877 stage 2 #23050, [maintainer] refactor(service-automation): split src/engine.ts (13,735 lines, +2,199 in 30 days) along its AutomationEngine seams, one move-only slice per PR #25222, [maintainer] refactor(plugin-security): split src/security-plugin.ts (10,746 lines, +2,803 in 30 days) along its SecurityPlugin seams, one move-only slice per PR #25223, [maintainer] refactor(plugin-auth): split src/auth-manager.ts (10,338 lines, +1,867 in 30 days) along its AuthManager seams, one move-only slice per PR #25224 and [maintainer] refactor(metadata-protocol): split src/protocol.ts (33,065 lines, +10,169 in 30 days) along its ObjectStackProtocolImplementation seams, one move-only slice per PR #25467.
- ⛔ Not a second copy of a helper.
- ⛔ Not a re-derivation.
Every slice proves it moved nothing:
- Every gate, baseline, ADR anchor and docs anchor that reads the file by path is migrated in the same PR and stays green.
- When an ADR anchor entry's paragraph describes a moved body, even half of it, the new module gets its own entry that carries that paragraph.
- The package's public .d.ts stays identical under ruling 6348891160 item 1, with a positive control.
- The decision-letter count of check:adr-anchors is reported at base and at head.
- ⛔ A gate whose target moved must not go silently empty.
Serial with the file's in-flight PRs. A slice waits for an open PR on the same region, or that PR merges main after the slice. Every slice measures the file's open PRs again at claim.
⛔ No new gate.

Series limits already measured (carry them forward)

Inferred return types. A public method whose return type is inferred cannot move under the host pattern. The .d.ts would then be inferred through the new module, and tsup drops property JSDoc and splices unrelated docblocks into it (measured on [maintainer] refactor(metadata-protocol): split src/protocol.ts (33,065 lines, +10,169 in 30 days) along its ObjectStackProtocolImplementation seams, one move-only slice per PR #25467 cut 2, getDiscovery). Such a member stays. Annotating it is a signature edit, which needs a ruling.
The tenant-audit census. UNTYPED_RECEIVERS is shrink-only. A body with a this.<field>.delete/insert/update( site becomes an untyped receiver in a Pick-host module, so that body stays unless a ruling says otherwise ([maintainer] refactor(plugin-auth): split src/auth-manager.ts (10,338 lines, +1,867 in 30 days) along its AuthManager seams, one move-only slice per PR #25224 cut 8, [maintainer] refactor(service-automation): split src/engine.ts (13,735 lines, +2,199 in 30 days) along its AutomationEngine seams, one move-only slice per PR #25222 slice 2).
Private statics stay in the class (entangled-shape rule, [maintainer] refactor(plugin-auth): split src/auth-manager.ts (10,338 lines, +1,867 in 30 days) along its AuthManager seams, one move-only slice per PR #25224).

Not in this card

Any behaviour change, and any design change to the shared members beyond giving them an explicit host.
objectql/src/engine.ts, filed beside this one as its own card.
driver-turso, which extends SqlDriver. Its readers of moved members are migrated in the same slice, but it is not split here.

Dedupe

MCP search_issues on this repository, closed issues included: "split objectql engine.ts or sql-driver.ts move-only cut ruling 24586". There were 113 hits. None is a split card for sql-driver.ts. The related ones are #24586 (the ruling), #25467 (protocol.ts), and #25222, #25223 and #25224 (the services files). New.

Generated by Claude Code

Claim: PM loop round 1
Session: session_01JVTd3PprStGa26aez5omLB
Account: sunhaolin (the seat's linked user as GET /user answers it; the card's assignee)
Branch: claude/issue-25521-sql-driver-cut-1
Worktree: objectstack-issue-25521
Domain: domain:server
Seat: domain:server#2
File surface: packages/drivers/driver-sql/src/** (sql-driver.ts and the new seam module of this slice), packages/drivers/driver-turso/src/** (readers of moved members only), scripts/adr-anchors/** entries for moved bodies, the path readers of sql-driver.ts under scripts/ and content/ that the slice must migrate, one changeset. ⛔ No docs/adr/** (governed) (stop on breach; explain in the report)
Container & model: M (one move-only slice plus the seam table), mode:subagent, model: TIER_DEFAULT (dispatch-gates --tier at 8801f2bb3: "no path-derived mandate … the tier stays the PM's per-card judgment call"), effort: high
Clause-②: no
Metadata-protocol: none
Responsibility: n/a — not a defect card
Thread-read: none
Serial constraints cleared: none. No open PR touches packages/drivers/driver-sql/src/sql-driver.ts (13 open PRs read this fire); last touch on main is 6abee03 (#25563, merged 2026-10-24).
Premise: re-read on origin/main 8801f2bb3: sql-driver.ts is 22,158 lines (the card measured 22,096 at 842a0aef9d); no split has started.

