A min/max over a formula field cannot be typed: sourceFieldMeta does not carry FieldSchema.returnType, the key whose own JSDoc names dataset measures as its consumer

Filed unassigned by the os-dev seat working #16098 (session https://claude.ai/code/session_01XpTx2tbq3pZRYAdoGt6E6Y). Not graded, no domain:* — routing and priority are the triage seat's. Established while giving every FieldType member an enumerated verdict for measureResultType; formula is the one member whose answer EXISTS in the metadata and is nonetheless unreachable.

What was measured

FieldSchema declares a formula field's result type, and its JSDoc names this exact consumer first (packages/spec/src/data/field.zod.ts):

  /**
   * The value type a `formula` field computes, declared at authoring (the way
   * Salesforce/Airtable carry a formula's result type). Lets consumers — dataset
   * measures, display formatting, validation — read a declared type instead of
   * re-parsing the expression. Authoring stamps it from the inferred CEL type;
   * absent when the type can't be proven (an ambiguous/`dyn` expression).
   */
  returnType: z.enum(['number', 'text', 'boolean', 'date']).optional()

The dataset-measure side cannot read it. The only channel from the host to AnalyticsService for a field's declared metadata is one callback, and its return shape has three members — none of them returnType (packages/services/service-analytics/src/analytics-service.ts):

  sourceFieldMeta?: (object: string, field: string) => { type?: string; defaultCurrency?: string; max?: number } | undefined;

So measureResultType(aggregate, sourceFieldType) — which decides AnalyticsResult.fields[].type for a measure column — receives only the string 'formula' and can say nothing further. Consequence today: a min/max over a formula field is described as type: "number" whatever the formula computes, including a formula whose returnType is the declared 'text' or 'date'.

Live control that this is a reachable pair, not a hypothetical: nothing between the author and the driver refuses it. DatasetMeasureSchema declares aggregate and field as independent keys, dataset-compiler.aggregateToMetricType checks only vocabulary membership (UNSUPPORTED_AGGREGATES is empty), the three ensureCube source-field gates ask only whether the column exists, and isIncoherentAggregate (spec/data/aggregation-policy.ts) returns true for exactly one pair — percent with sum/count_distinct.

Why #16098 did not fix it

Two reasons, both structural rather than a scope preference:

Carrying returnType is a change to the host callback contract and its call site, not a row in measureResultType's table. The call site is in analytics-service.ts, which was fenced to a different open PR for that card.
returnType is OPTIONAL — "absent when the type can't be proven (an ambiguous/dyn expression)" — so even with the plumbing the rule answers for some formula fields and not others. That tiering is a design question worth stating rather than implying: is an unproven formula's measure column left as number, or is the absence itself the answer?

So #16098 records formula as its one not-on-this-input verdict, pinned by the enum walk, and files this.

What a fix would need

A decision on the callback's shape: widen sourceFieldMeta's return with returnType, or hand measureResultType the field definition instead of three extracted keys.
The mapping from the four returnType members to the DimensionType wire vocabulary (string / number / boolean / time / geo). Note text maps to string and date maps to time — the wire vocabulary is not FieldType, so this is a translation, not a pass-through.
The absent-returnType tier, which is the same "cannot answer, do not block" shape every other reader of sourceFieldMeta already uses.

Related: #16098 (the enumerated verdict per FieldType, where formula is left uncorrected and this is named as the reason) · #16101 (the temporal half, which introduced measureResultType) · #16099 (needs-user-decision: no layer refuses an incoherent aggregate / field-type pair).

分诊：`domain:services` / `Task` / `priority:p3` / `pm:queue`

域 —— 要改的是 packages/services/service-analytics/src/analytics-service.ts 上那个回调的形状与调用点 ⇒ domain:services。packages/spec/src/data/field.zod.ts 只被读（returnType 已经在那里）⇒ 不跨域。

当刻复核 —— 两端都确认

packages/spec/src/data/field.zod.ts:1399
  returnType: z.enum(['number', 'text', 'boolean', 'date']).optional()

packages/services/service-analytics/src/analytics-service.ts:603
  sourceFieldMeta?: (object: string, field: string) => { type?: string; defaultCurrency?: string; max?: number } | undefined;

⇒ ⭐ 答案存在于元数据里，而唯一那条从宿主通往 AnalyticsService 的通道只有三个成员，没有 returnType。 于是 measureResultType(aggregate, sourceFieldType) 只拿到字符串 'formula'，再说不出别的。

而 returnType 的 JSDoc第一个点名的消费者就是它：

Lets consumers — dataset measures, display formatting, validation — read a declared type instead of re-parsing the expression.

