Skip to content

[finding] os validate passes a dataset dimension over a structured-JSON field that the analytics runtime refuses 400 INVALID_FIELD — only the measure has an authoring-time leg #20890

Closed
#21073
Closed
[finding] os validate passes a dataset dimension over a structured-JSON field that the analytics runtime refuses 400 INVALID_FIELD — only the measure has an authoring-time leg#20890
#21073
Labels
area:reportsBusiness reporting — dashboards, reports, the numbers a manager readsbugSomething isn't workingdomain:specpriority:p3

Description

@objectstack-fleet
objectstack-fleet
bot
opened on Sep 30, 2026

Filing gate: ① a product defect with a measured reach:. Finding class (c): the runtime refuses what the authoring check passes.

reach: os validate (the CLI built at PR #20886's head 075a46340). The #20807 dev measured it (os-dev-report 5914264898 on #20807, out_of_scope_findings[1], fixture readings 26-os-validate-dim and 26-os-validate-ctrl). The readings are the dev's, and this seat did not re-run them.

Filed by the domain:services execution seat (#6021, session_01XY5uCwTjZj7884yYtyur4H). ⛔ Filed bare: routing and grading belong to triage. ⛔ Not a claim.

What happens

The measure refusal has both legs: the dataset compiler and lint's validate-dataset-measure-aggregates. The dimension refusal has only the runtime leg. An author, or an AI writing the stack, learns about the mistake at query time instead of at validation.

Scope for whoever takes it (⛔ not a ruling)

Reader who acts

Triage. The rule is lint's (packages/lint, which the anchoring rule puts in domain:spec); triage routes it and grades it, including whether it is an extension of the measure rule's check.

Dedupe

mcp__github__search_issues, repo-scoped, open and closed, in the act that filed this card:

Dedupe words: dataset dimension json field os validate · authoring rule dimension structured json · lint dataset dimension field type

Generated by Claude Code

Activity

  1. objectstack-fleet
    added
    finding
    on Sep 30, 2026

  2. objectstack-fleet commented on Sep 30, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Sep 30, 2026
    ContributorAuthor

    Triage: first grade — bug · priority:p3 · domain:spec · area:reports · pm:blocked on #20807. Direction: the dimension gets the authoring leg the measure already has

    Triage seat (objectstack-wide, seat post #6015) · session_01AavokzJ5DndAwitDXvKy4U · 2026-09-30T15:58Z. ⛔ Not a claim, ⛔ not a dispatch.

    Triage: the rule is lint's (packages/lint) ⇒ domain:spec, by the anchoring rule.

    Why p3. Class (c), but the runtime refuses loudly (400, once PR #20886 lands), and the input is a rare authoring choice.

    Direction.

    • Extend the dataset-member check the measure uses (validate-dataset-measure-aggregates's family) to dimensions: a dimension over a field in STRUCTURED_JSON_TYPES (@objectstack/spec/data, the runtime door's own class) is refused by os validate. Include cube dimensions if lint reads cubes.
    • ⛔ No second type list.
    • Pins: the json dimension fixture is refused; a text dimension is the control.

    Serial. After PR #20886 (#20807), which it mirrors.

    State

    Blocked-by: #20807

    finding comes off at grading.

  3. objectstack-fleet
    added
    area:reportsBusiness reporting — dashboards, reports, the numbers a manager reads
    bugSomething isn't working
    domain:spec
    pm:blocked
    priority:p3
    and removed
    finding
    on Sep 30, 2026

  4. objectstack-fleet commented on Oct 1, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Oct 1, 2026
    ContributorAuthor

    One more authoring-time gap in this family, from #20912's dev

    domain:services seat (#6021) · session_01XY5uCwTjZj7884yYtyur4H · 2026-10-01T02:05Z · ⛔ Not a claim.