Generated by Claude Code

os-dev-report
{
"issue": 25521,
"status": "done",
"branch": "claude/issue-25521-sql-driver-cut-1",
"pr": "#25614",
"session": "session_01JVTd3PprStGa26aez5omLB — the dispatching seat's session (mode:subagent = parent's), on the Claude-Session: trailer of commit 021043d",
"premise_still_valid": true,
"summary": "Seam table measured fresh at origin/main e649c2d (sql-driver.ts 22,158 lines, 3,790 from the last 30 days, 67 commits in the window proven by git-history.mjs; class SqlDriver 347 members, 251 methods, 49 public, 32 statics, 1 public inferred return: supports) and cut 1 done move-only in draft PR #25614 (Part of #25521): the five protected catalog-introspection helpers (introspectColumnOrder, introspectColumns, introspectForeignKeys, introspectPrimaryKeys, introspectUniqueConstraints; class tail 21386-22083, 698 lines, 19 from 30 days) moved unedited into packages/drivers/driver-sql/src/sql-driver-introspection.ts through one named host (IntrospectionHost, members typed as SqlDriver indexed accesses; knex and the three dialect getters as getters, introspectColumnOrder as a late-bound closure so TursoDriver's override and spies still dispatch). Published dist/index.d.ts and index.d.mts are raw byte-identical to base, export lists identical (56 ESM, 56 CJS). No reader of the file by path needed migration and no adr-anchors entry was owed (no anchor paragraph describes a moved body). Proposed cut 2: Raw execution (10786-10911, 126 lines, 12 from 30 days); cuts of getBuilder-building areas (Distinct, Aggregation, Query plan, Window, CRUD, Bulk) first need check:tenant-chokepoint's SCAN_FILES question answered, and the hash-shadow and row-size families read protected statics. Draft only, per the release terms: no ready flip, no auto-merge, no queue.",
"tests": "head 021043d. driver-sql: tsc --noEmit exit 0; eslint --no-inline-config on sql-driver.ts and sql-driver-introspection.ts exit 0 (base sql-driver.ts also 0). pnpm --filter @objectstack/driver-sql exec vitest run --maxWorkers=2 base: Test Files 235 passed | 12 skipped (247), Tests 3787 passed | 212 skipped (3999), SQL_EXIT=0; head: identical counts, SQL_EXIT=0. pnpm --filter @objectstack/driver-turso exec vitest run --maxWorkers=2 (after rebuilding driver-sql dist at head) base: Test Files 95 passed (95), Tests 2492 passed | 34 skipped (2526), TURSO_EXIT=0; head identical. Per test name and status from vitest JSON, base vs head: driver-sql 0 lost / 0 changed / 0 added; driver-turso 0 / 0 / 0. Move-only proof (AST, scratch script): 5 functions, 1,546 tokens compared, 0 differ (3 parameter JSDoc tokens equal after removing exactly two columns per line); 591 lines compared, each the base line less two leading spaces or (86) byte-identical template-literal continuation lines; class 347 -> 347 members, 342 unmoved byte-identical, 5 stubs keep docblock+signature less async; top level 199 base statements identical, 1 import added; negative control (one token changed in a template literal) reports 2 differences, PROOF FAILED. .d.ts block (ruling 6348891160 item 2): raw sha256 base = head = 93da45168e82136b7613944525e47402d5dc6f9f8daa9efa4bdceeaa11b52a89 for index.d.ts and index.d.mts (the package emits no .d.cts); normalised base = head = 93da45168e82136b7613944525e47402d5dc6f9f8daa9efa4bdceeaa11b52a89 with the ruling's node one-liner verbatim; raw diff 0 lines; positive control: head index.d.ts line 8165 protected introspectColumns( made public -> raw and normalised e85d733d6d77d1a644666cc45167bf25bd4f1e13bef0a80af977a469339b5823; export lists ESM 56 / CJS 56, sorted-list sha256 d61a9dd3a02661602b1bb2dbe18152ac43986148f36b2746f40c0884bd8ffdec at all four. Head dist carries the move: 5 host .call( sites in each of dist/index.mjs and dist/index.js. No ablation: no tested behaviour changed.",
"gates": "Local, head 021043d. Card-named at base and head, each exit 0 both times: check:driver-conformance (covered 55, DEBT 0, exempt 0 at base and head), check:adr-anchors (84 anchored files; decision-letter citations on the anchored surface 1883 at base, 1883 at head), check:adr-symbol-anchors (2384 anchors resolve at both), check:tenant-chokepoint (24 getBuilder bindings across 3 files at both), check:logger-receiver-detach, check:object-def-param-keys, check:keyed-text-bounds. node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack: 55 commands, each run with its exit captured before any pipe; 52 exit 0 on first run; check-plugin-teardown-shape --self-test exit 3 (pinned fixture commit 621a487 outside the shallow clone) -> exit 0 after fetching that commit; check:lean-entry-closure exit 3 (no objectql dist) -> exit 0 after turbo run build --filter=@objectstack/objectql; check:dual-build-cjs-loads exit 3 PREREQUISITE NOT MET (needs every package's dist) = NOT MEASURED, CI runs it. --ran: '55 derived famil(ies) accounted for — 54 run, 1 NOT-MEASURED (1 DERIVED from a recorded exit 3)', exit 0. Not run locally, owned by CI: repo-wide pnpm lint (driver-sql has no lint script; eslint on both touched files exit 0 above), test:repo. check-platform-checklist --changed-from not owed (no exported symbol renamed or deleted). PR #25614 check-runs at report time: Check Changeset success, Check PR Size success, Check Documentation Links success, filter success, The card this PR closes must claim this branch success, Part-of PR must not also close its card success, No other open PR may claim the same issue success, No other open PR may claim the same single-writer path success; in_progress: Build Core, Test Core (1/6..6/6), Type Check · workspace / source gates / debt ledger / consumer gates, Lint · ESLint / repo & tooling gates / docs & code guards / CI, release & engine gates / PM dispatch-gates / scoped ×3, Governed Surface Queue Guard, Dogfood Regression Gate, Flag docs affected by code changes; skipped: opt-in harnesses, Console Pin Gate, Temporal Conformance, Build Docs, Dogfood Verify CLI.",
"dogfood_battery": "n/a — no ! in the PR title and no (narrowing) arm (Clause-②: no)",
"mcp_calls": "0 — no MCP GitHub tool called",
"api_writes": "4 REST writes in 2 relay strokes, both as objectstack-fleet[bot] through scripts/pm/with-fleet.sh --via dispatch: (1) pr_create = POST /repos/objectstack-ai/objectstack/pulls (draft) + POST /repos//issues/25614/assignees (sunhaolin) + POST /repos//issues/25614/labels (skip-changeset), run 40082061826 success, body read back byte-identical (14682 bytes), assignee and label read back; (2) this report, POST /repos//issues/25521/comments via post-stamped.mjs. Not REST: 2 git pushes of the branch (empty-branch probe, then 021043d).",
"open_questions": [],
"out_of_scope_findings": [
"carrier: #25521 (its next dispatch) · cuts of the getBuilder-building areas need check:tenant-chokepoint's SCAN_FILES (and its class-method walk) to cover the new module, or the gate silently narrows; protected statics block the hash-shadow and row-size families under the host pattern — both in the PR body's seam table, noted, not filed"
],
"files_changed": [
"packages/drivers/driver-sql/src/sql-driver.ts (+73 / -556; 22,158 -> 21,675 lines)",
"packages/drivers/driver-sql/src/sql-driver-introspection.ts (new, 665 lines)"
],
"line_budget": "+738 / -556 = 1,294 changed lines over 2 files, under the 3,000-line human-merge line (dispatch-gates: 'changed lines: 1294 … under')",
"deviations": [
"Commit trailers: the AGENTS.md model-free pair (Claude-Session + Co-authored-by: Claude) instead of the harness reminder's model-named Co-Authored-By line; PR body footer in the AGENTS.md session-URL form instead of the harness's line.",
"Claim 6382121925's file surface lists 'one changeset'; ruling 6348891160 item 4 and the dispatch say no changeset and skip-changeset; followed the ruling (declared in pr_create labels, no .changeset file).",
"Ruling 6348891160 names index.d.ts and index.d.cts; this package emits index.d.ts and index.d.mts (no .d.cts), so .d.mts was measured. Its unions are single-quoted (31, 0 double-quoted), so the ruling's normaliser is identity here and the proof rests on raw byte identity; the normaliser was not widened.",
"Dispatch assumption A2 (one adr-anchors entry per new module) not followed: card ask 3 makes it conditional on an anchor paragraph describing a moved body, and the sql-driver.ts entry (ADR-0053 formatOutput/presentReadValue, ADR-0120 D4) describes none.",
"The base and head test batches ran under one os-verify-lock call each with parts joined by ';', so the lock's VERDICT is batch-last-exit 0; each part's own exit is in its log (base SQL_EXIT=0 TURSO_EXIT=0; head BUILD_EXIT=0 SQL_EXIT=0 TURSO_EXIT=0).",
"Shared git state touched: history deepened with git fetch --shallow-since=2026-09-15 origin main, and commit 621a487 fetched by sha for the teardown self-test (shared object store; no ref of mine rewritten).",
"The cut commit was pushed before the second open-PR re-measure; the re-measure (13 open PRs, none touching sql-driver.ts or turso-driver.ts, origin/main unmoved at e649c2d) was taken before pr_create.",
"Teardown: worktree node_modules and the worktree are removed right after this comment posts (post-stamped.mjs runs from the worktree's node_modules); the final message is sent after that removal."
]
}

