Repository navigation
feat(plugin-grid): server-side grid grouping — group headers from the query, rows paged per group (objectui#7189) - #10878
Conversation
…oup row pages (objectui#7189) A grouped grid that owns its fetch now asks the server for its group set and every header number (compileListViewGroupQuery, one query per depth, through the new optional DataSource.queryGroupHeaders) and pages each open group's rows (compileListViewGroupRowsQuery). useGroupedData builds the tree from the header rows instead of bucketing a fetched page; the grouped exclusion in useServerPagination goes for such a grid; the Partial marker no longer renders where counts are server-true. ListView hands a grouped grid its own fetch (and the effective filter) when the data source serves group headers. Claude-Session: https://claude.ai/code/session_01MCg3z5cGrV51xEddpGcxEY Co-authored-by: Claude <noreply@anthropic.com>
…86-row five-unit fixture (objectui#7189) Claude-Session: https://claude.ai/code/session_01MCg3z5cGrV51xEddpGcxEY Co-authored-by: Claude <noreply@anthropic.com>
…sections; changeset; ListView relay pins (objectui#7189) Claude-Session: https://claude.ai/code/session_01MCg3z5cGrV51xEddpGcxEY Co-authored-by: Claude <noreply@anthropic.com>
…ed under; type the server-grouping hook (objectui#7189) Claude-Session: https://claude.ai/code/session_01MCg3z5cGrV51xEddpGcxEY Co-authored-by: Claude <noreply@anthropic.com>
…ers, not every render (objectui#7189) Claude-Session: https://claude.ai/code/session_01MCg3z5cGrV51xEddpGcxEY Co-authored-by: Claude <noreply@anthropic.com>
github-actions
Bot
added
documentation
|
changeset-claim-re-read
|
✅ 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
|
Contract reviewServed-tier: 233/233 Read: card objectui#7189 body and thread via the REST API (ruling ① Derived judgments
② Semver level
③ Boundary flags
Implemented-by: VERDICT: PASS |
…state GroupEntry's required members (objectui#7189) Claude-Session: https://claude.ai/code/session_01MCg3z5cGrV51xEddpGcxEY Co-authored-by: Claude <noreply@anthropic.com>
✅ 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
|
Contract reviewServed-tier: 138/138 Read: the previous record for ① Derived judgments
② Semver levelUnchanged from the previous record: ③ Boundary flags
Implemented-by: VERDICT: PASS |
…ts name 30 objectui issues that answer 404 (objectui#10803, batch 4) (objectstack-ai#10875) Part of objectstack-ai#10803 Clause-②: no Dispatched implementation of the `domain:ui` seat 2 claim (comment `5861571015`) on objectui#10803, batch 4, session `https://claude.ai/code/session_014mXUNuFomfj24w7s1pZzhN`. Citations only: no sentence's claim moves, and every edited pending changeset's frontmatter is byte-identical. The runtime text that moves loses its dead pointer and nothing else: the legacy `ActionSchema` `onSuccess` / `onFailure` tombstone guidance strings open `RETIRED (ADR-0049) —` (ruling `5861129870` as amended by `5861311219`), and two `DataTableSchema` zod `.describe()` strings drop their pointer (amendment `5860244997`, Q1 = A), all in `@object-ui/types`. The one test that pins the tombstone strings verbatim follows them, and nothing else in it moves. **Patch round 1** answers contract review `5861897723` (FAIL on `d21aeb0f6`) under seat ruling `5861900779` (route a: tests that pin changed comment text follow it, anchor only). `Test (shard 6/8)` was red on `layout-default-jsdoc-7361`, whose `TextSchema.variant` pin anchored on `6942`. The round does four things: - Three comment-text anchors follow the re-pointed text: the one in `layout-default-jsdoc-7361` and two in `cellClassNameCensusProse-6921`. - The `MarkdownSchema` docblock stops crediting `8063bcbdc` with retirements it never touched (**Special cases** 9). - The `7004-cli-root-path-line.md` remnant "(the mechanical half)" goes. - `main` (`de1b879a6`) is merged in by a merge commit (`ae0b81089`), with no rebase and no force-push. The round's anchor sweep (**Special cases** 10) shows those three pins are the only anchors on changed text. ## Why `Part of`, not a closing line The card stays open for the other 53 numbers of the family remainder (seat amendment `5860244997`, Q2 = A), listed in **Acceptance notes** 1, and for the bare-number 404s the C0 census below found. ## Premise, re-measured - REST `GET /repos/objectstack-ai/objectui/issues/N` for each of the 30: all 30 answer 404, and a second read of each answers 404 again (30 of 30). `GET .../pulls/N` answers 404 for all 30 too. Lit controls: objectui#10533 and objectui#7714 answer 200. No number was dropped from the batch. - The same 30 in objectstack all answer 200, with subjects unrelated to the objectui sentences (for example objectstack#6882 is an `examples/app-todo` trigger finding, while the objectui sentences are about `DataTableSchema.renderCellEditor`). Two exceptions were read, not guessed: - `objectstack#6888` already appears on 5 in-scope lines (`i18n.ts`, `ActionDefaultInspector.tsx`, `ActionPreview.tsx`, `block-config.ts`, `action-bar.tsx`), each about the `global_nav` retirement, which is objectstack#6888's own subject. They are live sister citations, untouched. - `UnpublishedAppBar.tsx` wrote "framework PR objectstack-ai#6942" bare. objectstack#6942 is the pull request "the ADR-0045 publish gate gets its own machine-managed key — `app.hidden` goes back to meaning navigation", exactly the `_unpublished` / `hidden` split that heading describes. It is re-qualified to `objectstack#6942` (Zone 1 item 5). - Every other site names an objectui card or pull request, confirmed by its landing commit below. - Every edited changeset is pending: it is present in `.changeset/` on `main`. - Branch point: `e32dae160`. `main` was merged in at `de1b879a6` in patch round 1 (merge commit `ae0b81089`). Not a shallow checkout. ## C0: the bare-number census (measurement only; PR objectui#10869 Acceptance note 2) The instrument, over the two in-scope classes at the branch point (a `#` not preceded by a word character, `#`, `&`, `/`, `.` or `-`, so `objectui#N`, `objectstack#N`, `cloud#N`, `hotcrm#N` and every other `word#N` are excluded, as is a `#N` inside a URL): ``` git grep -hoP '(?:^|(?<=[^\w#&/.\-]))#\d+(?![0-9A-Za-z_])' e32dae1 -- '.changeset/*.md' 'packages/*/src/**' ':!**/__tests__/**' ':!**/*.test.*' ':!**/*.spec.*' ':!**/test/**' ':!**/tests/**' | sort -u | wc -l ``` - **1002 distinct bare numbers.** That is over 600, so by the order's budget only the **980** at or below 10900 were read (replace the final `wc -l` with `tr -d '#' | awk '$1+0 <= 10900' | wc -l` to reproduce 980). The 22 above 10900 were NOT read. - Each of the 980 was read once with REST `GET .../issues/N`: **955 answer 200 and 25 answer 404.** Each of the 25 was read a second time and answered 404 again (25 of 25), and `GET .../pulls/N` answers 404 for all 25. - Lit control on the same instrument: objectstack-ai#5026, objectstack-ai#6905 and objectstack-ai#3720 are all in its output at the branch point. At this head none of the eight batch numbers it found bare remains. - The 25, with site counts (lines / files, at the branch point): - **This batch's own numbers (8), all fixed here:** objectstack-ai#6882 (5 / 2), objectstack-ai#6905 (7 / 2), objectstack-ai#6907 (5 / 1), objectstack-ai#6942 (1 / 1, the objectstack pull request above, re-qualified), objectstack-ai#6959 (1 / 1), objectstack-ai#6962 (1 / 1), objectstack-ai#7036 (8 / 1), objectstack-ai#7087 (2 / 2). The printed census below already counts a bare `#N` for these. - **Already on the family list (3), for their own batches:** objectstack-ai#7620 (2 / 2), objectstack-ai#7678 (3 / 3), objectstack-ai#7853 (1 / 1). - **New to the family (11), for later batches:** objectstack-ai#3720 (5 / 3), objectstack-ai#5420 (1 / 1), objectstack-ai#5503 (1 / 1), objectstack-ai#5506 (1 / 1), objectstack-ai#5737 (4 / 4), objectstack-ai#6467 (1 / 1), objectstack-ai#6936 (1 / 1), objectstack-ai#6945 (1 / 1), objectstack-ai#7622 (1 / 1), objectstack-ai#7662 (1 / 1), objectstack-ai#7684 (2 / 2). - **Not citations (3):** `#0` (257 / 187, the `#0` of "AGENTS.md #0.1"), `#000` (2 / 2) and `#000000` (10 / 6), which are CSS colours. - **Folded here: none.** No C0 404 outside this batch's 30 sits in a sentence this batch edits: none of the 11 new numbers, nor any of the other 53 family numbers, occurs in any hunk of this diff (3 lines of context). - The 200 set is not a citation census: many bare numbers are decision-batch, summons or commandment numbers, or React error codes, that happen to name a live issue. ## Census (the enumeration pin for this batch) The instrument PR objectui#10854 printed and PR objectui#10869 reused, with this batch's 30 numbers substituted (REF = a commit or tree): ``` git grep -nE '(objectui#|#|issues/)(6872|6879|6882|6887|6888|6889|6892|6905|6907|6921|6923|6924|6931|6940|6942|6958|6959|6962|6965|7004|7008|7023|7025|7036|7068|7069|7077|7079|7087|7088)([^0-9]|$)' REF -- '.changeset/*.md' 'packages/*/src/**' ':!**/__tests__/**' ':!**/*.test.*' ':!**/*.spec.*' ':!**/test/**' ':!**/tests/**' | grep -v 'objectstack#' | wc -l ``` - REF = `e32dae160` (branch point): **170** lines, 48 changeset lines in 38 files and 122 src lines in 45 files. REF = `de1b879a6` (the `main` merged in): 170. REF = `c6678b1bd` (`main` at the round's final read): 170. - REF = `ae0b81089` (this head, `de1b879a6` merged in): **0**. Round 0's head `d21aeb0f6` also read 0. - This head merged with a fresh `main` (`c6678b1bd`, `git merge-tree --write-tree`, clean, tree `e109a3bd0`): **0**. - **The instrument's blind spot, as batch 3 measured it.