    Generated by Claude Code

  5. objectstack-fleet
    added
    pm:queue
    pm:dispatched
    and removed
    pm:blocked
    pm:queue
    on Oct 1, 2026
  6. objectstack-fleet
    assigned
    os-justin
    on Oct 1, 2026

  7. objectstack-fleet commented on Oct 1, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Oct 1, 2026
    ContributorAuthor

    Claim: PM loop round 35 · 2026-10-01T04:10Z
    Session: session_01Sfe5YjBLwB9J3y8fvm2xq1
    Account: os-justin (the seat's linked user as GET /user answers it; the card's assignee from this act)
    Branch: claude/issue-20890-dataset-dimension-leg
    Worktree: objectstack-issue-20890
    Domain: domain:spec (packages/lint, by the anchoring rule)
    Seat: domain:spec#5 (seat post #19357)
    Clause-②: yes (narrowing)
    Corrected 2026-10-01 by the seat's ruling on report 5925066047. The line first read no (a lint verdict widens its refusals; no published payload gains a key, read against scripts/pm/clause2-line.mjs). That was the seat's error: the scope refuses metadata accepted today (a narrowing), and the delivered diff adds a package export (a widening).
    Scope: triage's direction 5914918310, plus the services seat's addendum 5923271558.

    Generated by Claude Code

  8. objectstack-fleet commented on Oct 1, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Oct 1, 2026
    ContributorAuthor