Generated by Claude Code

ACCEPT → PR #25614 comment 6382618649 (cut 1, head 021043d578; lands once CI is green on that head) · domain:server seat 2 · session_01JVTd3PprStGa26aez5omLB

Generated by Claude Code

Release: domain:server seat 2 · sunhaolin · session session_01JVTd3PprStGa26aez5omLB · 2026-10-24T10:35Z · cause: cut 1 landed (Part of #25521) · destination: pm:queue, unassigned, for cut 2

Landed: PR #25614 → `f7b029896` (#25521 cut 1)

reading	result
PR #25614 object	`merged: true`, merge commit `f7b02989688255549339747d03a42d65940b2c19`, merged 2026-10-24T10:34:28Z through the merge queue
`origin/main`	`f7b029896` is an ancestor of `origin/main` (`f7b0298968`), single parent (squash)
queue branch	no `gh-readonly-queue/main/*` head for PR 25614 remains

Landed: the five catalog-introspection helpers (introspectColumnOrder, introspectColumns, introspectForeignKeys, introspectPrimaryKeys, introspectUniqueConstraints) live in packages/drivers/driver-sql/src/sql-driver-introspection.ts. sql-driver.ts is 21,675 lines.
Remains: every other area of the seam table in PR refactor(driver-sql): move the catalog-introspection helpers out of sql-driver.ts into sql-driver-introspection.ts (#25521 cut 1) #25614's body. The dev proposes Raw execution for cut 2. Carried constraints: a cut in a getBuilder-building area needs check:tenant-chokepoint's SCAN_FILES to cover the new module first; protected statics keep the hash-shadow and row-size families in the class; insertOnlyUpsertColumns stays in sql-driver.ts while drivers(sql): upsert's empty-merge-set fallback is merge-ALL, which re-admits every insert-only column — currently unreachable, but it is the wrong shape for "nothing to merge" #8740's restart line names that file.
State: pm:dispatched → pm:queue, assignee cleared, in this act. The series card stays open.

Generated by Claude Code

ACCEPT → PR #25655 comment 6385306916 (cut 3, head 59d63a8d1b; landing in the same act) · domain:server seat 2 · session_01JVTd3PprStGa26aez5omLB

Generated by Claude Code

Release: domain:server seat 2 · sunhaolin · session session_01JVTd3PprStGa26aez5omLB · 2026-10-24T15:03Z · cause: cut 3 landed (Part of #25521) · destination: pm:queue, unassigned, for cut 4

Landed: PR #25655 → `645ede5e1` (#25521 cut 3)

reading	result
PR #25655 object	`merged: true`, merge commit `645ede5e1662cc9a99f721ea61b1766648fdae76`, merged 2026-10-24T15:01:15Z through the merge queue
`origin/main`	`645ede5e1` is an ancestor of `origin/main` (`dc6c7ee738`), single parent (squash)
queue branch	no `gh-readonly-queue/main/*` head for PR 25655 remains

Landed: the transaction doors (beginTransaction, assertBareKnexSafe, commit, rollback, commitTransaction, rollbackTransaction) live in packages/drivers/driver-sql/src/sql-driver-transactions.ts; releaseTransaction stays in the class. sql-driver.ts is 21,678 lines on origin/main (it grew by 27: the stubs are longer than the bodies they replaced).
Rule for the next cuts (the seat's ACCEPT, 6385306916): from cut 4 on, an area moves only when sql-driver.ts's net line count drops.
Remains: the dev proposes Schema introspection for cut 4 (listPhysicalColumns, introspectSchema; 136 lines, 65 from the last 30 days; no getBuilder build, no static read, no this.FIELD.delete/insert/update( site). It is priced against the net-line rule at dispatch. Constraints carried: no getBuilder-building area until check:tenant-chokepoint's SCAN_FILES covers the new modules; protected statics stay; insertOnlyUpsertColumns stays in sql-driver.ts (drivers(sql): upsert's empty-merge-set fallback is merge-ALL, which re-admits every insert-only column — currently unreachable, but it is the wrong shape for "nothing to merge" #8740).
State: pm:dispatched → pm:queue, assignee cleared, in this act. Dev was a local subagent; no cloud session to archive.

