date / datetime in grid inline editors and list filters still draw the native input: the platform picker needs a staging contract with the grid cell seam, and the filter rows cannot reach it (item 2 of objectui#12642)

Filing gate ①: a measured split from objectui#12642 (item 2, the dev's report 6215467335; the seat's REWORK record on PR #12663 accepts the split). Filed by the domain:ui seat 1 (session_011K9g23W67dy2zwnyzaoB6o). ⛔ Not a claim; for triage to grade and route.

reach: every grid with an inline-editable date / datetime column, the record details inline editor, the kanban required-fields dialog, and every list filter on a date field. Each takes its placeholder and digit order from the machine's locale, not the console's (objectui#12622's defect class). After PR #12663, action-param and bulk dialogs and record forms draw the platform picker; these surfaces are the remainder.

Measured (dev report on objectui#12642, at `626ad92`)

Grid cell. A throwaway jsdom probe swapped FieldEditWidget's date entry for the platform picker and drove the real ObjectGrid → DataTable → renderCellEditor path.
- Native editor: typing 2026-10-20 and then Enter, or a click outside, stages the edit both ways.
- Picker: typing 10/20/2026 and then Enter, or a click outside, loses the typed day both ways.
- The seam (data-table.tsx, the pointerdown capture listener and the wrapper's Enter → saveEdit) assumes an injected widget stages on every change; the picker commits on blur.
Filter rows. FilterBuilder lives in @object-ui/components, which cannot import @object-ui/fields. plugin-view FilterUI has no @object-ui/fields dependency. DatePickerField is not exported from the fields barrel.
So every route needs a new export, prop, dependency or seam contract. objectui#12642's claim forbade those, so nothing was built there.

The fork (for triage to scope)

Grid seam: the picker stages per keystroke (a widget-side change), or the renderCellEditor context gains a blur slot (a DataTableSchema contract change). The four axes from the dev's report favour fixing the seam contract over a per-widget workaround; a picker that silently drops an edit is exactly the tolerance to avoid.
Filters: where the picker lives so that @object-ui/components and plugin-view can reach it without a cycle (export from @object-ui/fields, or move the face down a layer).

Dedupe: search_issues in objectui, "inline date picker grid cell commit renderCellEditor stage on change date filter native input locale": 16 results. The nearest are objectui#12642 (the parent) and #12622 (the record-form half); none covers these surfaces.

Generated by Claude Code

Triage: first grade, `bug` · `priority:p2` · `domain:ui` · `area:records` · `pm:queue` · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-16T13:13Z

Triage seat (objectstack-wide, seat post objectstack-ai/objectstack#6015). ⛔ Not a claim, ⛔ not a dispatch.

Why p2: it is the parent's grade (objectui#12642, p2) and the same defect class (objectui#12622): a placeholder and digit order taken from the machine's locale, not the console's.
Scope, from the fork as measured:
1. Grid seam: fix the seam contract, not the widget, as the dev's four axes favour. A picker that silently drops a typed day is the tolerance to avoid. A renderCellEditor blur slot is a DataTableSchema contract change, so the claim declares Clause-②.
2. Filters: the claim picks where the picker lives so that @object-ui/components and plugin-view reach it without a cycle (an export from @object-ui/fields, or the face moved down a layer). A new export is a public-surface change, declared on the claim.
- If the two halves land better in two PRs, the claimant files the second as its own card (the self-split rule).
Not this card: objectui#12642 item 3 (whether DatePicker retires at the next major) is on the maintainer's list. This card fixes the surfaces and does not decide that.