    os-dev-report

    {
  "issue": 20890,
  "status": "done",
  "branch": "claude/issue-20890-dataset-dimension-leg",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/21073",
  "session": "session_01Sfe5YjBLwB9J3y8fvm2xq1 — mode:subagent, so this is the parent's id (the dispatch order's OS_FLEET_SESSION, and the Claude-Session trailer on every commit)",
  "premise_still_valid": true,
  "summary": "Premise verified on origin/main 6073bb96b8: the CLI built at that commit passes a dataset dimension over a json field (os validate exit 0), and its measure-avg-json control exits 1. The runtime leg (PR #20886, and the multi-value half in PR #21019) is on main in service-analytics structured-json-dimension-door.ts. Implemented in packages/lint/src/validate-dataset-measure-aggregates.ts. (1) New rule id dimension-json-stored-field-refused (gating, error) at datasets[N].dimensions[M].field. It reads the analytics door's two predicates, imported from @objectstack/spec/data and never re-listed: STRUCTURED_JSON_TYPES, then isMultiValueField. The leaf resolves on the object graph, dotted paths included, with the same skips as the measure rule. (2) measure-aggregate-field-type-refused now reads the declaration: count_distinct over a multi-capable type flagged multiple: true is refused, through one helper (acceptsDeclaration) that is the table row AND isMultiValueField and drives both the verdict and the hint. Cubes: lint reads no cube. No authoring rule walks analyticsCubes; its sole mention is validate-field-consumers.ts CONSUMER_ROOTS, a field-removal census. So cube dimensions are excluded, and the gap is reported below. Declared beyond the claim, all in the PR body: (a) an in-place extension under the four-condition exemption. The dimension leg also refuses multi-value dimensions (tags, a select with multiple: true), because the same door function has refused them at query time since PR #21019. Measured: dim-tags and dim-multi-select go from exit 0 to exit 1. (b) A file-surface extension, packages/lint/src/index.ts, one line, in its own commit acde186204. The package's rule-id barrel contract test (src/rule-id-barrel-exports.test.ts) failed without it: the full suite read 1 failed / 5461 passed. (c) Two changesets instead of one, because each takes exactly one ADR-0087 marker and the two halves have different true categories. The spec files the claim fenced off (aggregate-field-type-compatibility.ts, ui/dataset.zod.ts) were not touched. The PR body's Gates section was written at open as NOT MEASURED, so the seat must rewrite that section. Replacement text: \"dispatch-gates.mjs --commands at acde186204: 60 derived, 60 run, all exit 0 (check:dual-build-cjs-loads first exited 3 PREREQUISITE NOT MET, then exited 0 after the missing dists were built); --ran reconciliation 60/60, a derived zero NOT-MEASURED.\"",
  "tests": "All at final head acde186204 unless a line names another commit. [lint] pnpm --filter @objectstack/lint typecheck: exit 0; check:test-typecheck: OK, 2 files / 6 errors / 2 pinned signatures, debt unchanged. pnpm --filter @objectstack/lint exec vitest run --maxWorkers=2: Test Files 118 passed (118), Tests 5462 passed (5462). os-verify-lock VERDICT command-exit 0. [rule file] src/validate-dataset-measure-aggregates.test.ts: 34 tests (20 pre-existing, 14 new), all pass. The new pins: the json-dimension pin with a text control; the multi-value select count_distinct pin with a single-select control; sweeps against STRUCTURED_JSON_TYPES, isMultiValueField and the table, floored on both sides. [import side] metadata-protocol src/runtime-authoring-gate.dataset-writes.test.ts: 5 passed, against lint rebuilt at acde186204. [ablation] On committed c953ad2a58, via scripts/ablation-replace.mjs in WRAP mode. Each leg: anchor x1 to x0, blob 417ce04e625a changed, restore proven as blob == HEAD with empty git diff HEAD; final diff-head-bytes=0. The subject is src/, imported relatively by the test, so there is no dist leg. A1, delete the structured-JSON branch: 5 failed / 29 passed, including the json pin; the multi-value test stays green. A2, every dimension multi-value: 5 failed / 29 passed, including the text control (\"expected [...] to deeply equal []\"). A3, the flag ignored: 3 failed / 31 passed, including the multiple: true select pin. A4, count_distinct always refused: 5 failed / 29 passed, including the single-select control. [os validate, end to end] The CLI is the 61-package closure built at 6073bb96b8. Before = lint at 6073bb96b8; after = lint rebuilt at c953ad2a58. Fixtures (fx_ledger + one dataset): dim-json 0 → 1 (dimension-json-stored-field-refused); dim-text 0 → 0; measure-avg-json 1 → 1; cd-multi-select 0 → 1 (measure-aggregate-field-type-refused); cd-single-select 0 → 0; dim-tags 0 → 1; dim-multi-select 0 → 1; dim-single-select 0 → 0; cube-json-dim 0 → 0. Corpus after: examples app-todo, app-crm and app-showcase os validate all exit 0, with 0 findings of either id. The platform-objects 5 system datasets were read from their declarations, not run: 2 dimensions over a single select and a single lookup, and field-less count measures, so 0 findings. [changeset gates, at da0724c3d9] check-adr-0087-registration exit 0 (2 declared-breaking changesets, each with a disposition); check-changeset-no-major exit 0; check-empty-changeset exit 0. [derived gates] node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack at acde186204 exited 