plugin-timeline: ObjectTimeline has no absent-date-axis refusal path and floors internally at `|| 'date'` — steps ① and ② of the #7070 ruling, prerequisite for ③

[os-project-manager]

Filed by the domain:ui execution seat (session session_01EMrWaQw3XS5DxTHxp4yRyC) because the ruling on #7070 instructs this lane to file it: 「① ObjectTimeline 先加 refusal 屏(gantt 同款,独立小卡先行,车道立)」.

The ruling that orders this

#7070 comment 5494805467 — project director seat, 总监批 #28, 2026-09-01, maintainer verbatim 「同意」. Quoted unmodified (⛔ a translated ruling is a rewritten ruling):

1. A:退役 timeline 的 created_at 虚构,三步序执行(dev 测量已证单删会把虚构挪进更不可见的 || 'date'):
   ① ObjectTimeline 先加 refusal 屏(gantt 同款,独立小卡先行,车道立);
   ② 退役渲染器内部 || 'date' floor;
   ③ 两个 plugin 面(ListView.tsx / plugin-view/ObjectView.tsx)删 created_at —— 依赖 ①② 落地;
2. ListView.tsx:2292-2293 的书面决定被本裁决明确替代 —— house posture 一句入册:日期轴永不虚构(refuse)。

This card is ① and ②. Step ③ stays on #7070 and is blocked on this one.

Why ① and ② are one card and not two

The ruling names ① as the card to file first and lists ② separately. They are filed together here because each is inert without the other, and that is the measured lesson this whole family exists to record:

• A refusal screen added while the || 'date' floor stands is unreachable — the floor guarantees a date field name always resolves, so the refusal branch can never be taken. That is exactly the shape ObjectView invents startDateField: 'due_date' / titleField: 'name' for views with no calendar: block — the guess short-circuits ObjectCalendar's own refusal screen and piles every record on today #7029/fix(app-shell,plugin-list,plugin-view): no invented calendar field names, so the refusal screen becomes reachable #7062 measured on the calendar face and finding(views): the gantt and timeline branches still fabricate date-axis field names on all three faces — and app-shell's own #3129 note certifies them as already fixed #7070 measured on the gantt face.
• Retiring the floor without a refusal screen produces the failure the ruling explicitly rejects: a renderer with no honest refusal, bucketing everything into "No date".

⇒ Shipping them in one PR is the only ordering in which either is observable. If the director seat wants them split, splitting is trivial and this card is the place to say so — ⛔ this seat is not re-deciding the ruling, only reading its step order as an execution sequence rather than as two PRs.

The measurement this rests on

From the #7070 dev report (5486992293), accepted by the seat at the time. ⚠️ It was taken on 2c3cd1b75 and is now several days and many merges old — re-measure before implementing; treat every line number as stale by construction.

• No refusal path exists. Zero hits for an absent-axis refusal in packages/plugin-timeline, against a live control: the term refus hits 5 times in that same file (the finding(plugin-timeline): an object-bound timeline with variant: "gantt" throws — calculateDateRange reads a row shape ObjectTimeline never composes #6655 variant: 'gantt' refusals). The channel discriminates, so the zero is a reading.
• The internal floor, reported at packages/plugin-timeline/src/ObjectTimeline.tsx:247-249:
  startDateField = timelineConfig?.startDateField ?? timelineConfig?.dateField ?? schema.mapping?.date ?? schema.startDateField ?? schema.dateField ?? 'date'
  Four honest rungs and then a fabricated fifth. Nobody has ever declared the name date.
• The ADR-0047 timeline capability gate is already clean — it reads resolveTimelineDateBinding(schema).startDateField, which does not include the 'created_at' floor. So unlike gantt, this fabrication never lit the Timeline toggle, and the surviving reach is the deprecated flat prop, hit when a view is forced onto the timeline renderer. The blast radius is smaller than the face-to-face contradiction suggests — worth knowing before sizing the work.

Ask

1. Add the absent-axis refusal path to ObjectTimeline. The gantt twin is the settled in-repo shape — getGanttConfig returns null when neither a config block nor the required flat props are present, and the early return renders a refusal telling the author which fields to declare. Follow it rather than inventing a second dialect; name the fields the author must supply.
2. Retire the internal || 'date' floor so the refusal becomes reachable. ⭐ Prove reachability rather than asserting it: the refusal pin must distinguish refuses from renders an empty timeline, by asserting the timeline canvas is ABSENT and not merely event-free. A pin that only counts events passes on both.
3. ⛔ Do NOT touch 'created_at' on either plugin face. That is step ③, it lives on finding(views): the gantt and timeline branches still fabricate date-axis field names on all three faces — and app-shell's own #3129 note certifies them as already fixed #7070, and it is blocked on this card. Deleting it here resolves a sequence the ruling deliberately ordered.
4. Re-check the other four rungs. If any of them is also a fabrication rather than a declared binding, report it — do not extend scope.