Generated by Claude Code

Claim: PM loop round 1 · cut 4
Session: session_01JVTd3PprStGa26aez5omLB
Account: sunhaolin (the seat's linked user as GET /user answers it; the card's assignee)
Branch: claude/issue-25521-sql-driver-cut-4
Worktree: objectstack-issue-25521
Domain: domain:server
Seat: domain:server#2
File surface: packages/drivers/driver-sql/src/** (sql-driver.ts and the seam module of cut 4; the cut-3 dev proposes Schema introspection: listPhysicalColumns, introspectSchema), packages/drivers/driver-turso/src/** and packages/drivers/driver-sqlite-wasm/src/** (readers of moved members only), scripts/adr-anchors/** entries a moved body's anchored paragraph needs. skip-changeset under ruling 6348891160. ⛔ No docs/adr/** (stop on breach; explain in the report)
Container & model: M, mode:subagent, model: TIER_DEFAULT (dispatch-gates --tier: "no path-derived mandate … the tier stays the PM's per-card judgment call"), effort: high
Clause-②: no
Metadata-protocol: none
Responsibility: n/a — not a defect card
Thread-read: 6385766523
Serial constraints cleared: cut 3 (PR #25655) merged as 645ede5e1; no open PR touches sql-driver.ts (every open PR's file list read this fire). insertOnlyUpsertColumns stays in sql-driver.ts (#8740).
Premise: re-read on origin/main: sql-driver.ts is 21,678 lines; listPhysicalColumns(object, options?): Promise<string[]> and introspectSchema(): Promise<IntrospectedSchema> are in the class with explicit return types. The net-line rule (ACCEPT 6385306916): this cut moves only if the file's net line count drops.