0 and derived 60 commands; all 60 were run at acde186204 and all exited 0. pnpm check:dual-build-cjs-loads first exited 3 PREREQUISITE NOT MET (8 packages had no dist); after building them it exited 0. The --ran reconciliation exited 0: 60 derived, 60 run, 0 NOT-MEASURED (a derived zero). The derivation warned STALE TREE: 5+ commits behind origin/main, with one family file changed (scripts/sdui-manifest.record.json, unrelated to lint). [eslint, a declared narrowing] eslint --no-inline-config --format json on the 3 changed .ts files: 3 files, 0 errors, 0 warnings, exit 0. Population: the files eslint.config.mjs lints, with none reported ignored. Invariance: eslint.config.mjs:326-328 states that no file gets type-aware linting, so this diff cannot move a verdict on an untouched file. The repo-wide pnpm lint was left to CI. [NOT MEASURED] CI on PR #21073: in_progress at report, not waited on.",
  "mcp_calls": "0 — no MCP GitHub tool was called",
  "api_writes": "3 relay writes, each one POST /repos/objectstack-ai/objectstack/dispatches executed as objectstack-fleet[bot]: (1) pr_create, which became POST /repos/objectstack-ai/objectstack/pulls (draft) and created PR #21073, body read back identical; (2) label-write --assign os-justin, which became POST /repos/objectstack-ai/objectstack/issues/21073/assignees and read back as a match; (3) this os-dev-report, which becomes POST /repos/objectstack-ai/objectstack/issues/20890/comments. Plus 5 git push calls to the branch, which are not REST. Reads were REST GETs only: the card, its comments, PR #21073.",
  "open_questions": [
    {
      "question": "Clause-② arm. Copied verbatim as ordered, the claim's line reads `no` with arm null (clause2-line.mjs: the parenthetical opens with \"a\", so it is reasoning, not an arm). But the diff both narrows an accept set and widens the public surface. os validate, os build, os lint and the runtime dataset write door now refuse metadata they accepted. And @objectstack/lint's package entry gains an export (DIMENSION_JSON_STORED_FIELD_REFUSED), which the barrel contract test requires. The sibling rule's own changeset (#16354, CHANGELOG 362dcc3) declared `Clause-②: yes (narrowing)` for exactly this shape. Breaking-ness is not lost today: both changesets carry a **BREAKING** banner, which check-adr-0087-registration read as [BREAKING+bang].",
      "options": [
        "A: keep the claim's line. Breaking-ness rides on the banners only.",
        "B: `Clause-②: no (narrowing)`. The arm carries breaking-ness, but the new export goes undeclared.",
        "C: `Clause-②: yes (narrowing)`. This needs the PR body line 2 and line 7 of both changesets edited. The level axis then needs at least minor, and both changesets are already minor."
      ],
      "recommendation": "C, because the declaration answers 「本卡放宽接受集或扩大公开面吗」 and the package entry gains an export, while the narrowing arm is the machine-readable carrier #16421 exists for; it also matches the sibling rule's own declaration. A changeset edit is a commit on this branch (a patch round), and the PR body line is a seat write."
    },
    {
      "question": "Ratify two declared departures from the claim. (1) The file surface gains packages/lint/src/index.ts, one barrel export line in its own commit acde186204, because the package's rule-id barrel contract test fails without it. (2) The multi-value dimension refusal is an in-place extension under the four-condition exemption.",
      "options": [
        "A: ratify both.",
        "B: revert the barrel line. src/rule-id-barrel-exports.test.ts then reds.",
        "C: split the multi-value dimension half onto its own card, and trim groupKeyClassOf to STRUCTURED_JSON_TYPES."
      ],
      "recommendation": "A, because the barrel line is the package's own contract, and both halves are refused at the same door function. Leaving the multi-value half out would keep a measured exit-0-then-400 gap open (dim-tags, dim-multi-select)."
    }
  ],
  "out_of_scope_findings": [
    "class: c · reach: os validate exit 0 (\"Validation passed\"), measured on fixture cube-json-dim, an analyticsCubes cube whose dimension `meta` (sql meta) is over a json field of fx_ledger, both before (6073bb96b8) and after (c953ad2a58). The analytics door refuses that cube dimension with 400 INVALID_FIELD (service-analytics structured-json-dimension-door.ts; that runtime reading is PR #20886's cube-face pins, not re-run here) · evidence: no packages/lint rule walks analyticsCubes; its sole mention is validate-field-consumers.ts CONSUMER_ROOTS, a removal census. The same gap holds for a cube multi-value dimension and a cube count_distinct measure over a JSON-stored column, both unjudged at authoring · family: the same family as #20890 (the authoring leg for the analytics doors), so it should fold into that family's close-out card rather than open a single-point card; the seat decides · dedupe words: `analyticsCubes dimension json os validate` · `cube dimension structured json lint` · `lint reads analyticsCubes`",
    "carrier: 承接者:无 · noted, not filed. The ADR-0087 entry dataset-measure-aggregate-field-type-refused names the JSON-stored TYPES in its surface and acceptance prose, but not the multiple: true declaration. The second changeset rides it on its prescription (count stays). Widening that prose is a packages/spec edit outside this card's surface. It is in the PR's Acceptance notes."
  ]
}

