spec(data): retire the import mapping lookup transform's steering params (#10329, ADR-0049)

[os-sam]

Fixes#10329

Implements the triage ruling (comment 5383458820): option 1 — retire the four inert lookup params (object, fromField, toField, autoCreate) from ImportFieldMappingSchema.params under ADR-0049 enforce-or-remove, with every alias spelling converted to guidance so each of the fifteen spellings lands on a prescription. Option 2 (a second reference-resolution dialect in applyMappingToRows) is not implemented, per the ruling.

Premise re-verification (on origin/main at 064d484bd, pre-fix)

• The four keys and the alias set were still declared in packages/spec/src/data/mapping.zod.ts (shape lines 165–168; aliases folding lookupObject/targetObject, match/matchOn/matchField/keyField, returnField/valueField, create/createIfMissing/upsert onto them).
• applyMappingToRows still handles lookup in the same branch as none (packages/rest/src/import-mapping.ts:122-127 — the cell is copied through; the comment defers resolution to the pipeline's metaMap).
• The liveness mapping.json sub-walk boundary note still recorded the four keys as read by nothing.
• Repo-wide grep for fromField / autoCreate / the alias spellings outside mapping.zod.ts + its tests returned only unrelated identifiers (sql-driver.tsuniqueIndexesFromFields, tenant.zod.tsautoCreateSchema/autoCreateDatabase, field.zod.ts's own lookupObject → reference alias). Positive control: the same grep instrument finds the real reader of the neighbouring params.valueMap at packages/rest/src/import-mapping.ts:134-136.

premise_still_valid: true.

What changed

• Schema (packages/spec/src/data/mapping.zod.ts): the four keys are strict-deleted from the params shape (params is a strictObject, so this is the guidance-map route, not retiredKey() — the 17.0.0 #4488 审计发现的四个"授权门断连"：email_template / job / validation 的元数据条目到不了执行点，action 导航项点不动 #4509 treatment of extractQuery/errorPolicy/batchSize on this same schema, one level down). PARAMS_RETIRED_KEY_GUIDANCE carries fifteen entries — 4 canonical + 11 ex-alias spellings — each landing on a full prescription (fully-qualified key, why it was inert, the imperative fix, the os migrate meta --from 17 sentence). Per the ruling, the autoCreate guidance says what actually happens: nothing was ever created — an unresolved reference fails the row with import_reference_not_found, with or without the key.
• ADR-0087 D2 conversionmapping-lookup-params-removed (protocol 18, retiredFromLoadPath: true) strips the four keys from stored mappings[].fieldMapping[].params (pure lossless deletes; drilled two levels down, the metric-filters-removed shape one level deeper; fixture expectedNotices: 4). Wired into the step-18 chain (conversionIds + rationale extended).
• No RETIRED_KEYS_BY_MAJOR entries — deliberately. These keys sit one sub-walk level below the authorable-surface drill: authorable-surface/data.json records data/ImportFieldMapping:params only (that row stays live), so there is no defKey:name row for the four to register, delete, or age out — gates (a)/(b)/(b2)/(c) never see them. This mirrors how extractQuery/errorPolicy/batchSize left this same schema at 17.0.0 (conversion + guidance, no retired-key table entries). The dispatch's "retired-key entries" default yields to the playbook's route table and this file's own precedent, as instructed.
• Liveness ledger (packages/spec/liveness/mapping.json): the fieldMapping sub-walk boundary note now records the retirement (strict deletion — nothing dead remains below the drill) instead of parking the finding. Entry stays live; no row deletion (the four never had rows — sub-walk).
• Docs: the import-mappings.mdx params warning (the callout pointing at spec(data): mapping lookup transform params (object / fromField / toField / autoCreate) are authorable and read by nothing #10329) is deleted — that page's SDK warning is untouched (a batch sibling owns it). content/docs/references/data/mapping.mdx regenerated: the params row now lists only value/valueMap/separator.
• Tests (mapping.test.ts): refusal pins per key asserting the prescription text (the autoCreate pin asserts "nothing was ever created" + import_reference_not_found), alias-routing pins, and a surviving-surface pin. The old accept-lookup-params tests are replaced.
• Changeset.changeset/mapping-lookup-params-removed.md: minor per the lockstep launch-window convention (post-17.0.0-cut, prescription registered at protocol 18 — the element-form-retired precedent; check-changeset-no-major green confirms the convention), FROM → TO per key, one-line fix, plus the adr-0087 registered mapping-lookup-params-removed marker.
• Untouched by design: packages/rest/src/import-mapping.ts (the pass-through stays; its comments never named the four keys, so no comment edit either), content/docs/releases/, skills/** (no diff — the two skill readings are therefore not applicable).

