Repository navigation
fix(plugin-list,components,i18n): list filter starts on the first visible column, hides hidden fields, one empty check where empty and null match the same records (objectui#11810) - #11836
Merged
objectstack-fleet[bot] merged 5 commits intoOct 8, 2026
Conversation
…le column, drops hidden fields, offers one empty check per null-only type (objectui#11810) ListView hands its FilterBuilder the object definition's field map minus the fields marked hidden (kept only when filterableFields names one or a held condition filters on it), ordered: the view's columns as the grid shows them, other business fields, system fields (isSystemManagedField). "Add filter" seeds on that list's first entry, so a new condition starts on the first visible column instead of the hidden organization_id. FilterBuilder's operator dropdown no longer offers is_null / is_not_null on a column whose type the spec's expandEmptyOperator rates null_only: there is_empty matches exactly the same records in every dialect the builder writes. Text and list-valued columns keep both pairs. The row's own operator stays mounted, so a stored is_null still loads, and operatorsForFieldType (what a row can hold) is unchanged. Claude-Session: https://claude.ai/code/session_01CGZy1BGCjdN5cXqL9cnvB8 Co-authored-by: Claude <noreply@anthropic.com>
…a column keeps both (objectui#11810) On a text-like or list-valued column the operator dropdown keeps both empty-check pairs, and now closes with one line saying how they differ: "Is empty" also matches blank text (or an empty list); "Is null" matches only a missing value. The arm is the spec's expandEmptyOperator; the operator labels are interpolated, not restated. New language-pack keys filterBuilder.emptyCheckHint.text / .list in all ten packs; the de quote-pairing census moves 71 -> 75 for their spans. Claude-Session: https://claude.ai/code/session_01CGZy1BGCjdN5cXqL9cnvB8 Co-authored-by: Claude <noreply@anthropic.com>
…row instead of any (objectui#11810) Claude-Session: https://claude.ai/code/session_01CGZy1BGCjdN5cXqL9cnvB8 Co-authored-by: Claude <noreply@anthropic.com>
…bjectui#11810) The two new language-pack keys widen @object-ui/i18n's published packs, and @object-ui/components ships the hint that reads them, so both declare minor (the precedent objectui#11645 set for added keys). plugin-list's part is a fix and stays patch; the fixed group versions all three together at the group's highest level. Claude-Session: https://claude.ai/code/session_01CGZy1BGCjdN5cXqL9cnvB8 Co-authored-by: Claude <noreply@anthropic.com>
Contributor
✅ Console Performance Budget
The eager closure is every chunk the entry reaches through static imports — what the browser fetches and parses before the app renders. The entry chunk on its own is a small fraction of it. 📦 Bundle Size Report
Size Limits
|
Contributor
Author
Contract reviewServed-tier: Inputs read: card #11810 (body and all five comments, 6043301317 triage / 6046638726 claim / 6048143784 dev report / 6048182477 claim amendment / 6048273522 dev addendum), PR #11836 (body, 17-file list, net diff against ① Derived judgmentsAccept-set changes (what the end-user list and the shared builder read and offer):
Public-surface changes (what a published package's accept set gains or claims):
② Semver level
③ Boundary flags
What clears this record: one commit on the branch carrying the manifest line and the changeset sentence from ①.10, after which a fresh Implemented-by: VERDICT: FAIL |
…andEmptyOperator (objectui#11810) The published entry now imports expandEmptyOperator from @objectstack/spec/data, which the spec first exports in 17.5.0; ^17.0.0 admitted 17.0-17.4, where the import breaks. pnpm-lock.yaml regenerated with pnpm install --lockfile-only: the importer specifier line only. The changeset names the raise. Claude-Session: https://claude.ai/code/session_01CGZy1BGCjdN5cXqL9cnvB8 Co-authored-by: Claude <noreply@anthropic.com>