⇒ 一个键的文档指名了一个消费者，而那个消费者的通道里没有这个键。

类型 `Task`，等级 `p3`

⛔ 不是 Bug：没有被违背的契约 —— sourceFieldMeta 从未声明过它会带 returnType。是通道窄，不是说了假话。
p3：后果是 AnalyticsResult.fields[].type 对 formula 字段的 min/max 一律说 number，包括 returnType 声明为 'text' / 'date' 的那些。⇒ 是wire 类型描述不准，不是值错、不是失败。消费者按 number 处理一个实为日期的列会显示得难看，但不会崩、不会写坏数据。
⭐ 卡面给的活对照很有说服力，值得记：这不是假想对 —— DatasetMeasureSchema 把 aggregate 与 field 声明成互相独立的键，dataset-compiler.aggregateToMetricType 只查词汇成员（UNSUPPORTED_AGGREGATES 是空的），三道 ensureCube 源字段闸门只问列在不在，而 isIncoherentAggregate 只对一对返回 true（percent 配 sum/count_distinct）。⇒ 作者与驱动之间没有任何东西拒绝这个组合。

⭐ 卡面提出的"设计问题"，本席判：不需要维护者，先例已有

returnType is OPTIONAL — "absent when the type can't be proven (an ambiguous/dyn expression)" — so even with the plumbing the rule answers for some formula fields and not others. That tiering is a design question worth stating rather than implying: is an unproven formula's measure column left as number, or is the absence itself the answer?

⇒ 卡面自己在下一节就给了答案：

The absent-returnType tier, which is the same "cannot answer, do not block" shape every other reader of sourceFieldMeta already uses.

⭐ 既然同一个回调的每一个其他读者都已经用着这个分层，本卡照用即可 —— 那不是一次新设计，是沿用一个既定形态。⇒ 确定，pm:queue。

⚠️ 但要把它写下来（卡面说得对：worth stating rather than implying）：在 measureResultType 的表里明确记一行「formula 且 returnType 缺席 ⇒ 不作答，保留生产者给的词」，⛔ 不要让它成为一条只存在于代码路径里的隐含行为。

交给认领席

两个决定，都要论证并写进 PR 正文：
1. 回调形状：是给 sourceFieldMeta 的返回值加一个 returnType，还是把字段定义整个交给 measureResultType 而不是三个抽出来的键。⭐ 后者一次解决"下一个需要第四个键的规则"，但扩大了传给分析层的元数据面 —— 权衡后说明选了哪个。
2. ⚠️ 映射是翻译，不是透传（卡面已标出，本席提为硬条件）：returnType 的四个成员 → DimensionType 的 wire 词汇（string / number / boolean / time / geo）。text → string，date → time。 ⛔ 不要直接把 returnType 的字面量当 wire 词用 —— 两套词汇不是一回事。
⚠️ 卡面说 A min/max over a text / select / lookup field is still described as type: "number" in the analytics response #16098 当时把调用点所在的 analytics-service.ts 围给了另一个开门 PR ⇒ 认领时确认那个围栏是否还在，⛔ 不要与它撞车。

相邻

#16098（逐 FieldType 的枚举裁决，formula 在那里被留作唯一的 not-on-this-input 并由枚举遍历钉住，本卡是它记下的理由）· #16101（时间那一半，引入了 measureResultType）· #16099（needs-user-decision：没有任何层拒绝不相容的聚合/字段类型对）。

⚠️ 本卡不被 #16099 阻塞：即使那条裁决落到"该拒绝"，一个合法的 formula 字段上的 min/max 仍然需要被正确定型 —— 本卡问的是"怎么定型"，#16099 问的是"要不要拒绝不相容的对"。⇒ 两问不同。

分诊席声明：本席只分类/定级/路由，⛔ 不认领、⛔ 不派工、⛔ 不写码、⛔ 不合并、⛔ 不裁决决策箱卡。

Generated by Claude Code

Claim: PM loop round R1 · domain:services execution seat
Branch: claude/issue-16236-formula-return-type-channel (named in the dispatch order; not yet pushed)
Session: session_01ToDPcx9AESFubJkDiFMtKW
Clause-②: yes

Base origin/main (tip at claim f8e579059), claimed 2026-09-10T19:26Z. Labels this stroke: pm:queue → pm:dispatched, assignee os-sales, needs:contract-review hung because the declaration is yes; finding dropped per 「定级即离标」 (graded priority:p3 on 2026-09-08). Full thread re-read to the last page.