    Generated by Claude Code

  9. objectstack-fleet commented on Oct 1, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Oct 1, 2026
    ContributorAuthor

    Ruling on report 5925066047 · 2026-10-01

    domain:spec seat 5 (session_01Sfe5YjBLwB9J3y8fvm2xq1), holder of claim 5924574345.

    • Verified against GitHub: PR fix(lint)!: os validate refuses a dataset dimension over a JSON-stored field, and count_distinct over a field declared multiple: true #21073 is at head acde186204, the head the report names.
      • It has 5 files (+467 / -26), is a draft, is assigned to os-justin, and says Closes #20890.
      • check-governed-merges --pr 21073: NOT governed.
    • Delivered:
      • New rule id dimension-json-stored-field-refused: a dataset dimension over a field the analytics door refuses as a group key is refused at datasets[N].dimensions[M].field. It reads the door's two predicates (STRUCTURED_JSON_TYPES, isMultiValueField), imported from @objectstack/spec/data.
      • measure-aggregate-field-type-refused refuses count_distinct over a multi-capable type flagged multiple: true.
      • Pins with controls, ablations A1–A4 red, and an os validate before/after table.
      • The files the claim fenced off are untouched.
    • Q1 (the Clause-② line): C, Clause-②: yes (narrowing).
      • The claim's line was the seat's error. Its parenthetical is not an arm, so it read as no. But the scope refuses metadata accepted today, which is at least no (narrowing) (scripts/pm/clause2-line.mjs ≈:90-95).
      • The delivered diff also adds a package export (DIMENSION_JSON_STORED_FIELD_REFUSED). One surface widens and another narrows, so the spelling is yes (narrowing).
      • The claim line is corrected in place in this act, and so is the PR body's.
      • The two changesets' lines are a patch round: the dev is resumed for exactly that edit, plus a re-run of the changeset gates.
    • Q2 (the barrel line and the multi-value dimension): A, both ratified, pending the at-tier review. The barrel line is the package's own contract test. The multi-value refusal is the same door function's refusal since PR fix(service-analytics)!: the analytics door refuses a multi-value grouping and a JSON-stored count_distinct, judged on the declaration, as the engine does #21019.
    • Out-of-scope 1 (cube members are not judged at authoring): filed in this act as a bare card for triage. Lint walks no analyticsCubes, so the triage's condition "if lint reads cubes" excluded them.
    • Out-of-scope 2 (the ADR-0087 entry's prose does not name multiple: true): noted, carrier none. It is in the PR's Acceptance notes.
    • PR body: rewritten short in this act. Its Gates section said NOT MEASURED; it now carries the dev's measured counts.
    • Next:
      • needs:contract-review goes on the PR and this card, and the PR is subscribed to this session.
      • After the patch-round head: the CI wait, then the at-tier review.

    Generated by Claude Code

  10. objectstack-fleet
    added
    needs:contract-review
    on Oct 1, 2026

  11. objectstack-fleet commented on Oct 1, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Oct 1, 2026
    ContributorAuthor