Contributor
✅ Console Performance Budget
The eager closure is every chunk the entry reaches through static imports — what the browser fetches and parses before the app renders. The entry chunk on its own is a small fraction of it. 📦 Bundle Size Report
Size Limits
|
Contributor
Author
Contract reviewServed-tier: Inputs: card #11810 (body and all seven comments: triage 6043301317, claim 6046638726, reports 6048143784 / 6048273522 / 6048846867, amendments 6048182477 / 6048666363), PR #11836's body and 19-file list, the net diff ① Derived judgmentsCheck-runs on this head — the gate verdicts. 45 check-runs, all completed: 42
Scope against the claim. The 19 files are exactly the claim's File surface as amended twice: ② Semver level
③ Boundary flagsFrom report 6048143784:
From report 6048273522:
From report 6048846867:
From the PR's Acceptance notes:
Nothing left unanswered; nothing further to escalate for this head. Implemented-by: VERDICT: PASS Rendered 2026-10-07T23:49Z on the inputs named above. |
This was referenced Oct 7, 2026
objectstack-fleet
Bot
deleted the
claude/issue-11810-filter-builder-fields
branch
October 8, 2026 00:09
akarma-synetal
pushed a commit
to akarma-synetal/objectui
that referenced
this pull request
Oct 9, 2026
…d by its predicate says why it is unavailable (objectui#11811) (objectstack-ai#11841) Fixes objectstack-ai#11811 Clause-②: yes A record action greyed out by its declared `disabled` predicate now says why. It shows the generic reason "Not available for this record" as a tooltip, on hover and on keyboard focus, and the same text is its accessible description (`aria-describedby`). Before, the button carried no tooltip, no `title` and no description. ## Where the card's defect actually lives The claim's original Responsibility line named `DeclaredActionsBar` as the code that draws the showcase Task page's *Archive*. It does not: `DeclaredActionsBar` renders only `sys_approval_request` actions (record page approvals bar, approvals inbox). Measured on objectstack `main` (the card's `879bd38c` is its ancestor), the Task page (`showcase_task_detail`, `kind: 'full'`) draws Archive in two other renderers: - **the header**: a `page:header` node whose `actions` are ids, drawn by `PageHeaderRenderer`'s `renderButton` (`@object-ui/components`); - **the section bar**: `record:quick_actions` at `location: 'record_section'`, drawn by `QuickActionButton` (`@object-ui/plugin-detail`). The seat recorded the correction in claim amendment 6049093782 and added the header file to the surface. `DeclaredActionsBar` has the same defect class, so it gets the same fix. ## What changed - `page:header` (`packages/components/src/renderers/layout/containers.tsx`): the inline button gets a tooltip trigger plus an `sr-only` description. In the ⋯ overflow menu the reason is a visible second line under the label and is the item's `aria-describedby`; the item's name stays the label (`aria-labelledby`). A tooltip there would be out of keyboard reach: the menu's roving focus skips a disabled item (`focusable: !disabled`) and the menu traps Tab. **Only an authored action's declared `disabled`** gets the reason. The host-injected Edit and Delete (`RecordContext.headerSystemActions`, a host-computed boolean) and the inline-edit lock are unchanged. Authored and host actions are told apart by key, from the dedupe that lets an authored action win. - `record:quick_actions` (`QuickActionButton`) and `DeclaredActionsBar`: the same tooltip trigger and description on the inline button. - One new key, `actions.notAvailableForRecord`, with real translations in all ten packs. **Why a wrapper span.