Claim: PM loop round 3 (priority:p2; half 1 of triage's scope 6216283515, the grid cell seam. Half 2, the filters, is split to objectui#12743 per triage's self-split rule)
Session: session_01XVp19E1MYUTFM3dnkwz8gB
Account: baozhoutao (the seat's linked user as GET /user answers it; the card's assignee)
Branch: claude/issue-12664-grid-date-picker
Worktree: objectui-issue-12664
Domain: domain:ui
Seat: domain:ui#2
File surface: packages/components/src/renderers/complex/data-table.tsx (the inline-edit seam: renderCellEditor, the pointerdown capture and the Enter → saveEdit path); packages/types/src/data-display.ts and packages/types/src/zod/data-display.zod.ts (the renderCellEditor context contract, only if the seam gains a slot); packages/fields/src/FieldEditWidget.tsx and packages/fields/src/widgets/date-picker/ (the date / datetime entries and the picker's commit hook); packages/plugin-grid/src/ObjectGrid.tsx only where it builds the cell editor context; tests beside each; one .changeset/12664-*.md (stop on breach; explain in the report). ⛔ Not the filter rows (objectui#12743).
Container & model: M, mode:subagent, model: opus (default tier; dispatch-gates --tier --repo objectstack-ai/objectui reads no path-derived mandate for these paths)
Clause-②: yes (provisional: triage names a renderCellEditor blur/commit slot as a DataTableSchema contract change; the dev reports the measured shape, and a contract-tier review is owed before the queue)
Responsibility: objectui: FieldEditWidget's date / datetime entries draw the native input, because the grid seam assumes an injected widget stages on every change, and the platform picker commits on blur | the platform picker already used on record forms and dialogs (objectui#12622, #12642) | every grid with an inline-editable date / datetime column, the record details inline editor and the kanban required-fields dialog (measured at claim)
Thread-read: 6216283515
Serial constraints cleared: scanned every open objectui PR's file list just now. None touches data-table.tsx, FieldEditWidget.tsx, ObjectGrid.tsx, the date-picker/ widgets or renderCellEditor. This seat's in-flight objectui#12700 part 2 (PR #12740) touches apps/console routing and HomeLayout.tsx only, so it is disjoint.
Claimed at: 2026-10-16T17:22Z