Note: spec-changes.json / the upgrade guide deliberately do not change — PROTOCOL_MAJOR is 17 and the generator folds up to the current major only; the protocol-18 conversions (this one and its step-18 siblings, e.g. metric-filters-removed) enter those artifacts when 18 is cut. check:spec-changes / check:upgrade-guide green.

Reverse verification (from the committed state)

Predicted direction stated before the run: restoring the pre-fix mapping.zod.ts from base 064d484bd must turn the five new refusal/alias pins RED (the keys parse cleanly again, so the expected throws never happen) and leave the surviving-surface pins green. Mutation proven on disk before reading results (grep -c 'autoCreate: z.boolean().optional()' = 1, grep -c PARAMS_RETIRED_KEY_GUIDANCE = 0). Observed: exactly 5 failed / 33 passed. No rebuild leg needed — mapping.test.ts imports ./mapping.zod relatively (src, never through the package exports/dist), so the mutated source is the resolved module by construction. Restore leg: git checkout HEAD -- …, restoration proven on disk (greps invert: 3 / 0), re-run 38/38 green, git status clean.

Verification (all at commit ea258bef9, the head this PR ships; exit codes captured before any pipe, verdicts quoted from the runners' own lines)

• pnpm --filter @objectstack/spec build — os-verify-lock: VERDICT command-exit 0
• pnpm --filter @objectstack/spec check:generated — after --fix regenerated the one proved-stale artifact (gen:docs): all runnable gates green (check:authorable-surface ✓ — zero surface change, correct for a sub-walk deletion; check:migration-registry ✓; check:spec-changes ✓; check:upgrade-guide ✓)
• pnpm --filter @objectstack/spec test — Test Files 418 passed (418) · Tests 11118 passed (11118), VERDICT command-exit 0
• pnpm --filter @objectstack/rest test (after building the rest dependency closure) — Test Files 138 passed (138) · Tests 2201 passed (2201), VERDICT command-exit 0
• pnpm --filter @objectstack/spec typecheck / pnpm --filter @objectstack/rest typecheck — VERDICT command-exit 0 both
• node scripts/pm/dispatch-gates.mjs (no paths — the script derived the changeset itself; its stderr attributes the answer to objectstack-ai/objectstack at ea258bef9): 31 path-matched + 6 convention-triggered families derived. All 36 runnable local families run individually with captured exit codes: check-dev-prereqs failed locally only on the fresh-worktree precondition "the workspace is not built" (43 packages without dist — an environment fact, not a diff finding; CI builds fresh); after turbo run build --filter=./packages/* --filter=./packages/*/* it passes (exit 0), making it 36/36 PASS
• Convention gates from the test-file edit: check:query-options-erasure, check:type-check-coverage, check:engine-double-contract, check:where-matcher, check:cross-package-test-inputs — PASS; check:type-check-debt (self-test + --re-measure) run after the full packages-closure build — exit 0, PASS
• pnpm check:nul-bytes — PASS

CI convergence is reviewed by the PM after this report, per the standing dispatch contract.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 1 package(s): @objectstack/spec, touching 27 documentable anchor(s).

17 hand-written doc(s) name something this change touched — list omitted above 15 rows. Re-derive on the tree named below: node scripts/docs-audit/affected-docs.mjs --json 050d8d8685312ee6a0a77bd9036becf30750fd31.

⛔ 3 release-owned page(s) also affected — read-only, see AGENTS.md Documentation Guardrails.