** A natively `disabled` button fires no pointer or focus events, and the Button primitive adds `disabled:pointer-events-none`, so a tooltip or a native `title` on the button itself never opens. The span is the tooltip trigger, the pattern Radix documents for a disabled trigger. It takes the hover, and `tabIndex={0}` lets a keyboard user focus it. Both the button and the span carry `aria-describedby` pointing at a persistent `sr-only` copy, so a screen reader gets the reason with the tooltip closed. The repo had no disabled-tooltip pattern. Its `title`-on-disabled-Button sites cannot show at all; a Chromium probe found that the hit target at a `pointer-events: none` disabled button's centre is its parent. **Not changed:** a button greyed out only while its own action runs (spinner) shows no reason. No export, prop or type member is added. The author-written reason beside the predicate is a spec key for objectstack, not part of this PR. ## Tests (head `e165c34`) - New pins, three files: `page-header-disabledReason-11811.test.tsx` (7), `record-quick-actions.disabledReason-11811.test.tsx` (7), `DeclaredActionsBar.disabledReason-11811.test.tsx` (7). - **Disabled by the predicate:** the button is disabled and described by the reason; hover opens `role=tooltip`; Tab reaches the trigger, which opens it. - **Controls:** a predicate that does not hold gives no reason; a running or loading button gives no reason; the header's host Edit and Delete with `disabled: true` give no reason. - **Overflow:** the item shows the reason and is described by it. - **Text:** the zh pack and the en pack. - **Fixture:** the predicate is the served CEL envelope. The bare string goes to the legacy evaluator, where `has()` faults and a fail-soft `disabled` greys out both rows, so it would pin the wrong reason. - `vitest run packages/components/` plus the 36 app-shell, console and example files that name `page:header`, `PageHeaderRenderer` or `headerSystemActions`, at `e165c34`: `Test Files 401 passed | 1 skipped (402)`. - `vitest run packages/plugin-detail/ packages/i18n/` plus the 27 app-shell and console files that name `DeclaredActionsBar`, `record-quick-actions` or the key, at `e2a498f`: `Test Files 356 passed | 1 skipped (357)`. `e165c34` changes no file in those packages. - type-check: `@object-ui/components` exit 0 at `e165c34`; `@object-ui/i18n`, `@object-ui/plugin-detail` and `@object-ui/app-shell` exit 0 at `e2a498f`, on dists rebuilt from this branch. - **Ablations**, run after commit through objectstack's `scripts/ablation-replace.mjs`: the anchor `if (!disabledReason) return button;` is replaced by an unconditional `return button;`. - Header: `Tests 4 failed | 3 passed (7)`. Red: description, hover, focus, zh. Green: overflow and both controls. - Quick actions and `DeclaredActionsBar`: `Tests 10 failed | 4 passed (14)`. Red: description, hover, focus, zh, en on each. Green: the controls. - Each restore was proven: blob equals HEAD and `git diff HEAD` is empty. - Gates at `e165c34`, all exit 0: - i18n: `check:i18n-keys`, `check:i18n-drift` (1 key added, no en value changed), `check:i18n-dead-keys` (report; the new key is not a candidate), `check:i18n-designer-parity`; - changeset: `check-changeset-presence` (4 released packages, 1 changeset), `check-changeset-no-major`, `check-changeset-fixed`, `check:changeset-claims`, `check:pending-changeset-literals`; - other: `check:control-bytes`, `check:test-path-roots`, `check:new-line-citations` (0 new); eslint on touched files (0 errors). - **Not run locally:** the full app-shell suite. The named-file sets above cannot see a suite that draws these renderers through a page tree without naming them; CI runs the full farm. ## Changeset `.changeset/11811-disabled-action-reason.md`: `@object-ui/i18n` and `@object-ui/components` are `minor`, `@object-ui/app-shell` and `@object-ui/plugin-detail` are `patch`. The fixed group publishes all of them at `minor`, never `major`. The body names the three surfaces and the one key. ## Acceptance notes - The generic action renderers (`action:button` / `action:icon` / `action:group` / `action:menu`) have the same gap. They are their own family card, objectui#11839, and