Generated by Claude Code

refactor(driver-sql): move introspectSchema's body out of sql-driver.…

os-dev-report
{
"issue": 25521,
"status": "done",
"branch": "claude/issue-25521-sql-driver-cut-4",
"pr": "#25698",
"session": "session_01JVTd3PprStGa26aez5omLB — the dispatching seat's session (mode:subagent = parent's), on the Claude-Session: trailer of commit f73c545",
"premise_still_valid": true,
"summary": "Claim 6386825345 (newest Claim:) names this branch; verified before any edit. Seam table re-measured at origin/main 6021b5e (sql-driver.ts 21,678 lines, 3,815 = 17.6% from the 30 days before the base commit, 69 commits in the window proven by git-history.mjs; class SqlDriver 5499-21603, 347 members, 251 methods, 49 public, 32 statics; the one inferred public member is the getter supports). A1 holds with one change: Schema introspection (15794-15929, 136 lines, 65 from 30 days) re-measures clean, but only introspectSchema moves (104 lines, 38 from 30 days). listPhysicalColumns stays, because its 6-line body is shorter than the 10-line stub its 3 getters would need (+5 lines on its own; the seat's cut-3 ACCEPT keeps such bodies in the class). Cut 4 done move-only in draft PR #25698 (Part of #25521): introspectSchema's body moved unedited into cut 1's module sql-driver-introspection.ts, not a new one. It is the public read those five helpers serve, its body calls four of them, and four of its five field/getter reads are already IntrospectionHost members, so the one host grows by five members (dialectName as a getter; introspectColumns, introspectForeignKeys, introspectPrimaryKeys, introspectUniqueConstraints as late-bound closures) instead of a second host retyping the same four. postgresTableNameCollisionError, read only by this body, stays in sql-driver.ts and gains a leading export (cut 2's accepted option A). A1b holds: sql-driver.ts 21,678 -> 21,595 (-83). A3 measured: TursoDriver overrides introspectSchema (calls super) and introspectColumns (one parameter, calls super); SqliteWasmDriver overrides isSqlite; an instance spy on introspectColumns exists in a driver-sql test; a scratch probe gave 14 identical readings at base and head. Published dist/index.d.ts and index.d.mts are raw byte-identical to base, export lists identical (56/56). check:driver-conformance covered 55 / DEBT 0 at both ends. Proposed cut 5: introspectIndexes (Schema sync, 15097-15249, 153 lines, 0 from 30 days, into the same module; priced at about -82 lines). Draft only: no ready flip, no auto-merge, no queue.",
"tests": "head f73c545. driver-sql: pnpm exec tsc --noEmit exit 0; pnpm exec eslint --no-inline-config packages/drivers/driver-sql/src VERDICT command-exit 0, JSON reporter 265 files, 0 errors, 0 warnings (new module included). Each suite under its own os-verify-lock call, VERDICT command-exit 0 at both ends: pnpm --filter @objectstack/driver-sql exec vitest run --maxWorkers=2 base and head: Test Files 235 passed | 12 skipped (247), Tests 3787 passed | 212 skipped (3999). pnpm --filter @objectstack/driver-turso exec vitest run --maxWorkers=2 (each end against that end's rebuilt driver-sql dist) base and head: Test Files 95 passed (95), Tests 2492 passed | 34 skipped (2526). pnpm --filter @objectstack/driver-sqlite-wasm exec vitest run --maxWorkers=2 base and head: Test Files 39 passed (39), Tests 704 passed (704). Per test name and status from vitest JSON, base vs head: driver-sql 0 lost / 0 changed / 0 added (3999 names); driver-turso 0 / 0 / 0 (2513 distinct, 13 repeat); driver-sqlite-wasm 0 / 0 / 0 (704). Move-only proof (AST, scratch script, read from git show of base and HEAD): introspectSchema, 567 tokens compared, 0 differ, async on both; 104 lines compared (87 the base line less two leading spaces, 16 template-literal continuation lines byte-identical, the first line identical less the added this parameter); class 347 -> 347 members, 346 unmoved byte-identical, 1 stub keeps signature less async; sql-driver.ts top level 202 -> 202 non-class statements, 200 byte-identical, 1 identical but for a leading export (postgresTableNameCollisionError), 1 import gaining exactly the specifier introspectSchema (leading comment re-worded); module: cut 1's 5 functions byte-identical with leading trivia, IntrospectionHost 5 -> 10 members with every base member line kept, imports gain type IntrospectedSchema, type IntrospectedTable, value postgresTableNameCollisionError, no other statement; PROOF exit 0. Negative control ('sqlite_%' -> 'sqlite%' in the SQLite arm's template literal of a copy of the module): 1 token + 1 line different, PROOF FAILED, exit 1. Late-binding probe (scratch, node against built dist, not committed): 14 readings, base and head output identical (diff empty). .d.ts block (ruling 6348891160 letter A): raw sha256 base = head = 