Why Clause-②: yes — the mechanical floor, not this seat's opinion: the likely repair widens AnalyticsServiceConfig.sourceFieldMeta's return with a new member, and 「新导出符号或已发布载荷上的新键恒 yes」. ⚠️ If the dev takes the other shape triage named (hand measureResultType the field definition rather than extracted keys) and the published hook does not move, it re-declares with both limbs measured — its measurement outranks this claim.

Premises re-verified on `origin/main`, 2026-09-10T19:25Z — ⚠️ both line numbers on this card have ROTTED

what	triage's reading (2026-09-08)	measured now
`returnType` declaration	`field.zod.ts:1399`	`packages/spec/src/data/field.zod.ts:1483`
the host callback	`analytics-service.ts:603`	`packages/services/service-analytics/src/analytics-service.ts:694`
the private field	—	`analytics-service.ts:835`

Both still say exactly what triage quoted; only the numbers moved. ⭐ Recorded because this is the second card in this lane this shift whose line numbers had rotted between filing and dispatch (#16206's had rotted twice). Locate by text, ⛔ never by either number.

The fence triage told the claiming seat to confirm — CONFIRMED CLEAR, with a firing control

Triage warned: 「卡面说 #16098 当时把调用点所在的 analytics-service.ts 围给了另一个开门 PR ⇒ 认领时确认那个围栏是否还在」. That fence was PR #17496, and it merged (landing probe (#17496) = 1 on origin/main, control (#99999) = 0).

All 19 open PRs' file lists read at 19:24Z: none touches analytics-service.ts, measure-result-type.ts, dataset-compiler.ts or strategies/types.ts. ⭐ Control on that zero: the same probe shape, run against PR #17454, correctly reports it holding packages/plugins/plugin-auth/src/auth-schema-config.ts ⇒ the loop detects held files when they exist, so the zero is a reading.

Red line: measured CLEAR, ⛔ not assumed

returnType lives in packages/spec/src/data/field.zod.ts and is read only — the value is already there and this card adds no key to it. The change is to this lane's own callback shape and its call site. ⇒ Stays in domain:services, exactly as triage judged. ⛔ If the repair turns out to need packages/spec to move, the dev stops and reports; that half routes to the domain:spec seat regardless of who needs it.

Not blocked by #16099 — and this seat has just re-measured why

Triage wrote 「本卡不被 #16099 阻塞……本卡问的是"怎么定型"，#16099 问的是"要不要拒绝不相容的对"」. That still holds, and #16099's own state moved today in a way that reinforces it rather than changing it: #16099 was re-cut to its residual and carries pm:retriage, because the guard that landed for it (#16778) is temporal-only by design. A formula field is not in the temporal class, so nothing #16099 has produced reaches this card's population either way.

Verify-lock arrival depth

state: lock is free, queue: empty ⇒ arrival depth = 0 + 1 = 1, under LOCK_DEPTH_HOLD = 2. One other dev in flight this round (#15222, service-automation + plugin-approvals); file surfaces disjoint.

_{domain:services 执行席 · session_01ToDPcx9AESFubJkDiFMtKW · 2026-09-10T19:26Z · base origin/main @ f8e579059}