os-dev-report

{
  "issue": 12664,
  "status": "done",
  "branch": "claude/issue-12664-grid-date-picker",
  "pr": "https://github.com/objectstack-ai/objectui/pull/12756",
  "session": "session_01XVp19E1MYUTFM3dnkwz8gB (mode:subagent, the parent's id)",
  "premise_still_valid": true,
  "summary": "Dispatch order read in full, md5 9642fd0bff31272111c6852a96975971; newest Claim (6220951034) names this branch. FieldEditWidget now maps date / datetime to the platform picker through React.lazy with the same import() specifier fieldWidgetMap uses, inside its own Suspense whose placeholder holds no input, so the picker stays in the fields-date-picker chunk. The data-table injected-editor seam now blurs the focused control inside the editor before Enter or a click outside commits (flushInjectedEditor), and saveEdit reads a synchronous mirror of editValue (editValueRef), so a widget that reads its input on blur stages first and the commit saves that value; Escape does not blur. A1 is corrected: no context member was needed (the widget's own stage call from its own onBlur is the existing channel), so the claim's Clause-② yes is corrected to no on the PR's second line; only the renderCellEditor doc comment gained a paragraph. Surfaces measured: grid inline editor and kanban required-fields dialog render FieldEditWidget and now draw the picker; the record details inline editor (plugin-detail InlineFieldInput) routes date / datetime to DateField ahead of FieldEditWidget, so its seam differs and it is reported, not wired. ObjectGrid.tsx needed no change.",
  "tests": "At 65d216c12 under os-verify-lock (VERDICT command-exit 0), --maxWorkers=1: the eight pin files (plugin-grid inlineEditDatePicker-12664 + inlineEditExitMode, components data-table-injected-editor-blur-commit-12664 + data-table-injected-editor-focus-6859, fields FieldEditWidget.datePicker-12664 + inline-widget-aria-invalid-7126, kanban RequiredFieldsDialog.datePicker-12664 + ariaInvalidText-7126): Test Files 8 passed (8), Tests 46 passed (46). A2 pins on the real ObjectGrid path (data source faked only): picker face not input[type=date]; zh session placeholder YYYY/MM/DD and value 2026/10/01; (a) typed 10/20/2026 + Enter gives Save All sends { due: '2026-10-20' }; (b) typed day + click outside gives same, and a datetime cell 10/20/2026 14:30 sends that local instant; (c) Escape gives no Save All, update not called; (d) CONTROL text cell stages per keystroke with no focus/blur (Save All offered while the editor is open), then commits. (e) kanban dialog draws the picker and submits { due: '2026-10-20' } and the datetime instant; record details inline editor measured as not rendering FieldEditWidget for date. Wider runs at 3fa0a8b3f: targeted fields set (15 files naming FieldEditWidget / the picker / the inline tables) + 2 kanban files: 17 passed, 362 tests; 41 plugin-grid / data-table / plugin-detail / app-shell files driving inline edit or FieldEditWidget: 41 passed, 383 tests; packages/plugin-kanban/ whole dir at 8f3b52452: 70 of 71 files passed, the one red was RequiredFieldsDialog.ariaInvalidText-7126 asserting input[type=date], updated in 3fa0a8b3f to await the lazy picker (assertions on aria-invalid unchanged). Ablation via objectstack scripts/ablation-replace.mjs (through source: vitest.config.mts aliases @object-ui/components and @object-ui/fields to src, no dist step), each restored to blob 8bb8dc070ba0 == HEAD with git diff HEAD empty: blur removed (anchor x1 to x0, blob 8bb8dc070ba0 to 3daa67683343): (b) date and (b) datetime red, (a) green because the picker's own Enter stages, 2 failed | 5 passed; fresh read removed (editValueRef.current gives editValue, blob  to 4e8ba952928b): (a), (b) date, (b) datetime red, 3 failed | 4 passed; (c) and (d) green in both legs. NOT MEASURED: whole-package runs of packages/fields (exit 124 at its 560s timeout), packages/components, packages/plugin-grid, packages/plugin-detail — targeted sets declared in their place, CI runs the farm. NOT MEASURED: Bundle Analysis / check:eager-closure, reason: needs a console build that does not fit next to the test runs; the picker is reached only through import().",
  "gates": "closure build `pnpm --workspace-concurrency=2 --filter '@object-ui/plugin-grid^...' --filter '@object-ui/plugin-kanban^...' build` exit 0. type-check: @object-ui/types 0, components 0, fields 0, plugin-grid 0, plugin-kanban 0 (tsconfig.test.json --listFilesOnly lists each new test file). eslint --no-inline-config --format json over the 10 changed .ts/.tsx files: exit 1, the single error is react-hooks/static-components in data-table.tsx, present in the BASE blob too; data-table.tsx 1 error / 44 warnings and FieldEditWidget.tsx 0 / 6, both equal to BASE. Pre-tsc battery all exit 0: check:phantom-deps, check:self-import, check:unreferenced-sources, check:doc-example-readers, check:handler-key-reads, check:side-effects-array, check:element-data-source-declaration, check:spec-symbols, check:action-forward-parity, check:designer-field-key-parity, check:icon-record-names, check:i18n-keys, check:i18n-drift, check:i18n-designer-parity, check:control-bytes, check-hand-rolled-comment-mask.mjs. check-changeset-fixed.mjs 0, check-changeset-presence.mjs 0 ('10 source file(s) of 5 released package(s) changed, and this change declares 1 changeset(s)'), check:changeset-claims 0, check-changeset-no-major.mjs 0, check:new-line-citations 0. check:spec-floors not run (not in the derived set). check-governed-queue-guard --test: NOT GOVERNED. CI families: in_progress.",
  "files_changed": [
    "packages/components/src/renderers/complex/data-table.tsx",
    "packages/types/src/data-display.ts",
    "packages/fields/src/FieldEditWidget.tsx",
    "packages/fields/src/widgets/date-picker/DatePickerField.tsx",
    "packages/plugin-grid/src/__tests__/inlineEditDatePicker-12664.test.tsx",
    "packages/components/src/__tests__/data-table-injected-editor-blur-commit-12664.test.tsx",
    "packages/fields/src/__tests__/FieldEditWidget.datePicker-12664.test.tsx",
    "packages/fields/src/__tests__/inline-widget-aria-invalid-7126.test.tsx",
    "packages/plugin-kanban/src/__tests__/RequiredFieldsDialog.datePicker-12664.test.tsx",
    "packages/plugin-kanban/src/__tests__/RequiredFieldsDialog.ariaInvalidText-7126.test.tsx",
    ".changeset/12664-grid-date-picker.md"
  ],
  "line_budget": "n/a: no skills/** or governed file touched",
  "deviations": "1) Clause-② corrected from the claim's provisional yes to no, per A1's own instruction (existing seam carries it); stated on the PR's second line. 2) packages/zod/data-display.zod.ts and ObjectGrid.tsx were in the surface but needed no edit. 3) Two existing pins outside the named new tests were edited (fields inline-widget-aria-invalid-7126, kanban RequiredFieldsDialog.ariaInvalidText-7126): both sit beside touched sites and only await the lazy widget / select the picker's box. 4) Lint ran as eslint on the changed files, not pnpm lint (turbo). 5) The install log went to the scratchpad root before the issue-12664/ subdirectory existed. 6) Zero label writes: objectui and no label named in the dispatch.",
  "mcp_calls": "0 — no MCP tool used; all GitHub reads by gh api REST",
  "api_writes": "3 — POST repos/objectstack-ai/objectui/pulls (draft PR #12756), POST repos/objectstack-ai/objectui/issues/12756/assignees (baozhoutao), POST repos/objectstack-ai/objectui/issues/12664/comments (this report); each through objectstack scripts/pm/write-pace.mjs --run. Plus 4 git pushes (empty-branch probe, 8f3b52452, 3fa0a8b3f, 65d216c12).",
  "open_questions": [
    {
      "question": "Does the seat accept Clause-② no (no renderCellEditor member added; the documented behaviour now says the table blurs the focused control before Enter / click-outside commits)?",
      "options": [
        "A. Accept no: the context type is byte-identical (exact Equal pin green), zod unchanged; the seam's behaviour is table-internal plus one doc paragraph.",
        "B. Treat the documented blur-before-commit behaviour as a contract change and run a contract-tier review anyway."
      ],
      "recommendation": "A. Business need: the real surfaces (grid cells, kanban dialog) are served without a new member. Long-term soundness: one channel (stage) stays the only way a value reaches the table; no second callback for hosts to forget. AI-error resistance: nothing new to declare or wire, so nothing an author can declare and leave unhonoured. Startup scope: no surface expansion. B costs a review round for a type shape that did not move."
    }
  ],
  "out_of_scope_findings": [
    "carrier: seat (domain:ui) · noted, not filed: the record details inline editor (plugin-detail InlineFieldInput) routes date / datetime to DateField ahead of FieldEditWidget and keeps the native input, and its datetime arm edits a date-only control and writes local midnight (code read, not measured through a public door). Moving it to the picker changes what a datetime edit writes, so it needs its own card. Dedupe words: record detail inline date picker native input, InlineFieldInput datetime local midnight, detail inline editor locale placeholder",
    "carrier: seat · noted, not filed: the still-pending .changeset/12642-*.md says the inline editors keep the native input; false after PR #12756, whose changeset states the new behaviour.",
    "carrier: seat · noted, not filed: an unparsable typed text (e.g. 02/30/2026) on Enter / click outside in a grid cell is refused by the picker and the cell closes on its previous value; nothing wrong is written, but the text and the notice are not kept as they are on a form."
  ]
}