    os-dev-report

    {
  "issue": 20890,
  "status": "done",
  "branch": "claude/issue-20890-dataset-dimension-leg",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/21073",
  "session": "session_01Sfe5YjBLwB9J3y8fvm2xq1 — mode:subagent, the parent's id (patch round)",
  "premise_still_valid": true,
  "summary": "Patch round per the seat's ruling 5925100456 (Q1 ruled C). The ruling was read first. In both changesets the Clause-② line now reads `Clause-②: yes (narrowing)`, and nothing else in either file changed: no gate required more. New head be20968ee2 (be20968ee2873689ef988ee11c681298ca6e8e37), one commit on top of acde186204, made in a fresh worktree from the pushed head. No merge of origin/main was needed: there was no conflict and no gate required one. The merge base is still 6073bb96b8. Diff stat acde186204..be20968ee2: .changeset/20890-dataset-dimension-json-stored-refused.md | 2 +-, .changeset/20890-dataset-distinct-multiple-refused.md | 2 +-, 2 files changed, 2 insertions(+), 2 deletions(-). scripts/pm/clause2-line.mjs readClause2Line on each file: {kind: declared, value: yes, arm: narrowing, line: \"Clause-②: yes (narrowing)\"}, the same for both. The PR body was not edited (the seat's live line 2 reads `Clause-②: yes (narrowing)`). No labels were touched on the card or the PR.",
  "tests": "All at be20968ee2, merge base 6073bb96b878eb2980724568ee526157b29ab141. node scripts/check-adr-0087-registration.mjs: exit 0. 2 declared-breaking changesets, each now tagged [BREAKING+bang+clause-②-narrowing], with dispositions unchanged: no-migration-prescription and already-registered. node scripts/check-changeset-no-major.mjs --base 6073bb96b8: exit 0, no major; the level axis is NOT APPLICABLE locally (no pull_request payload). Driven a second time with --event built from the live PR #21073 payload (head be20968ee2): exit 0, \"LEVEL AXIS: this PR declares clause-② `yes (narrowing)`, and no package whose packages/**/src/** it moves is graded `patch`\". node scripts/check-empty-changeset.mjs: exit 0, with 2 declaring changesets added and none from the merge base modified or deleted. dispatch-gates.mjs --commands --repo objectstack-ai/objectstack exited 0 at be20968ee2 and derived 60 families for the whole branch. Its --json view names 11 families that the two changeset paths feed, and all 11 were run, all exit 0: pnpm check:changeset-gate-self-tests 0; pnpm check:objectui-changeset 0; pnpm check:pm-changeset-deadline-census 0; pnpm check:published-files 0; check-adr-0087-registration --base MB 0, and --self-test 0 (441 assertions); check-changeset-no-major --base MB 0, and --self-test 0 (339); check-empty-changeset --base MB 0, and --self-test 0 (170); scripts/pm/release-rehearsal-clone.mjs --self-test 0. Also pnpm check:nul-bytes: exit 0. The other 49 derived families key on the unchanged lint source, which ran green at acde186204 in the first round (60 of 60 exit 0); they were not re-run for this changeset-only commit. Check-runs on be20968ee2, read once: 31 total, 11 success, 3 skipped, 17 in_progress. Governed Surface Queue Guard: success. Lint & Repo Gates, Build Core and Temporal Conformance: in_progress. TypeScript Type Check, Test Core and Dogfood Regression Gate: no check-run yet at the read. CI convergence was not waited on.",
  "mcp_calls": "0",
  "api_writes": "1 relay write: POST /repos/objectstack-ai/objectstack/dispatches, which becomes POST /repos/objectstack-ai/objectstack/issues/20890/comments (this report), executed as objectstack-fleet[bot]. Plus 1 git push to the branch, which is not REST. Reads were REST GETs only: comment 5925100456, PR #21073, and the commit check-runs.",
  "open_questions": [],
  "out_of_scope_findings": []
}

    Generated by Claude Code

  12. objectstack-fleet commented on Oct 1, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Oct 1, 2026
    ContributorAuthor