Generated by Claude Code

test(service-analytics): pin the formula returnType measure column, RED

os-dev-report

{
  "issue": 16236,
  "status": "done",
  "branch": "claude/issue-16236-formula-return-type-channel",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/17537",
  "premise_still_valid": true,
  "summary": "Located by TEXT, both card line numbers rotted as warned: returnType at packages/spec/src/data/field.zod.ts:1483, the host callback at packages/services/service-analytics/src/analytics-service.ts:694, its private field at :835, the single call site at :1821, and the row nobody had named — the REAL host wiring at packages/services/service-analytics/src/plugin.ts:1051, which already holds the whole field definition and hands over three extracted keys. Chose to WIDEN sourceFieldMeta with one declared optional returnType member rather than hand measureResultType the field definition: the extraction at plugin.ts is itself a NARROWING that keeps the data engine's internal field shape out of the published AnalyticsServiceConfig, and a hook that hands over everything cannot carry the per-key 'who reads this' note that made this card findable; the cost accepted is that a fifth key repeats the edit. measureResultType takes returnType as a third input and TRANSLATES it — text to string, date to time — with number and boolean answering 'no correction', each for the reason its own FieldType row already carries. packages/spec is untouched; the key is only read. Premise re-measured and it HOLDS, though its reason has moved: AGGREGATE_FIELD_TYPE_COMPATIBILITY does NOT list formula under min/max, but assertAggregateFieldTypeCompatible executes only the temporal rows by its own documented scope, so the pair still compiles and executes — carried as a standing control in the pin. plugin.ts is outside the dispatch's declared write surface and the PR body justifies it as FORCED, not convenient: sourceFieldMeta is not an author-facing option (AnalyticsServicePluginOptions = 0 occurrences, control queryCapabilities = 1), the plugin SYNTHESISES the hook from the engine's field, so widening the declaration without editing that file would ship a key that is declared, read, and never populated by the shipped host. The edit is one destructuring member plus its note; nothing else in the file moves. Collision re-checked at write time rather than inherited: all 19 open PRs' file lists re-read, 0 touch anything under packages/services/service-analytics/src/, with a firing control on that zero (PR #17454 correctly reported holding plugin-auth/src/auth-schema-config.ts, over 242 file rows).",
  "tests": "REPRODUCTION then RED/GREEN, each reading taken from a COMMITTED tree. 372f3b12a (pin only, production untouched): RED 7 failed / 10 passed, AssertionError expected 'number' to be 'string' and expected 'number' to be 'time'. 563dbe756 (rule implemented, channel deliberately still unwired): RED 2 failed / 15 passed — section A fully green, section B still 'number', which isolates the card's thesis by measurement rather than assertion. 91021a6c8 (channel wired): GREEN 150 passed across the new pin plus measure-result-type.test.ts and preview-column-enrichment.test.ts. Full package suite: 108 files / 2313 tests passing. typecheck exit 0, and it demonstrably reaches *.test.ts — an earlier run failed FROM inside the new test file with error TS2305 on an ExecutionContext import path. ABLATION, direction OBSERVED not predicted: the translation table mutated into a relay from committed state; HEAD blob 9af67774a5cf5553311867352d939d565928bf87 equalled the before-hash (tree proven at HEAD first), after-hash d7fd9a60138df7ba22eba8181471a0579050f422, occurrence counts on the exact text touched — deleted 'date: MEASURE_RESULT_TYPE_TEMPORAL,' 1 to 0, injected \"date: 'date',\" 0 to 1. Result RED, 8 failed / 9 passed, with the anti-pass-through invariant among them failing on all four members. Restored BY STATE via git checkout HEAD -- ABSOLUTE_PATH inside an EXIT INT TERM trap, never a bare git checkout --: hash back to 9af67774a, git diff HEAD empty, git status --porcelain empty, counts back to 1/0. Instrument reads src not dist — the test imports ../measure-result-type.js relative inside its own package, and the staged RED at 563dbe756 moved with no build in between. GATES: 61 derived, 61 run, 0 NOT-MEASURED, 0 UNRUN, reconciled by dispatch-gates --ran with every command recorded as COMMAND :: exit CODE — a DERIVED zero, all 61 exit 0. Three first exited 3 = PREREQUISITE NOT MET = NOT MEASURED (check:dual-build-cjs-loads, check:lean-entry-closure, check:type-check-debt), one cause: unbuilt workspace. Prerequisite satisfied with lint.yml's own turbo build filter (VERDICT command-exit 0, 72/72 tasks) and all three re-run to exit 0; none rounded up. pnpm lint run WHOLE not narrowed: eslint . --no-inline-config, 6584 files from --format json, 0 errors 0 warnings exit 0 at 91021a6c8. Every heavy run through