this PR leaves them alone. - A `disabled` predicate that faults (fail-soft) also greys the action out, and now also shows the generic reason. That is consistent, since the action is unavailable either way, but the wording slightly overstates the case for a broken predicate. - `InlineEditSaveBar`'s Save button sets `title` on a disabled Button, which cannot show for the pointer-events reason above. The approval-lock band already shows that reason on screen. Noted only. - Overlap: objectstack-ai#11836 adds keys to the same ten packs. Whoever lands second merges `main`. The session behind this change is `https://claude.ai/code/session_01CGZy1BGCjdN5cXqL9cnvB8`. --- _Generated by [Claude Code](https://claude.ai/code/session_01CGZy1BGCjdN5cXqL9cnvB8)_ --------- Co-authored-by: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #11810
Clause-②: yes
The claim was amended to
Clause-②: yesin comment 6048182477, because the two new language-pack keys below widen@object-ui/i18n's published packs. Implemented by the dispatched dev on this branch, sessionsession_01CGZy1BGCjdN5cXqL9cnvB8.What changes
List view (
@object-ui/plugin-list,ListView'sfilterFields). The Filter panel's field list was the object definition's field map verbatim, and the served map puts the injected system columns first. Now:hidden: true(organization_id,owning_business_unit_id, the__searchcompanion) is not offered, unless the view names it infilterableFieldsor a condition the panel already holds filters on it (a filter restored from the per-user cache, possibly saved while "Add filter" still defaulted to Organization; without this its field trigger would draw blank);effectiveFields: after FLS, user-hidden columns andfieldOrder), then the other business fields, then system fields (the sharedisSystemManagedFieldfrom@object-ui/types, which reads the spec'ssystemflag), then a hidden field kept by the rule above. The definition's own order holds inside a tier.addConditionreads the first entry offields), so a new condition now starts on the view's first visible column. No default is chosen anywhere else.Builder (
@object-ui/components,FilterBuilder). The operator dropdown mountsofferedOperatorsForRow(module-private):operatorsForFieldTypeminusis_null/is_not_nullon a column whose type the spec'sexpandEmptyOperatorratesnull_only. The row's own operator is always mounted, so a storedis_nullstill loads as "Is null".operatorsForFieldType(what a row can HOLD) is unchanged.Hint (
@object-ui/components+@object-ui/i18n). Where both pairs stay, the operator list ends with one line saying how they differ. The operator labels are interpolated, not restated.The per-type table
Read from
@objectstack/specexpandEmptyOperator(17.7.0), the function every server face expands$emptywith. Not a local list.typealone)null_onlytext''multi_value[]The pair that stays on
null_onlyisis_empty/is_not_empty: every bucket that has an empty check offers it, so one label means one thing across columns, and every dialect the builder writes lowers it to$empty(Mongo criteria inFilterConditionField,ViewFilterRulein saved views and the dataset inspector, the ASTisemptyin the live grid).Callers of
FilterBuilderplugin-listListView(end-user list toolbar): field list, order and default change here; operator offer as above.fieldsFilterConditionField(sharing-rule criteria): offer narrows the same way. It loses no predicate: on anull_onlycolumn{ $empty: true }matches what{ $null: true }matches. A stored$nullreads back asis_nulland stays mounted.app-shelldataset inspector andmetadata-adminwidgets,config-field-renderer, the SDUIfilter-builderrenderer: same offer change. The dataset read-back (builderHolds,readBackOperator) asksoperatorsForFieldType, which is unchanged, so a stored$nullon a select column stays editable in the visual builder. Control: thedatasetFilterCondition*suites pass unchanged.Public surface
No export, prop or type member is added.