Close condition

On the merged ref: ObjectTimeline has a refusal path reachable when no date axis is declared (pinned, canvas-absent form), and the ?? 'date' floor returns 0 hits in packages/plugin-timeline/src/ObjectTimeline.tsx — with a control term on the same file still firing, or the zero is not a reading.

Clause ②

Expected no — this is the view/renderer layer ceasing to invent a field name, not a published accept surface moving. ⚠️ Re-declare against the actual diff: if the honest fix turns out to move a published type or validator, the tier follows the finding, not this line.

Not in this card

• Step ③ ('created_at' at ListView.tsx and plugin-view/ObjectView.tsx) — finding(views): the gantt and timeline branches still fabricate date-axis field names on all three faces — and app-shell's own #3129 note certifies them as already fixed #7070.
• flavour 3, progress / dependencies — the ruling routes these separately and ⛔ forbids applying the date-axis conclusion to them: 「非轴字段按每行缺值语义单独评估 … ⛔ 不套用日期轴结论」. Measured to produce a per-row missing value, not a wrong axis for the whole chart.
• The house posture sentence itself (日期轴永不虚构) is already on record in the ruling; ⛔ this card does not restate or amend it.

Triage note

domain:ui is set here because the ruling assigns this card to this lane by name; it is transcribed, not judged. type and priority are deliberately left for the triage seat — ⛔ this execution seat is not their producer.

Dedupe: semantic search for an ObjectTimeline absent-axis refusal returned 5 results — #7365 (a different timeline defect: a null gantt bar refused through the wrong fault), #7070 (this card's parent), and #6781 / #7029 / #6750, all closed and all different defects. No card asks for this refusal path. The control fired: the query returned this card's true relatives rather than nothing, so the channel discriminates. ⚠️ That is a discrimination control, not a completeness control — it cannot rule out a truncated corpus.

Refs: #7070 (the ruling and step ③) · #7062 / #7029 (the calendar precedent) · PR #7110 (the gantt half, landed) · #6655 (the existing variant: 'gantt' refusals in this same file)

[os-project-manager]

Claim — domain:ui execution seat

Session session_01EMrWaQw3XS5DxTHxp4yRyC. Branch claude/issue-7459-timeline-absent-axis-refusal. Filed and dispatched in the same round, because the ruling on #7070 orders this card into existence and specifies its content — there is nothing left to triage before work can start.

File surface declared for cross-lane serialization: packages/plugin-timeline/src/ObjectTimeline.tsx and its tests. Checked against every open PR this lane holds — no intersection:

PR	card	holds
#7447	#6940	data-display.zod.ts, the parity ledger
#7456	#6939 group 1	overlay.zod.ts
#7453	#7071	plugin-calendar/ObjectCalendar.tsx — enqueued
#7457	#7088	types/base.ts
(in flight)	#7344	app / reports / views / crud / complex zod faces
(in flight)	#6939 group 6	objectql.zod.ts

packages/plugin-timeline/ is held by nothing. ⚠️ Declared with the lesson from this shift attached: I once declared "no intersection" having checked exactly one open PR's file list, and the collision I missed would have had a dev write refusal arms for symbols another PR was deleting. The table above is every open PR in the lane, each read.

Scope, restated for the record

⛔ Step ③ — deleting 'created_at' at the two plugin faces — is not in this dispatch. It stays on #7070 and is blocked on this card landing. The brief carries that fence explicitly, because it is the one edit a dev in this file is most likely to make on momentum, and making it here would silently collapse a sequence the ruling deliberately ordered three steps deep.

⛔ flavour 3 (progress / dependencies) is likewise fenced out: the ruling forbids applying the date-axis conclusion to non-axis fields, and the measurement behind that is that a fabricated name there yields a per-row missing value rather than a wrong axis for the whole chart. Different defect, different question.

The gate the brief puts first

The dev is told to re-measure both premises before implementing — the absent refusal path and the ?? 'date' floor — rather than inheriting them from the #7070 report, which was taken on 2c3cd1b75 several days and many merges ago. Every line number in this card is stale by construction and the brief says so.

Clause ② expected no (the view layer stops inventing a field name; no published accept surface moves), re-declared against the actual diff. So this one is enqueueable by this seat once green — ⛔ unlike the Clause-② cards currently waiting on the project director seat.

State transition, same stroke: pm:queue → pm:dispatched, assignee set.

---

Generated by Claude Code

[os-project-manager]

Claimed by the domain:ui execution seat — session session_01EMrWaQw3XS5DxTHxp4yRyC, branch claude/issue-7459-timeline-absent-axis-refusal.

Draft PR: #7467. Assignee and labels deliberately untouched (PM owns both).

Both premises were re-measured on 3804c222b before the first edit and both reproduce; the full reading, including one correction to the card's rung arithmetic, is in the PR body and in the report comment that follows.

---

Generated by Claude Code

[os-project-manager]

os-dev-report

{
 "issue": 7459,
 "status": "done",
 "branch": "claude/issue-7459-timeline-absent-axis-refusal",
 "pr": "https://github.com/objectstack-ai/objectui/pull/7467",
 "premise_still_valid": true,
 "gate_re_measurement": {
 "taken_on": "3804c222b, before the first edit (card figures were from 2c3cd1b75)",
 "premise_1_no_refusal_path": "REPRODUCES. 0 hits for an absent-date-axis refusal in packages/plugin-timeline. Control fired in the same run on the same file: `refus` matched 6 LINES of ObjectTimeline.tsx (card said 5 — the file moved, so the control is live rather than copied forward). All 6 are the #6655 `variant: 'gantt'` refusal or its prose; none concerns a date axis.",
 "premise_2_internal_floor": "REPRODUCES, and the line numbers happen to still hold: packages/plugin-timeline/src/ObjectTimeline.tsx:247-249, spelled exactly as reported.",
 "correction_to_the_card": "The card and the dispatch both say 'four honest rungs, then a fabricated fifth'. The chain reads FIVE declared bindings and a fabricated SIXTH: timeline.startDateField, timeline.dateField, mapping.date, flat startDateField, flat dateField. Off by one; reported, not acted on beyond retiring the literal.",
 "other_rungs_re_checked": "All five are genuine DECLARED bindings — no further fabrication. The first two are declared on ListViewTimelineConfig (packages/types/src/objectql.ts); the last three on this component's own props and on its module-local TimelineExtensionSchema. Scope not extended."
 },
 "summary": "Added the absent-date-axis refusal to ObjectTimeline and retired the internal date-name floor from the end of its resolver chain, as one change — steps (1) and (2) of the #7070 ruling. The refusal follows the settled gantt shape (early return, role=alert, names the fields to declare) and interpolates its field list from a new module-local OBJECT_BOUND_TIMELINE_DATE_BINDINGS, so the diagnostic cannot drift from the resolver. One enabling change is declared rather than smuggled: the component's SUCCESS surface is now named `data-testid=timeline-canvas` — every other terminal state of this component already named itself, and without a name for the rendered timeline a pin cannot distinguish 'refused' from 'rendered an empty timeline', which is the distinction the ruling turns on. One new i18n key across all ten packs plus the plugin default table. Step (3) untouched: the two `created_at` floors are exactly where they were (re-derived: plugin-list/ListView.tsx:2496, plugin-view/ObjectView.tsx:1354 — the card's numbers were stale), and flavour 3 progress/dependencies not touched. Blast radius measured: nothing in the product starts refusing, because both plugin faces still supply created_at and app-shell routes through ListView; the refusal is reached only by a view forced directly onto the renderer with nothing declared, exactly as the card predicted.",
 "tests": "All figures from runs at the FINAL commit d2dbdfb8c; heavy runs serialized through the shared verify lock, verdict lines read, never a bare $?; exit codes captured before any pipe. (1) UNION plugin-timeline + i18n + plugin-list + plugin-view -> 'Test Files 175 passed (175) / Tests 2288 passed (2288)', VERDICT command-exit 0. (2) Pre-edit BASELINE for comparison: plugin-timeline 21 files / 276 tests; after: 22 / 287 (the new pin is 11 cases). (3) type-check plugin-timeline + i18n -> exit 0, 'Scope: 2 of 47 workspace projects', each echoing 'tsc --noEmit && tsc -p tsconfig.test.json' + 'Done', so the zero-match pnpm filter trap did not fire; and the new test file is genuinely measured rather than excluded — `tsc -p tsconfig.test.json --listFiles` counts it 1 (the NOT-MEASURED trap, checked rather than assumed). Dependency closure was built first (`--filter '<pkg>^...' build`); the first type-check attempt failed TS6305 on an unbuilt core/dist and is reported as that, not as a red gate. (4) GATES, verdict lines quoted from each gate's own output: check:i18n-keys 'every in-scope call-site key resolves against the en pack (2960 keys) …'; check:i18n-drift '0 en value(s) changed (1 key(s) added, 0 removed)'; check:i18n-dead-keys (report-only) 0 new; check:control-bytes 'OK (scanned 6140 tracked text file(s))'; check:vi-mock-specifiers 'OK'; check:vi-mock-inherit 'OK (122 call site(s) on @object-ui/react judged, 122 inherit, 0 auto-mocked)'; check:side-effects-array OK; check-changeset-presence '13 source file(s) of 2 released package(s) changed, and this change declares 1 changeset(s)'; check-changeset-no-major 'No changeset declares a major bump'; check-changeset-fixed OK; check-changeset-overwrite OK — all exit 0. (5) NOT MEASURED, read as neither red nor green: check:readme-exports (all 314 findings are 'type entry ./dist/index.d.ts is not on disk — run pnpm build first') and check:eager-closure ('No eager-closure report … the console was not built', which the gate itself calls a broken gauge). Both need a full-monorepo / console build that CI does regardless. (6) eslint PLAIN form (never --no-inline-config): exit 0, 0 errors, 29 warnings. DECLARED NARROWING: 14 changed files, not the repo-wide farm. Evidence it excludes nothing: file count 14 read from --format json; eslint.config.js enables NO type-aware linting (no projectService / parserOptions.project), so this diff cannot move an untouched file's verdict; CI runs the farm anyway. The two react-hooks warnings on ObjectTimeline.tsx were confirmed PRE-EXISTING by linting the origin/main copy of the file through --stdin: identical rule counts. (7) ABLATION, direction PREDICTED then observed. Implementation committed FIRST so the restore leg had a real reference. Mutation = put the retired floor BACK while the refusal stays, i.e. the pairing itself. Predicted: 4 RED (the three refusal cases plus the message case) and 7 GREEN BY CONSTRUCTION (the five declared-rung controls, the authored-items control, the #6655 ordering case — all fully declared, so unaffected in both worlds). Observed: 'Tests 4 failed | 7 passed (11)', and the four failures are exactly the four predicted, by name — the asymmetry that separates 'stopped inventing' from 'refuses everything'. MUTATION PROVEN ON DISK, not by an editor's exit code: anchor LINES 1 -> 0, injected lines 0 -> 1, the injected hit printed and read (not just counted), blob 1b4575de… -> 8cea816b…; the anchor was asserted unique BEFORE mutating, so a zero-match no-op would have aborted rather than reported a healthy run. RESTORE PROVEN BY STATE, not by exit code: git hash-object == git rev-parse HEAD:PATH (non-empty and matching), git diff HEAD empty, git status clean; absolute-path trap on EXIT INT TERM. NO dist leg exists for this ablation and none is claimed — the pin imports the subject by relative path from source (./ObjectTimeline), so no rebuild sits between the mutation and the measurement. (8) CLOSE CONDITION verified after the final commit: the floor literal returns 0 hits in ObjectTimeline.tsx while the control term fires at 25 lines in that same file. The tombstone comment is deliberately spelled so it does NOT quote the removed literal — a comment quoting it would have made that scan answer 1 where the truth is 0.",
 "mcp_calls": "3 — create_pull_request, add_issue_comment (claim), add_issue_comment (this report). Every read went through the zero-quota channel: the issue body and both #7070 comments via the public REST endpoints, everything else via git and the local tree. The PR body was read back in full through REST (0 quota) and is byte-identical to what was sent apart from the platform's own appended session footer — no sanitizer truncation.",
 "clause_2": "NO — re-declared against the actual diff rather than inherited from the card. No spec key, no zod validator, no exported type moves. OBJECT_BOUND_TIMELINE_DATE_BINDINGS is module-local and deliberately not exported (an exported array trips react-refresh/only-export-components, as the sibling constant's comment records). The only new published surface is one data-testid attribute and one i18n key. startDateField's type widens to string|undefined, but it is a local inside the component, not a prop. Changeset .changeset/7459-timeline-absent-date-axis-refusal.md, minor for @object-ui/plugin-timeline and @object-ui/i18n — the grade #7070 used for the identical class; graded here and confirmed by the gate's own line, check-changeset-no-major 'No changeset declares a major bump'.",
 "files_changed": [
 "packages/plugin-timeline/src/ObjectTimeline.tsx (the refusal, the retired floor, the named canvas, the module-local binding list)",
 "packages/plugin-timeline/src/ObjectTimeline.absentDateAxisRefusal-7459.test.tsx (new — the pin, 11 cases)",
 "packages/plugin-timeline/src/useTimelineTranslation.ts (the new key in the plugin default table, byte-identical to en)",
 "packages/i18n/src/locales/{en,de,es,fr,ja,ko,pt,ru,zh,ar}.ts (timeline.unconfigured.noDateAxis, all ten packs)",
 ".changeset/7459-timeline-absent-date-axis-refusal.md (minor x2)"
 ],
 "notes_worth_the_pm_seat_reading": [
 "The de value carries NO typographic quote pair on purpose: de-quote-pairing-3876.test.ts ratchets paired-span counts, and it stays at 59 rather than needing a bumped ceiling.",
 "No docs page teaches the fabricated default — searched content/, docs/, apps/, examples/ and the package README; the only hits are CHANGELOG prose, which is never hand-edited. Nothing to correct, and content/docs/releases/ was not touched.",
 "The refusal is placed BELOW the #6655 variant refusal deliberately, and that ordering is pinned rather than left to line placement."
 ],
 "open_questions": [],
 "out_of_scope_findings": []
}

---

Generated by Claude Code

[os-project-manager]

Closed — landed as 20cb8db9b (PR #7467), verified by content on origin/main

domain:ui execution seat, session session_01EMrWaQw3XS5DxTHxp4yRyC. pm:dispatched stripped.

The readings

probe	before (177afeba1)	origin/main
the fabricated ?? 'date' floor	1	0
⭐ CONTROL startDateField, same file	12	18
data-testid="timeline-canvas"	—	present
noDateAxis across the locale packs	—	10
⛔ FENCE created_at — plugin-list/ListView.tsx	—	4, unchanged
⛔ FENCE created_at — plugin-view/ObjectView.tsx	—	1, unchanged

The control rising 12 → 18 is what makes that zero a reading: a probe that had simply stopped matching would show the same 0 with a flat control.

⭐ And the tombstone comment in the shipped code is deliberately spelled so it does not quote the retired literal — the dev's reason, stated up front: a comment quoting it would make this very close-condition scan answer 1 where the truth is 0. That trap fired on this lane three separate times today; here it was designed out before it could.

Steps ① and ② of the ruling are done; ③ is now unblocked

Ruling #7070, 总监批 #28 (comment 5494805467) sequences ① the refusal screen and ② the retired floor together — neither is observable alone — before ③. Both are on main, and the fence above proves ③ did not leak in early: the two created_at floors at the plugin faces are untouched at 4 and 1.

⇒ #7070 step ③ is dispatched next. House posture on record with that ruling: 日期轴永不虚构.

What this card actually bought

An object-bound timeline that declared no date axis used to resolve one anyway — a bare literal nobody had ever declared, sitting below the created_at the view faces supply. Every record then read a key its object does not carry, landed in the "No date" bucket, and the screen read as a timeline that had been built and simply had nothing in it. It now refuses, naming every binding it accepts.

⚠️ Blast radius, measured rather than assumed: nothing in the product starts refusing. Both plugin faces still supply created_at and app-shell routes through ListView, so the refusal is reached only by a view forced directly onto the renderer with nothing declared.

One correction to this card that belongs in the record

The card and my dispatch both said "four honest rungs, then a fabricated fifth." The chain actually reads five declared bindings and a fabricated sixth. The dev re-derived it instead of trusting the count, checked the other five (all genuine declared bindings, no second fabrication), and reported the off-by-one without extending scope to "fix" anything. The error was mine.

⚠️ Also worth carrying: this PR sat red on Bundle Analysis for most of the shift, and the red was its own — one refusal string in ten locale packs, against a ceiling named framework that was in fact 78.7% translation catalogue. It went green with not one byte of its copy changed once #7399's option A′ (PR #7489) put those bytes on the right budget. I published three wrong causal stories about that ceiling before the measurement settled it; the record is on #7399.

---

Generated by Claude Code