93da45168e82136b7613944525e47402d5dc6f9f8daa9efa4bdceeaa11b52a89 for index.d.ts and index.d.mts (no .d.cts emitted); ruling one-liner verbatim (double-quoted) base = head = 93da45168e82136b7613944525e47402d5dc6f9f8daa9efa4bdceeaa11b52a89 (0 double-quoted unions); same one-liner with each double quote in the regex swapped for a single quote, base = head = 118e992ad8aad590044eb31509dca65e6b321e3410ce846e1e5f5dd34b956e95 (21 lines carry a single-quoted union); raw diff 0 lines; positive control in the same block: head index.d.ts line 5899 introspectSchema(): Promise...; made protected -> raw and double-normalised 402f862acf50c10714bd3aef080c5b397a15b9ef1712180ee86cf30a25e6709c, single-normalised e12430bcae9b43a1e4e9ff7914e5c12a534cc1765bc3c32a642a11977e4812ae; export lists ESM 56 / CJS 56, sorted-list sha256 d61a9dd3a02661602b1bb2dbe18152ac43986148f36b2746f40c0884bd8ffdec at all four. Head dist carries the move: 15 host .call( sites in each of dist/index.mjs and dist/index.js (14 at base + 1). No ablation: no tested behaviour changed. test:repo not run locally (diff touches no repo-project test; CI's Test Core runs it).",
"gates": "Local, head f73c545. Card-named at base and head, exit 0 both times: check:driver-conformance (covered 55, DEBT 0, exempt 0 at base and head), check:adr-anchors (85 anchored files; decision-letter citations on the anchored surface 1847 at base, 1847 at head), check:adr-symbol-anchors (2384 anchors resolve at both), check:tenant-chokepoint (rule 1: 24 getBuilder bindings across 3 files at both; rule 2 clean at both, its literal raw-fragment population in sql-driver.ts 76 -> 73, see out_of_scope_findings), check:logger-receiver-detach (0 detaches at both, 3687 files). node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack (change set vs merge base 6021b5e): 55 commands, each run with its exit captured before any pipe (redirect-then-capture); 53 exit 0 on first run; check:lean-entry-closure exit 3 (no objectql dist) -> exit 0 after turbo run build --filter=@objectstack/objectql under the lock; check:dual-build-cjs-loads exit 3 PREREQUISITE NOT MET (needs every package's dist) = NOT MEASURED, CI runs it. --ran: '55 derived famil(ies) accounted for — 54 run, 1 NOT-MEASURED (1 DERIVED from a recorded exit 3)', exit 0; size line 'changed lines: 278 (+165 / -113; 2 file(s)) vs the human-merge threshold 3000: under'. Not run locally, owned by CI: repo-wide pnpm lint (driver-sql has no lint script; eslint over the package src exit 0 above), test:repo. check-platform-checklist --changed-from not owed (no exported symbol renamed or deleted; one export added in a non-published module path). PR #25698 check-runs at report time: success 8 (Check Changeset, Check Documentation Links, Check PR Size, filter, The card this PR closes must claim this branch, Part-of PR must not also close its card, No other open PR may claim the same issue, No other open PR may claim the same single-writer path); in_progress 23 (Build Core, Test Core 1/6..6/6, Type Check · workspace / source gates / debt ledger / consumer gates, Lint · ESLint / repo & tooling gates / docs & code guards / CI, release & engine gates / PM dispatch-gates / scoped ×3, Governed Surface Queue Guard, Dogfood Regression Gate, Flag docs affected by code changes, Auto Label); skipped 12 (opt-in harnesses, Console Pin Gate, Temporal Conformance, Build Docs, Dogfood Verify CLI, dogfood shards, duplicate Check Changeset / Check PR Size / Auto Label runs). in_progress is the honest value; CI convergence is the seat's.",
"dogfood_battery": "n/a — no ! in the PR title and Clause-②: no (no (narrowing) arm)",
"mcp_calls": "0",
"api_writes": "2 — both through the fleet-write relay as objectstack-fleet[bot] (POST /repos/objectstack-ai/objectstack/dispatches each): pr_create (draft PR #25698 with assignees [sunhaolin] and labels [skip-changeset]; read-back 20643 bytes sent = stored), and the os-dev-report comment on #25521 via scripts/pm/post-stamped.mjs. Not REST: two git push (the empty landing-marker branch, then f73c545). All other GitHub traffic was REST reads (card, comments, three PR bodies, two ACCEPT comments, open-PR file lists, check-runs).",
"files_changed": [
"packages/drivers/driver-sql/src/sql-driver.ts (+24 / -107; 21,678 -> 21,595 lines)",
"packages/drivers/driver-sql/src/sql-driver-introspection.ts (+141 / -6; 665 -> 800 lines)"
],
"line_budget": "sql-driver.ts 21,678 -> 21,595 (-83, the A1b net-line rule holds); sql-driver-introspection.ts 665 -> 800 (+135); +165 / -113 = 278 changed lines over 2 files, under the 3,000-line human-merge line (dispatch-gates --ran: 'changed lines: 278 … under')",
"open_questions": [
{
"question": "listPhysicalColumns stays in the class. The dispatch's A1 named both Schema introspection members. Moving it alone would add 5 lines to sql-driver.ts (a 10-line stub with 3 getters in place of a 6-line body, plus one import name). Moving both would still be net-negative (-78 instead of -83).",
"options": [
"A — keep it in the class, as done: the cut-3 ACCEPT keeps a body that is shorter than its host literal in the class.",
"B — move it too, in a follow-up push to this PR: the section empties into one module, and the file is 5 lines longer than in A."
],
"recommendation": "A, because the seat's rule prices each body against its stub. A move that lengthens sql-driver.ts buys only a section-tidiness claim, and listPhysicalColumns is a 6-line read that has no override reach problem in the class."
},
{
"question": "docs/qa/platform-checklist/areas/integration-system.json cites packages/drivers/driver-sql/src/sql-driver.ts#introspectSchema (introspectSchema; SQLite branch reads sqlite_master …) as an item source. The anchor still resolves to the method (check:adr-symbol-anchors green), but the SQLite branch it describes now lives in sql-driver-introspection.ts. Card ask 3 migrates docs anchors that read the file by path. This one did not break, and the file is outside the claim's file surface.",
"options": [
"A — leave it: the anchor names the door, which did not move, and its behaviour claim is still true of introspectSchema (cut 3 left ADR-0119's sql-driver.ts#beginTransaction the same way).",
"B — add a second source line naming sql-driver-introspection.ts#introspectSchema, with the item's history revision, in this PR after a claim surface extension, or in a docs-only follow-up."
],
"recommendation": "A, because the anchor is a declaration anchor and stays true, and a checklist edit needs a history revision on a file this claim does not own. If the seat prefers B, a docs-only follow-up after the series lands catches every cut's stale parenthetical in one pass."
}
],
"out_of_scope_findings": [
"carrier: #25521 (the series card, which already carries the check:tenant-chokepoint SCAN_FILES precondition) · noted, not filed — SCAN_FILES names sql-driver.ts but not the seam modules. Rule 2 (no IS NULL arm on a tenant column) visits every raw-family call in those files, so it no longer reads the three catalog-listing knex.raw( fragments this cut moved. Its literal raw-fragment population in sql-driver.ts went 76 -> 73, measured by AST at base and head. Cut 1 had already moved 16 into the same module, and the module now holds 19. None names a tenant column or says IS NULL, so the verdict is unchanged and the gate does not go empty. Widening SCAN_FILES is a gate edit outside a move-only cut. Not class a/b/c: no wrong answer at any public door today."
],
"deviations": [
"Commit trailers: the AGENTS.md model-free pair (Claude-Session + Co-authored-by: Claude) instead of the harness reminder's model-named Co-Authored-By line; PR body footer in the AGENTS.md session-URL form instead of the harness's two lines.",
"The dispatch named the worktree /home/user/objectstack-issue-25521 as created; it did not exist at start, so this run created it from origin/main 6021b5e (git worktree add --no-track) and pushed the empty branch first as the landing marker.",
"A1 named both listPhysicalColumns and introspectSchema; only introspectSchema moved (open question 1).",
"The dispatch asks for pnpm lint on the touched packages: driver-sql declares no lint script, so eslint --no-inline-config packages/drivers/driver-sql/src ran instead (265 files, 0 errors, 0 warnings); the repo-wide run is CI's.",
"pr_create takes no draft key (the relay refused it in a dry run, zero writes); the op always sends draft: true, and the read-back shows draft: true.",
"Base leg: after the cut was committed and pushed, the base sql-driver.ts and sql-driver-introspection.ts were restored into this worktree (tree only, git restore --source=6021b5e83, blob hashes checked on disk) for the base build, probe and three suites. They were then restored with git checkout HEAD -- under an EXIT/INT/TERM trap, proved by HEAD blob hashes f2cc60c80 / d922dbdd6 matching, git diff HEAD 0 lines and 0 status entries. The head dist was rebuilt before any head reading.",
"Open-PR re-measure: gh api --paginate on the Version Packages PR #25640's file list hit a proxy 403 (numeric-ID pagination path); its 183 files were read with explicit page parameters, at start and before pr_create, and none is under the three drivers' src.",
"origin/main moved two commits past the base (#25603, #25684, now 1fbd9de); not merged in, as neither touches the three driver packages or a derived gate script.",
"Shared git state touched: git fetch origin main twice (moves the shared origin/main ref); no history deepening was needed (floor 2026-09-15 predates the window).",
"Teardown: worktree node_modules and the worktree are removed right after this comment posts; the final message is sent after that removal."
]
}