@object-ui/componentsraises its@objectstack/specfloor from^17.0.0to^17.5.0, because its published entry now importsexpandEmptyOperator, which the spec first exports in 17.5.0 (pnpm-lock.yamlmoves only that importer's specifier line).offeredOperatorsForRow,emptyCheckArmand the hint renderer are module-private. Two NEW language-pack keys in all ten packs:filterBuilder.emptyCheckHint.textandfilterBuilder.emptyCheckHint.list.Changeset
.changeset/11810-filter-builder-fields.md:@object-ui/i18nminor(the added keys, as in objectui#11645),@object-ui/componentsminor(it ships the hint that reads them),@object-ui/plugin-listpatch(its part is a fix). All three are in thefixedgroup, so they are versioned together at the group's highest level,minor.pnpm exec changeset statuslists the group atminor, with no mixed-changeset error.Tests
Head
24a52a4. All runs go from the worktree root through the shared verify lock. The code union below ran on3c5fe54. Since then no source file has changed:78fb73bsets the changeset levels, and24a52a4raises@object-ui/components'@objectstack/specfloor (git diff --stat 78fb73b 24a52a4: thepackage.jsonrange, thepnpm-lock.yamlimporter specifier line, one changeset sentence).On
24a52a4:node scripts/check-spec-range-floors.mjs(check:spec-floors), run after building@object-ui/componentsand@object-ui/plugin-listwith their dependencies: 0floor-too-low, and both packages are judged clean ("Inspected 21 of 39 published package(s) — 823 artifact file(s) … 304 (subpath, symbol) pair(s) judged"). The run still exits 1, but only for 14no-artifactfindings: packages this branch does not touch and that were not built locally (app-shell, auth, cli, layout, plugin-chatbot, plugin-dashboard, plugin-designer, plugin-detail, plugin-form, plugin-gantt, plugin-grid, plugin-timeline, plugin-tree, plugin-view). Control on78fb73b, same artifact, before the raise: "@object-ui/components[floor-too-low] packages/components/dist/index.js referencesexpandEmptyOperatorfrom @objectstack/spec/data, which @objectstack/spec@17.0.0 does not export", and the gate suggested^17.5.0.pnpm --filter @object-ui/components run type-check0.pnpm check:changeset-claims0 ("No pending changeset names a file this change touches.").pnpm check:pending-changeset-literals0.node scripts/check-changeset-presence.mjs0.pnpm changeset:check0.pnpm check:lockfile-integrity0 ("VERDICT clean — no @objectstack/* identity moved backward and no package gained a copy.").pnpm check:lockfile-dedupe0 ("VERDICT deduped").On
78fb73b(changeset levels), everything that reads a changeset was re-run:pnpm check:changeset-claims0, "Every package declared across those 1 body(ies) is either not negated …";pnpm check:pending-changeset-literals0, "No test source names a pending changeset.";node scripts/check-changeset-presence.mjs0, "16 source file(s) of 3 released package(s) changed, and this change declares 1 changeset(s)";node scripts/check-changeset-no-major.mjs0, "No changeset declares amajorbump.";node scripts/check-changeset-fixed.mjs(theChangeset Fixed Group Checkjob) 0, "All workspace packages are in the changeset fixed group." / "privatePackages declared: version=true, tag=false.";pnpm changeset:check0;pnpm exec changeset status0.vitest runover the seven changeset-reading script suites (body-dialect-census,changeset-polarity-census,check-changeset-claims,check-changeset-fixed,check-changeset-no-major,check-changeset-overwrite,check-changeset-presence):Test Files 7 passed (7),Tests 299 passed (299),VERDICT command-exit 0.On