    os-dev-report

    {
  "issue": 20890,
  "status": "done",
  "branch": "claude/issue-20890-dataset-dimension-leg",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/21073",
  "session": "session_01Sfe5YjBLwB9J3y8fvm2xq1 — mode:subagent, the parent's id (docs round)",
  "premise_still_valid": true,
  "summary": "Docs round for the review flag in 5925454166 (section ① \"A hand-written doc now misstates the rule\", section ③ \"New flag\", and the seat's adoption); the record was read first. One file was edited, content/docs/deployment/validating-metadata.mdx, in §6, the page's account of this rule. New head b0afb0d7ba (b0afb0d7bad6c17f8514f7697e7af02538923976): one commit on be20968ee2, made in a fresh worktree from the pushed head, with no merge of main and no conflict. Diff stat: content/docs/deployment/validating-metadata.mdx | 18 ++++++++++++++++--, 1 file changed, 16 insertions(+), 2 deletions(-). PASSAGE BEFORE (:210-217 at be20968ee2): \"... and `count_distinct` over every type but the JSON-stored ones (the structured-JSON types and `multiselect` / `checkboxes` / `tags`), whose values no two backends compare for equality alike. The analytics service refuses the same pair with `400 DATASET_INVALID` when a query is built; this is the identical verdict, from the identical table, one door earlier. The rule stays silent wherever the field's type cannot be resolved (an object this stack does not define, a dangling field path, an untyped field) rather than guessing.\" PASSAGE AFTER (:210-231 at b0afb0d7ba): the same, with two additions. Inserted after \"for equality alike.\": \"The check reads the field's declaration as well as its type: a `select`, `radio`, `lookup`, `user`, `file` or `image` field declared `multiple: true` holds a list stored as JSON, so `count_distinct` over it is refused too, although the table accepts the type.\" A new paragraph closes §6: \"The same check judges the dataset's **dimensions**, under its own id. A dimension is a group key, and the analytics service refuses a query grouped by a JSON-stored column with `400 INVALID_FIELD` before any SQL is built. So a dimension whose field is declared with a structured-JSON type, or as a multi-value field (`multiselect` / `checkboxes` / `tags`, or one of the types above declared `multiple: true`), is refused (`dimension-json-stored-field-refused`) at `datasets[N].dimensions[M].field`. Group by a field that stores one value instead. Like the measure check, it stays silent when the field's type cannot be resolved.\" The page lists no rule ids in a roster; §6 names `measure-aggregate-field-type-refused` in prose, so the new id is named in that same prose. SOURCES, one per new sentence (lint = packages/lint/src/validate-dataset-measure-aggregates.ts, fv = packages/spec/src/data/field-value.zod.ts, aftc = packages/spec/src/data/aggregate-field-type-compatibility.ts): (1) \"reads the field's declaration as well as its type\": lint:260-261 shapeOf builds {type, multiple}, and lint:240-243 acceptsDeclaration judges it. (2) \"select ... image declared multiple: true holds a list stored as JSON, so count_distinct over it is refused too, although the table accepts the type\": the list is fv:335-336 MULTI_CAPABLE_TYPES (select, radio, lookup, user, file, image), checked against acceptsDeclaration; fv:355-358 isMultiValueField; lint:242 refuses count_distinct when it holds; \"stored as JSON\" is aftc:95-97; \"the table accepts the type\" is aftc:190-198 and :211, the count_distinct row minus the ten JSON-stored types, none of the six among them. (3) \"the same check judges the dataset's dimensions, under its own id\": lint:309 walks ds.dimensions inside validateDatasetMeasureAggregates, and lint:337 sets rule DIMENSION_JSON_STORED_FIELD_REFUSED. (4) \"a dimension is a group key, and the analytics service refuses a query grouped by a JSON-stored column with 400 INVALID_FIELD before any SQL is built\": packages/services/service-analytics/src/dataset-compiler.ts:695 files the dimension as a cube dimension; structured-json-dimension-door.ts:3-10 (\"refused INVALID_FIELD / 400 ... before either strategy builds anything\"; \"every JSON-stored column\"), :202-205 groupedClassOf, :260 invalidMemberError. (5) \"a dimension whose field is declared with a structured-JSON type, or as a multi-value field (...), is refused (dimension-json-stored-field-refused) at datasets[N].dimensions[M].field\": lint:279-282 groupKeyClassOf, fv:320-322 STRUCTURED_JSON_TYPES, fv:146-148 MULTI_OPTION_TYPES, fv:355-358, lint:337 for the rule and lint:339 for the path. (6) \"Group by a field that stores one value instead\": lint:352, the structured-JSON hint (\"Group by a field that stores one scalar value\"), and lint:357, the multi-value hint (\"Point this dimension at a field that stores one value\"). (7) \"Like the measure check, it stays silent when the field's type cannot be resolved\": lint:317 (an unresolved or unjudgeable path) and lint:321 (no declared type), the same skips as the measure loop at lint:374 and :378, plus the per-dataset object skip that both share. No universal is added beyond the code's set: the one \"every\" in the passage is pre-existing and is the count_distinct row's own set. NOT touched, by scope: the page's doors table row (:442, \"Dataset measure `aggregate` × the field's declared type\") still names only the measure family. Its runtime-publish cell reads \"—\", but the registry entry has been CLI_AND_RUNTIME with runtimeTypes ['dataset'] since #19143, so the cell was already stale at main before this PR. That is noted for the seat below.",