What this run could not see

• 1 changed file(s) yielded no anchor (packages/spec/liveness/mapping.json) — pages documenting those are invisible to this run
• 9 name(s) were too generic to anchor anything (single lowercase words)
• the SDK route bridge reached 45 of 221 client-bound route-ledger rows — the other 176 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run: node scripts/docs-audit/affected-docs.mjs --bridge-coverage

Coarse fallback — 126 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 050d8d8685312ee6a0a77bd9036becf30750fd31 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 56dffce3d7c625c473d07b8aaa4968c8dc84dd8a — the merge of head ea258bef9fdea9fa52d0f650b0edbfa0052cda44 into base 050d8d8685312ee6a0a77bd9036becf30750fd31, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 56dffce3d7c625c473d07b8aaa4968c8dc84dd8a && git checkout 56dffce3d7c625c473d07b8aaa4968c8dc84dd8a
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 050d8d8685312ee6a0a77bd9036becf30750fd31 ea258bef9fdea9fa52d0f650b0edbfa0052cda44 && git checkout -B drift-repro 050d8d8685312ee6a0a77bd9036becf30750fd31 && git merge --no-ff ea258bef9fdea9fa52d0f650b0edbfa0052cda44
node scripts/docs-audit/affected-docs.mjs --json 050d8d8685312ee6a0a77bd9036becf30750fd31

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs 050d8d8685312ee6a0a77bd9036becf30750fd31 → pass the list as
args.docs, on the commit named under Which tree this was computed on.

[github-actions]

⛔ merge queue 构建失败 — 先分诊，再决定要不要重排

队列构建 32629757073 红了。队列跑的是全量套件（PR 侧 CI 只跑 affected 子集），
所以失败的测试可能在本 PR 没碰过的包里 —— 那不是重排能修的。每次盲目重排都会让排在后面的所有 PR 重建一轮。

失败的 job（日志抽取，best effort）：

• Dogfood Verify CLI — 失败步骤: Verify pnpm version

  AssertionError [ERR_ASSERTION]: The expression evaluated to a falsy value:
  

• Test Core (4/6) — 失败步骤: Verify pnpm version

  AssertionError [ERR_ASSERTION]: The expression evaluated to a falsy value:
  

↳ 失败原因 是判读的关键：超时（Test timed out in … / Hook timed out in …）多半是负载/时序，不是本 PR 的回归；
断言（AssertionError: …）才指向真实的行为改变。两者的 FAIL 行长得一模一样，只有这一行能区分。

跨 PR 相同签名（24h，按失败测试文件聚合）：

• ⚠️本次没有可用的聚合签名（日志里没有能解析出测试文件名的 FAIL 行）—— 这不是「没有同签名的其他 PR」，是这一轮没测到。跨 PR 聚合本次不可用，请手工比对其他 PR 的同类评论。
• ⚠️ 24h 评论账本没读完（超过 5 页仍未读到窗口尽头），所以上面的「不同 PR 数」是下界，不是全量。

历史信号：

• 本 PR 过去 24h 无队列失败记录（首次）。
• 过去 24h 队列共有 70 个失败构建（不含本次）。

分诊清单：

1. 失败测试在本 PR 改动的包里 → 真回归，修 PR。
2. 失败测试与本 PR 无关 → 看上面的「跨 PR 相同签名」；已有汇总 issue ⇒ flaky/环境问题实锤，去那张 issue 上谈，修好前重排只会再烧一轮全队列。
3. 两者都不是 → 可能与同组 PR 语义冲突；等前面的 PR 落地或失败出队后再重排一次即可，不要连续重排。

---

Generated by Claude Code · merge-queue-triage workflow (#4859)

[os-sam]

Queue-kick triage (spec seat, session_01RadETjNRLALFLhFA3xehZP), per the merge-queue-triage checklist above:

Verdict: infra, not this PR — re-queue once, deliberately delayed. The two failed queue jobs both died at the Verify pnpm version setup step (corepack's pnpm download crashing with Node undici AssertionError [ERR_ASSERTION]: assert(!this.paused)) — before any test body ran, so the triage comment's own assertion-vs-timeout discriminator does not apply (the assertion is in the toolchain fetch, not in a test). The same signature killed Dogfood Regression Gate (3/3) and Test Core (6/6) on PR #11290's PR-side run in the same 08:59–09:00Z window — four jobs, two unrelated diffs, one npm-registry blip. Neither diff touches pnpm, corepack, or CI setup.

Disposition: this is the died-before-any-test-body re-run case. To avoid burning another full-queue build into the same blip, the re-queue happens on the seat's next patrol (~09:20Z) rather than immediately — one re-queue, and a second failure of any shape is treated as real. No changes will be pushed for this signature.

---

Generated by Claude Code