Generated by Claude Code

Release: domain:server seat 2 · sunhaolin · session session_01JVTd3PprStGa26aez5omLB · 2026-10-24T18:50Z · cause: cut 4 landed (Part of #25521) · destination: pm:queue, unassigned, for cut 5

Landed: PR #25698 → `b1790841b` (#25521 cut 4)

reading	result
PR #25698 object	`merged: true`, merge commit `b1790841bb0bf2c7c696e858de51e8105704471a`, merged 2026-10-24T18:49:39Z through the merge queue
`origin/main`	`b1790841b` is `origin/main`'s tip; two parents (a merge commit)
queue branch	no `gh-readonly-queue/main/*` head for PR 25698 remains

Landed: introspectSchema lives in packages/drivers/driver-sql/src/sql-driver-introspection.ts beside cut 1's helpers (IntrospectionHost grew by five members); listPhysicalColumns stays in the class (its body is shorter than its stub). sql-driver.ts is 21,595 lines on origin/main (−83).
Remains: the dev proposes introspectIndexes (Schema sync, 153 lines, 0 from the last 30 days) for cut 5, into the same module, priced at about −82 lines. Constraints carried: the net-line rule; no getBuilder-building area until check:tenant-chokepoint's SCAN_FILES covers the seam modules (its rule-2 population in sql-driver.ts went 76 → 73 with this cut; verdict unchanged); protected statics stay; insertOnlyUpsertColumns stays (drivers(sql): upsert's empty-merge-set fallback is merge-ALL, which re-admits every insert-only column — currently unreachable, but it is the wrong shape for "nothing to merge" #8740).
State: pm:dispatched → pm:queue, assignee cleared, in this act. Dev was a local subagent; no cloud session to archive.

Generated by Claude Code

Navigation Menu

Repository navigation

[maintainer] refactor(driver-sql): split `src/sql-driver.ts` (22,096 lines, +2,533 in 30 days) along its `SqlDriver` seams, one move-only slice per PR #25521

Measured (on `origin/main` `842a0aef9d`, 2026-10-23)

The ask (from #24586's ruling and #23050's shape, as on #25467)

Series limits already measured (carry them forward)

Not in this card

Dedupe

Landed: PR #25614 → `f7b029896` (#25521 cut 1)

14 remaining items

Landed: PR #25655 → `645ede5e1` (#25521 cut 3)

Landed: PR #25698 → `b1790841b` (#25521 cut 4)

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[maintainer] refactor(driver-sql): split src/sql-driver.ts (22,096 lines, +2,533 in 30 days) along its SqlDriver seams, one move-only slice per PR #25521

Description

Measured (on origin/main 842a0aef9d, 2026-10-23)

The ask (from #24586's ruling and #23050's shape, as on #25467)

Series limits already measured (carry them forward)

Not in this card

Dedupe

Activity

objectstack-fleet commented on Oct 24, 2026

objectstack-fleet commented on Oct 24, 2026

objectstack-fleet commented on Oct 24, 2026

objectstack-fleet commented on Oct 24, 2026

Landed: PR #25614 → f7b029896 (#25521 cut 1)

14 remaining items

objectstack-fleet commented on Oct 24, 2026

objectstack-fleet commented on Oct 24, 2026

Landed: PR #25655 → 645ede5e1 (#25521 cut 3)

objectstack-fleet commented on Oct 24, 2026

objectstack-fleet commented on Oct 24, 2026

objectstack-fleet commented on Oct 24, 2026

Landed: PR #25698 → b1790841b (#25521 cut 4)

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

[maintainer] refactor(driver-sql): split `src/sql-driver.ts` (22,096 lines, +2,533 in 30 days) along its `SqlDriver` seams, one move-only slice per PR #25521

Measured (on `origin/main` `842a0aef9d`, 2026-10-23)

Landed: PR #25614 → `f7b029896` (#25521 cut 1)

Landed: PR #25655 → `645ede5e1` (#25521 cut 3)

Landed: PR #25698 → `b1790841b` (#25521 cut 4)