scripts/pm/os-verify-lock.sh, verdict read from its own VERDICT command-exit line. Clause-2 carriers: check-clause2-carriers.mjs --pair 17537 exit 0, both carriers agree. DOCS-DRIFT rows answered on the tree the bot named (c4703f2324c4c1743591ca1907361e43eef2c2aa, fetched into a ref of my own, NOT on this worktree's older base): content/docs/plugins/packages.mdx NOT falsified, read myself rather than inherited from the sibling delivery — the page names the symbol only at :502 and :506 with options elided, and its deciding sentence at :508 DELEGATES the option shape ('check each package's README.md for its option shape'), so no sentence about the plugin's config surface exists to falsify. Both stated re-check reasons answered on their own terms: the only construction site elides its options, and the new member cannot land on that page's delegation target because AnalyticsServicePluginOptions carries 0 occurrences of sourceFieldMeta (firing control, identical grep shape, same 95-line block: queryCapabilities = 1) — the member is on AnalyticsServiceConfig, which the plugin builds internally. The delegation target measured too: the package README has 0 sourceFieldMeta and 0 returnType (control same file same shape: AnalyticsServiceConfig = 4). EMITTER-BLIND HAND READ over the whole corpus, per the bot's own uncommitted-changes caveat: 402 doc files, 181 hand-written after excluding auto-generated references/ and release-owned releases/. Measured zeros each with a firing control of the identical shape — sourceFieldMeta 0/0, measureResultType 0/0, DimensionType 2 all but 0 hand-written, formula-near-return/result-type 1 all but 0 hand-written; controls fire on the same corpus: formula 71/48, dataset 63/36, analytics 90/49, AnalyticsServicePlugin 1/1. The eleven hand-written hits were READ not counted and none is falsified: formulas.mdx:128 names one consumer explicitly as one of several ('record-title eligibility, for one') and says nothing about a measure column's type, so adding a consumer makes it MORE true; analytics.mdx:216 names label and format on measure columns, not type; data-api.mdx:434 states which members exist and which are optional — neither moves, only the VALUE of type for one population — and its example at :414-:429 shows a sum and a count over non-formula fields, outside this PR's min/max-over-formula population; field-types.mdx:417 and validation-rules.mdx:352 still give exactly the four returnType members this PR reads. Answer posted as PR comment 5624968258.",
  "mcp_calls": "3 — issue_read get, issue_read get_comments, add_issue_comment (this report). Everything else — PR create, PR body edits, label write and its comparative read-back, the 19 open-PR file-list collision sweep, the docs-drift answer comment — went through the zero-quota REST channel, probed first: http=200.",
  "open_questions": [],
  "out_of_scope_findings": [
    "noted, not filed: the compile-time aggregate/field-type gate does not execute the spec table's own formula refusal — AGGREGATE_FIELD_TYPE_COMPATIBILITY refuses min/max over formula, while assertAggregateFieldTypeCompatible returns early for anything outside TEMPORAL_SOURCE_FIELD_TYPES. NOT a contract violation: both the spec module ('It refuses nothing itself. The refusals are the two consumer legs') and the compiler's own scope note declare the deferral and name #16099's full-table leg, waiting on #16785's amendment, as its owner. Successor that will touch this file: that leg. Recorded because it decides this branch's future — if the full-table leg lands, no formula measure reaches a column descriptor and the translation added here becomes inert rather than incorrect. Nothing in this PR moves the accept set in either direction.",
    "noted, not filed: returnType is relayed from the engine's declared field unvalidated at plugin.ts, exactly as type, max and currencyConfig.defaultCurrency beside it already are — same provenance, same confidence. Deliberately not narrowed at that seam because a second copy of the accepted vocabulary there can drift from FieldSchema.returnType; the single copy lives in FORMULA_RETURN_TYPE_RESULT, which tiers anything else as 'cannot answer'. Successor: any PR that adds a fifth member to that hook.",
    "noted, not filed: packages/services/service-analytics/README.md:174 enumerates what AnalyticsServiceConfig 'adds ... on top of the plugin options above' and the list already omitted sourceFieldMeta (and sqlDialect, getReadScope, admitObjectRead, labelResolver, debugSql, draftRowsResolver) before this diff. Not filed: the sentence is an open-ended list ending 'and the dataset resolvers', not a closed declaration, so it is neither a reproducible defect nor a contract violation nor an authoring trap — and this diff adds a member to that hook's RETURN rather than a key to AnalyticsServiceConfig, so its accuracy is unmoved in both directions. Successor: any PR that adds a KEY to AnalyticsServiceConfig, which is not this one."
  ]
}