  "tests": "All at b0afb0d7ba, merge base 6073bb96b878eb2980724568ee526157b29ab141. node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack exited 0 and derived 88 families (60 before, plus 28 that the .mdx path brings: the doc-* families, the docs-* families and seven spec checks). Its --json view names 34 families that this path feeds. Two of them are CI-matrix steps (check-shard-attestation, check-test-completeness) that --commands does not print. All 88 derived commands were run at b0afb0d7ba, every one exit 0. The docs and spec ones were: check-doc-frontmatter 0 (and --self-test 0), check-doc-route-spelling --advisory 0 (and --self-test 0), check-docs-section-name 0 (and --self-test 0), check-section-landing-index 0 (and --self-test 0), lint check:doc-formula-expressions 0, lint check:doc-security-posture 0, spec check:docs 0 (\"227 generated files in sync\"), spec check:empty-state 0, check:liveness 0, check:skill-examples 0, check:strictness-ledger 0, check:variant-docs 0, check:yaml-examples 0, check:corpus-claim-drift 0, check:doc-anchors 0 (\"394 internal #fragment links ... all resolve\"), check:doc-authoring 0, check:docs-audit-scope 0, check:docs-redirects 0, check:docs-single-h1 0, check:docs-spec-enumerations 0, check:docs-transcript-drift 0, check:published-readme-links 0, check:react-page-adapter-contract 0, check:role-word 0, check:skill-identifier-liveness 0, check:vendor-version-stamps 0, docs-audit/check-affected-docs 0, check:nul-bytes 0. The changeset gates were all 0 again. The gates ran on a built tree: turbo build of the workspace minus docs and console, 72 tasks, 71 cached, exit 0. --ran reconciliation exited 0: \"88 derived, 88 run, 0 NOT-MEASURED, 0 UNRUN\", a derived zero, since every row recorded its exit code. Check-runs on b0afb0d7ba, read once: 34 total, 28 success, 2 skipped, 4 in_progress. TypeScript Type Check, Dogfood Regression Gate, Build Core, Temporal Conformance, Governed Surface Queue Guard and Build Docs: success. Lint & Repo Gates: in_progress. No check-run named Test Core or Docs Drift Check existed at the read. CI convergence was not waited on.",
  "mcp_calls": "0",
  "api_writes": "1 relay write: POST /repos/objectstack-ai/objectstack/dispatches, which becomes POST /repos/objectstack-ai/objectstack/issues/20890/comments (this report), executed as objectstack-fleet[bot]. Plus 1 git push to the branch, which is not REST. Reads were REST GETs only: comment 5925454166 and the commit check-runs. The PR body and labels were not touched.",
  "open_questions": [],
  "out_of_scope_findings": [
    "carrier: 承接者:无 · noted, not filed. The doors table in content/docs/deployment/validating-metadata.mdx (:442) gives the dataset-measure family \"—\" under runtime publish. Since #19143, authoring-rules.ts:691-723 declares that entry CLI_AND_RUNTIME with runtimeTypes ['dataset'], so the cell was already stale at main 6073bb96b8 before this PR. The row also names only the measure family, and this PR's dimension leg rides the same entry. Fixing it needs a dataset superscript that the table's legend does not have yet, which is outside this round's one-passage scope. The seat decides whether it rides this PR or goes in a docs-only follow-up."
  ]
}

    Generated by Claude Code

  13. objectstack-fleet
    removed
    needs:contract-review
    on Oct 1, 2026
  14. os-justin
    closed this as completedin #21073on Oct 1, 2026

  15. objectstack-fleet commented on Oct 1, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Oct 1, 2026
    ContributorAuthor

    Landed: PR #21073 → 5e470f8c1c, and the card is done

    domain:spec seat 5 (session_01Sfe5YjBLwB9J3y8fvm2xq1) · 2026-10-01. Holder of claim 5924574345.

    Generated by Claude Code

  16. objectstack-fleet
    removed
    pm:dispatched
    on Oct 1, 2026
  17. objectstack-fleet
    unassigned
    os-justin
    on Oct 1, 2026
  18. akarma-synetal
    added a commit that references this issue on Oct 7, 2026

    fix(lint)!: os validate refuses a dataset dimension over a JSON-store…

    5e470f8
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    area:reportsBusiness reporting — dashboards, reports, the numbers a manager readsbugSomething isn't workingdomain:specpriority:p3

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions