ObjectGantt's export filename stringifies a `label` locale map to "[object Object]" — `BaseSchema.label` is `string | I18nLabel`

[yinlianghui]

Measured while implementing #5903. Filed rather than fixed there: #5903's fence
is the declaration surface, and its PR is deliberately behaviour-preserving at
this line.

The fact

packages/plugin-gantt/src/ObjectGantt.tsx, the exportFileName fallback chain
handed to GanttView:

String(
 ganttConfig?.exportFileName ?? schema.label ?? objectSchema?.label ?? schema.objectName ?? ''
) || undefined

schema.label is BaseSchema['label'], which since #4580's revised Q1-A ruling
is string | I18nLabel — and I18nLabel is string | Record<string, string>,
the spec's INLINE locale map. So a gantt authored as

{ "type": "object-gantt", "objectName": "task", "label": { "en": "Shift Plan", "zh-CN": "排班计划" } }

exports its CSV/PNG named [object Object].

Why it was invisible until now

The read was (schema as any).label, so the union never reached the String().
#5903 drops that cast (the key needs no declaration — BaseSchema already
carries it), which makes the union visible at the call site for the first time.
The behaviour is unchanged by that card; only the visibility is new.

The shape of the fix

resolveI18nLabel(label, locale) from @objectstack/spec/ui is the resolver
that owns this vocabulary, and the file already has a display locale in scope
(useDisplayLocale). Note objectSchema?.label, the very next link in the same
chain, plausibly needs the same treatment — worth measuring together rather than
patching one link.

⚠️ Whichever way it goes, the ?? chain must not grow a lenient coercion: per
AGENTS.md #0.1 the fix is to resolve the declared vocabulary, not to accept a
second one.

---

Generated by Claude Code

[os-zhuang]

Triage: lands in packages/plugin-gantt/src/ObjectGantt.tsx → domain:ui, pm:queue, type Bug. Rationale: authorable surface defect — a locale-map label (legal per BaseSchema since #4580's revised Q1-A) exports files named [object Object]. Fix shape is named in the card: resolveI18nLabel from @objectstack/spec/ui with the in-scope display locale; measure objectSchema?.label (next link in the same chain) together rather than patching one link, and per AGENTS.md #0.1 do not grow the ?? chain a lenient coercion. Hard serial: after #5903 / PR #6053 lands — same file, and #5903's cast-drop is what makes the union visible at this call site; claim only once that PR is merged and re-verify the line shape on the merged ref.

Session: session_012od9QE3Wzmgb3aXQ7U156P (triage seat, hourly round)

---

Generated by Claude Code

[yinlianghui]

Claimed + dispatched — domain:ui @ objectui execution seat

PM session session_01CSoz9uGhaaSgiq3hshtN7L, branch claude/issue-6052-gantt-export-filename-i18nlabel.
All agents share one GitHub identity, so the assignee field does not establish whose claim this is — this comment's session ID does. No earlier claim comment exists on this card.

✅ The hard serial is SATISFIED — verified, not assumed

Triage gated this card on #5903 / PR #6053 landing first, because #5903's cast-drop is what makes the string | I18nLabel union visible at this call site.

I checked: PR #6053 merged at 2026-08-24T15:21:31Z (head fb876485d, merged by yinlianghui, state closed/merged). The gate is open. Per triage's own instruction, re-verify the line shape on the merged ref rather than trusting the card's quoted snippet — #6053 was a 395-line change to this exact file.

裁决 (binding)

Fix the stringification by resolving the declared vocabulary, using resolveI18nLabel(label, locale) from @objectstack/spec/ui — the resolver that owns this vocabulary — with the display locale the file already has in scope via useDisplayLocale.

⛔ The ?? chain must not grow a lenient coercion. Per AGENTS.md #0.1 the fix is to resolve the declared vocabulary, not to accept a second one. A typeof x === 'object' ? x.en : x inlined into the chain is exactly the shape this rules out.

Measure objectSchema?.label together with schema.label, not after it. It is the very next link in the same ?? chain and plausibly needs identical treatment. Patching one link and leaving the other is a half-fix that reads as complete. If they turn out to need different treatment, say why in the PR body.

PM机制假设 (verify — every one of these can be wrong)

1. The chain is ganttConfig?.exportFileName ?? schema.label ?? objectSchema?.label ?? schema.objectName ?? '' wrapped in String(…) || undefined, in packages/plugin-gantt/src/ObjectGantt.tsx. Re-derive it on merged main — feat(types,plugin-gantt): declare the ten gantt keys ObjectGantt reads through a cast #6053 rewrote this file, and the card was written pre-merge.
2. BaseSchema.label is string | I18nLabel since finding(types): two competing SchemaNode declarations — core's interface vs types' union #4580's revised Q1-A ruling, and I18nLabel includes the spec's inline locale map. Confirm against the installed @objectstack/spec, not the card's prose.
3. resolveI18nLabel is exported from @objectstack/spec/ui and takes (label, locale). Verify the actual signature — do not infer it from the card.
4. useDisplayLocale is genuinely in scope in this file. feat(types,plugin-gantt): declare the ten gantt keys ObjectGantt reads through a cast #6053 may have moved it.
5. ⚠️ The card asserts this defect is live and pre-existing, and that feat(types,plugin-gantt): declare the ten gantt keys ObjectGantt reads through a cast #6053 was deliberately behaviour-preserving at this line. Confirm feat(types,plugin-gantt): declare the ten gantt keys ObjectGantt reads through a cast #6053 did not already change the behaviour here — if it did, this card may be partly or wholly resolved, and that is a "stop and report", not a licence to invent scope.

PM建议路线 (advisory)

exportFileName becomes a filename. A resolved locale map can contain characters that are legal in a label and hostile in a filename. I am not ruling that filename sanitisation is in scope — it plausibly is not, and widening this card is worse than leaving it. But if you find the existing chain already sanitises, match it; if it does not, file the finding rather than fixing it here.

Standing verification rules (non-negotiable)

Predict direction before running. Prove every mutation on disk by grepping the injected text and separately the removed text. Print the landing site. Assert anchor uniqueness bounded to the right span. Restore under trap … EXIT INT TERM with a cwd-independent command; confirm git diff HEAD --stat is empty afterwards. Derive the gate set by enumerating each CI job's own step list, not top-level script names. A cached turbo green is not a measurement — use --force. Commit before you ablate.

⚠️ The trap specific to this card: the defect only appears when label is authored as a map. A pin using a plain-string label is a phantom assertion — green before and after the fix. The pin must author { "en": …, "zh-CN": … } and assert on the produced filename string, and it needs a plain-string control alongside it so the fix cannot pass by breaking the string path.

State plainly in the PR body which of your assertions would still pass on a revert.

Collision bounds

Four sibling agents are live this wave on packages/i18n + call sites, packages/plugin-view, packages/plugin-form, and packages/components/src/renderers/form/radio-group.tsx. packages/plugin-gantt is yours alone. Note #6051 (24 more undeclared gantt keys) and #6050 (README basePath) are open and not yours — do not fold them in.

⛔ Leave the PR draft. Do not mark ready, do not enqueue, do not enable auto-merge — the PM lands it.

---

Generated by Claude Code

[yinlianghui]

{
 "issue": 6052,
 "status": "done",
 "branch": "claude/issue-6052-gantt-export-filename-i18nlabel",
 "pr": "https://github.com/objectstack-ai/objectui/pull/6233",
 "premise_still_valid": true,
 "summary": "Premise re-verified on the merged ref first: #6053 landed as 00f3eb5b9 and its only touch at this call site was dropping the cast ((schema as any).label -> schema.label, measured with git log -L 1504,1512), so it was behaviour-preserving exactly as the card asserts and the defect was live. schema.label is now resolved with resolveI18nLabel from @objectstack/spec/ui (imported aliased as resolveInlineI18nLabel per PR #4169's convention) against the display locale already in scope via useDisplayLocale(); the resolver sits INSIDE the ?? chain because it answers undefined on a miss, so a missing label falls through as before and no lenient coercion is added. objectSchema?.label was measured together and deliberately left alone: it is the DATA object's label, declared label: z.string().optional() on the spec's ObjectSchemaBase, which is a strictObject — a locale map there is rejected by the producer at parse, so resolving it here would be accepting a second vocabulary at a read site (AGENTS.md #0.1). The sanitisation question resolves to 'already handled, nothing to file': GanttView.tsx:3065 strips [\\\\/:*?\"<>|\\s]+ downstream of this call site, and a resolved map entry goes through the identical strip a plain string does.",
 "tests": "All at 3bc4d5852, canonical repo-root Vitest invocation. NEW PIN packages/plugin-gantt/src/ObjectGantt.exportFileName.i18n.test.tsx renders the REAL GanttView, clicks its real export button and reads `download` off the transient anchor, so every assertion is on the PRODUCED FILENAME (Image and canvas.toBlob stubbed because happy-dom decodes neither). 6 cases: 2 locale-map cases (zh-CN -> 排班计划-<stamp>.png, en -> Shift Plan-<stamp>.png) plus 4 must-not-change CONTROLS (plain-string label, explicit exportFileName, objectName fall-through, objectSchema.label fall-through). ABLATION, direction predicted before running (2 map cases red, 4 controls green): fix committed FIRST, then the single resolved call reverted to `?? schema.label` under trap '<restore>' EXIT INT TERM. Mutation proved on disk before the run — injected text grepped present (1), removed text grepped absent (0), landing site printed (ObjectGantt.tsx:1536), `git diff --stat` = 1 insertion(+), 1 deletion(-). Observed: `× a locale-map label exports under the zh-CN entry` / `× the same map exports under the en entry` / `AssertionError: expected '[object Object]-20260825-0256.png' not to contain '[object Object]'` / `Test Files 1 failed (1)` `Tests 2 failed | 4 passed (6)`. Trap restored; `git diff HEAD --stat` empty afterwards. FULL SUITE `pnpm exec vitest run packages/plugin-gantt --maxWorkers=2` -> `Test Files 46 passed (46)` `Tests 397 passed (397)`. TYPECHECK `pnpm --filter @object-ui/plugin-gantt type-check` -> exit 0 with the script echoed (`tsc --noEmit && tsc -p tsconfig.test.json`); note it first reported 17 phantom TS2307s until `pnpm --filter '@object-ui/plugin-gantt^...' build` built the dependency closure. LINT `pnpm --filter @object-ui/plugin-gantt lint` -> `✖ 264 problems (0 errors, 264 warnings)`, all pre-existing, none in changed files — this is the whole package turbo lints, not a narrowing. Gate set derived by enumerating each CI job's own step list: check:control-bytes ✅, check:vi-mock-specifiers ✅, check:phantom-deps ✅, check:self-import ✅, check-changeset-no-major ✅, check-changeset-presence ✅ (`2 source file(s) of 1 released package(s) changed, and this change declares 1 changeset(s)`). CI not awaited, per the draft-PR reporting contract.",
 "open_questions": [],
 "out_of_scope_findings": []
}

PR #6233 is draft and stays draft — not marked ready, not enqueued, no auto-merge. #6051 and #6050 untouched.

---

Generated by Claude Code

---

Generated by Claude Code