3c5fe54:pnpm exec vitest run --maxWorkers=2 packages/components/ packages/plugin-list/ packages/i18n/plus the cross-package filter suites (FilterConditionField*,datasetFilterCondition*,filter-builder-protocol-ids-census-9306,drill*Dialect*,drillEmptyBucketEscapeHatch-9159,view-operator-builder-parity,filter-builder-mirror-6939):Test Files 577 passed | 1 skipped (578),Tests 6746 passed | 37 skipped (6783),VERDICT command-exit 0.pnpm --filter @object-ui/components --filter @object-ui/i18n --filter @object-ui/plugin-list run type-checkafterturbo run build --filter='@object-ui/plugin-list^...': exit 0 (all three echoedtsc --noEmit && tsc -p tsconfig.test.json; the test projects include the new test files).pnpm exec eslinton the 16 touched TS files: 0 errors; warnings equal to the base file forListView.tsx(188 = 188) andfilter-builder.tsx(19 = 19), 0 on the new test files.check:control-bytes,check:test-path-roots,check:changeset-claims,check:pending-changeset-literals,check:i18n-keys,check:i18n-drift,check:i18n-dead-keys,check:i18n-designer-parity,check:vi-mock-specifiers,check:vi-mock-inherit,check:vi-mock-override-shape,check:new-line-citations,check:phantom-deps,check:esm-specifiers,scripts/check-changeset-presence.mjs,scripts/check-changeset-no-major.mjs.check:eager-closureandcheck:eager-locale-catalogues(PREREQUISITE NOT MET: they read a built console bundle; declared to CI). Marginal cost measured instead:@objectstack/spec/datais already in the eager closure throughrenderers/layout/containers.tsx(isFileIdToken), and an esbuild probe of that import with and withoutexpandEmptyOperatormoves 641429 to 641645 bytes minified (+216), 152073 to 152167 gzip (+94).Pins
packages/plugin-list/src/__tests__/ListView.filterFieldList-11810.test.tsx— the Tasks definition rebuilt from the spec'sinjectedSystemColumnDefs,task.object.ts's declared fields and the__searchcompanion (25 fields, lit control): no hidden field; order columns, then business fields, then system fields; "Add filter" starts on "Title";fieldOrderis followed; a held condition onorganization_idkeeps it last and its trigger reads "Organization";filterableFieldsnaming a hidden field keeps it.packages/components/src/__tests__/filter-builder-empty-null-pair-11810.test.tsx— the arm table read from the spec (lit control); ninenull_onlytypes offer one pair and no hint; text types offer both pairs and the blank-text hint; list types both pairs and the empty-list hint; a storedis_nullon a select shows "Is null" (control); a stored camelCaseisNotNullloads; a text row on "Is null" switched to a select keeps it;operatorsForFieldTypestill carries both pairs (author-facing callers' hold set, control).filter-builder-operator-alias-trigger-7561.test.tsx: the recorded number-column list loses "Is null" / "Is not null".de-quote-pairing-3876.test.ts: the census moves 71 to 75 for the hint's four quoted spans.Ablations (one-off, committed tree, restored by
git checkout HEAD --, restore proven by blob hash equal to HEAD and an emptygit diff HEAD)Direction predicted before each run: red.
filterFieldsreturnscandidateFieldsunfiltered: 5 of 6 field-list pins red, the fixture control green. "Add filter" readsOrganizationinstead ofTitle, the issue's symptom.offeredOperatorsForRowreturns the full bucket: 11 red (the ninenull_onlytypes, the stored-is_nullpartner check, the 7561 recorded list), 52 green.The first attempt at ablation 1 was a no-op: the landing check counted a multi-line anchor with
grep -cF(one pattern per line) and refused before any test ran. It was re-run with a single-line anchor and a marker count (0 to 1) as the landing proof.Acceptance notes
multiple: truelookups / users / selects. The builder's field descriptor carriestypeonly, so such a column is judged by its type (null_only) and offered "Is empty" / "Is not empty" without the null pair. On the server it ismulti_value, and "Is empty" there matches null or[], which is the reading an end user means. Offering both pairs with the list hint there would needmultipleon the descriptor, a new type member. Not done here.filterableFields(the spec calls it "fields enabled for end-user filtering", an explicit author choice), and one a held condition filters on (otherwise its field trigger draws blank). Both go last.candidateFields(relational fields excluded, the platform's sortability signal applied). Whether it offers the hidden__searchtext column on a served deployment was not measured. Carrier: none.Generated by Claude Code