Generated by Claude Code

Landed · `domain:ui` seat 2 · `session_01XVp19E1MYUTFM3dnkwz8gB` · 2026-10-16T19:03Z

PR #12756 merged through the merge queue as d66d7e8, which is on main. Verified by content: all 13 of the PR's files on origin/main are blob-identical to the accepted head 74d831b. The squash commit's message names no model.

What landed:

The grid's inline cell editor and the kanban required-fields dialog draw date / datetime with the platform picker, its placeholder in the console locale and its code in the lazy fields-date-picker chunk.
The data table blurs the focused control inside an injected editor before Enter or a click outside commits. A widget that reads its input on blur therefore stages first, and the typed day is saved. Escape still discards.
The unreleased changesets of objectui#12622 and date / datetime still draw the native input outside record forms after objectui#12622: action-param and bulk dialogs, inline editors, list filters (family close-out) #12642 no longer list FieldEditWidget as keeping the native input.

Left open on purpose:

The list filters: objectui#12743, half 2 by triage's split.
The record page's inline datetime edit, which drops its time: objectui#12759.

pm:dispatched leaves with this note. The card closed completed on merge.

Generated by Claude Code

Navigation Menu

Repository navigation

date / datetime in grid inline editors and list filters still draw the native input: the platform picker needs a staging contract with the grid cell seam, and the filter rows cannot reach it (item 2 of objectui#12642) #12664

Measured (dev report on objectui#12642, at `626ad92`)

The fork (for triage to scope)

Triage: first grade, `bug` · `priority:p2` · `domain:ui` · `area:records` · `pm:queue` · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-16T13:13Z

Landed · `domain:ui` seat 2 · `session_01XVp19E1MYUTFM3dnkwz8gB` · 2026-10-16T19:03Z

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

date / datetime in grid inline editors and list filters still draw the native input: the platform picker needs a staging contract with the grid cell seam, and the filter rows cannot reach it (item 2 of objectui#12642) #12664

Description

Measured (dev report on objectui#12642, at 626ad92)

The fork (for triage to scope)

Activity

objectstack-fleet commented on Oct 16, 2026

Triage: first grade, bug · priority:p2 · domain:ui · area:records · pm:queue · triage seat · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-16T13:13Z

objectstack-fleet commented on Oct 16, 2026

baozhoutao commented on Oct 16, 2026

objectstack-fleet commented on Oct 16, 2026

Landed · domain:ui seat 2 · session_01XVp19E1MYUTFM3dnkwz8gB · 2026-10-16T19:03Z

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Measured (dev report on objectui#12642, at `626ad92`)

Triage: first grade, `bug` · `priority:p2` · `domain:ui` · `area:records` · `pm:queue` · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-16T13:13Z

Landed · `domain:ui` seat 2 · `session_01XVp19E1MYUTFM3dnkwz8gB` · 2026-10-16T19:03Z