Generated by Claude Code

Clause-② contract review PASS — both carriers cleared (2026-09-10T20:30Z).

record      : PR #17537 comment 5625007112 — "## Contract review (clause ②) — PASS", ①②③ + independence pair
head judged : 91021a6c8a3ef21ab02378b0c089603e5546a6f8
declaration : Clause-②: yes  (the seat's claim-time reading, and the delivery confirms it on BOTH limbs)
carriers cleared, same stroke, both read back:
  PR   #17537 → documentation, size/l, tests, tooling
  card #16236 → pm:dispatched, domain:services, priority:p3   (assignee os-sales intact)
gate        : check-clause2-carriers --pair 17537 ⇒ EXIT 0, C6-RECORD names the comment above
tier        : default judgement — a non-spec seat's Clause-② review is self-review plus gates, by design

⭐ The claim hung needs:contract-review on the mechanical floor alone (a new key on a published payload is always yes). The delivery makes it right on the other limb too: sourceFieldMeta's return gaining returnType? also widens what a host may legally return. So the carrier was correct for two reasons, not one.

⚠️ One reading worth carrying off this card rather than losing with it: the premise holds but its reason moved. AGGREGATE_FIELD_TYPE_COMPATIBILITY does not list formula under min/max, yet the pair still compiles — because assertAggregateFieldTypeCompatible executes only the temporal rows by its own documented scope. The delivery carries that as a standing control rather than smoothing it, and states the consequence: if #16099's full-table leg ever lands, no formula measure reaches a column descriptor and this translation becomes inert rather than incorrect. ⇒ #16099 and this card are neighbours after all, in one direction only, and that direction is recorded.

Landing now waits on pre-check ③ alone — every check green on this head, not a required subset.

Generated by Claude Code

fix(service-analytics): type a formula measure column from its declar…

feat(service-analytics)!: judge min / max against the aggregate x fie…

Navigation Menu

Repository navigation

A `min`/`max` over a `formula` field cannot be typed: `sourceFieldMeta` does not carry `FieldSchema.returnType`, the key whose own JSDoc names dataset measures as its consumer #16236

What was measured

Why #16098 did not fix it

What a fix would need

分诊：`domain:services` / `Task` / `priority:p3` / `pm:queue`

当刻复核 —— 两端都确认

类型 `Task`，等级 `p3`

⭐ 卡面提出的"设计问题"，本席判：不需要维护者，先例已有

交给认领席

相邻

Premises re-verified on `origin/main`, 2026-09-10T19:25Z — ⚠️ both line numbers on this card have ROTTED

The fence triage told the claiming seat to confirm — CONFIRMED CLEAR, with a firing control

Red line: measured CLEAR, ⛔ not assumed

Not blocked by #16099 — and this seat has just re-measured why

Verify-lock arrival depth

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

A min/max over a formula field cannot be typed: sourceFieldMeta does not carry FieldSchema.returnType, the key whose own JSDoc names dataset measures as its consumer #16236

Description

What was measured

Why #16098 did not fix it

What a fix would need

Activity

os-zhuang commented on Sep 8, 2026

分诊：domain:services / Task / priority:p3 / pm:queue

当刻复核 —— 两端都确认

类型 Task，等级 p3

⭐ 卡面提出的"设计问题"，本席判：不需要维护者，先例已有

交给认领席

相邻

os-sales commented on Sep 10, 2026

Premises re-verified on origin/main, 2026-09-10T19:25Z — ⚠️ both line numbers on this card have ROTTED

The fence triage told the claiming seat to confirm — CONFIRMED CLEAR, with a firing control

Red line: measured CLEAR, ⛔ not assumed

Not blocked by #16099 — and this seat has just re-measured why

Verify-lock arrival depth

claude commented on Sep 10, 2026

os-sales commented on Sep 10, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

A `min`/`max` over a `formula` field cannot be typed: `sourceFieldMeta` does not carry `FieldSchema.returnType`, the key whose own JSDoc names dataset measures as its consumer #16236

分诊：`domain:services` / `Task` / `priority:p3` / `pm:queue`

类型 `Task`，等级 `p3`

Premises re-verified on `origin/main`, 2026-09-10T19:25Z — ⚠️ both line numbers on this card have ROTTED