** Its `grep -v 'objectstack#'` drops a whole line that also names an objectstack number. Three in-scope lines were hidden that way, each naming objectui#6965 beside objectstack#9343: in `10039-publish-drafts-advisories.md`, `6965-batch-publish-advisories.md` and `render-publish-advisory-findings-5026.md`. They are re-pointed here too, so the batch's true base population is 173 lines (51 changeset lines in 40 files, 122 src lines in 45 files). Without the filter the instrument reads 178 at the branch point and **7** at this head, all live: the five `objectstack#6888` lines, `objectstack#6942` in `UnpublishedAppBar.tsx`, and the same spelling in the new sweep changeset. - Other spellings at this head: a search for the 30 numbers with no `#` finds only filenames (`action-callback-retired-7068.test.ts`, `data-table-declared-keys-6882.test.ts` and the like) and two SVG path coordinates. Hex-colour false positives (a number followed by a hex letter) at the branch point: 0. - Lit control, the same printed instrument over live objectui#7714 at this head: 17 lines. - **Out of scope, as it stands** (the instrument with the 30 numbers, unfiltered, whole tree at this head): 226 test lines in 96 files (`scripts/__tests__` included), 22 scripts lines in 5 files, 0 `.github`, 0 governed, and 18 other lines in 10 files: published `CHANGELOG.md` history, three `content/docs` lines, and four lines of the private `@object-ui/test-support` README. See **Acceptance notes** 3. ## Citation form - The 9-character backticked sha of the commit on `main` that landed the change the sentence rests on, as in PRs objectui#10707 / objectstack-ai#10766 / objectstack-ai#10797 / objectstack-ai#10854 / objectstack-ai#10869. All 24 distinct shas below are ancestors of `main`: `git merge-base --is-ancestor`, exit 0 each. Control leg in the same checkout: the head of PR objectui#10714 (`d2afdb6bf`) answers exit 1, and a known ancestor, `5f789538d`, answers exit 0. `git rev-parse --short=9` returns the same 9 characters for each. - A changeset's pointer to its OWN card is dropped, not replaced: `@changesets/changelog-git` prefixes each released entry with the hash of the commit that added the file, which is that landing. Where dropping the pointer left a dangling word, the clause was minimally repaired (**Special cases** 5). - Where the sentence rests on a RULING, the ruling is cited by its date, next to the sha that carried it out where one exists (**Special cases** 4). - Where a dead card sits beside a live pull request that IS its landing, the dead card is dropped and the live pointer stays (**Special cases** 1), as batch 3 dropped the dead predecessor in "(objectui#5401 → objectstack-ai#5454)". - Where the claim lived only on the dead card (a finding, a triage, a deferral, a quoted ruling), the sha only locates the card: "the card behind SHA" (**Special cases** 3). This is review `5861306588`'s lesson: a sha must carry what the sentence cites it for. - Where nothing answers, the pointer is dropped and the sentence names the card by role (objectstack-ai#7023 only). - Runtime text carries no sha: see **Special cases** 7. ## Mapping (number to resolution) Lines / files are the branch-point census for that number (a line naming two batch numbers counts for each, and the three hidden objectstack-ai#6965 lines are included). "Method" is how the landing was found. The last column is why that commit carries what the sentence cites it for. | dead number | resolution | method | lines / files | why the commit carries it | |:--|:--|:--|:--|:--| | objectstack-ai#6872 | own-card pointer dropped (landing `d9a0490b7`) | changeset's adding commit | 1 / 1 | the only site is the objectstack-ai#6872 changeset's own pointer | | objectstack-ai#6879 | `3619792bf` | changeset's adding commit, `git log --grep` | 1 / 1 | it curates `box` into `PUBLIC_BLOCKS`, the list the comment sits in | | objectstack-ai#6882 | `bf97b98c8` | changeset's adding commit | 20 / 7 | it declares `renderCellEditor` and schema-level `cellClassName` on `DataTableSchema`, adds the `Equal` exact-shape pin, and corrects `cellClassName`'s reach to the three utility cells | | objectstack-ai#6887 | `5c09cca27` | changeset's adding commit | 1 / 1 | it derives metadata-viewer's option shape from the spec's `SelectOption` with the `Omit` the comment names | | objectstack-ai#6888 | `320374d2a` | changeset's adding commit | 9 / 7 | it keys `LocationField`'s residue refusal and the two coordinate nouns into the locale packs | | objectstack-ai#6889 | `f75810e7c` | changeset's adding commit, `git log -S` | 8 / 1 | it parses a dropped-fields entry's `fields` and `object` instead of asserting them, the gate the comments describe | | objectstack-ai#6892 | own-card pointers dropped (landings `8d40c18a7`, `951fa8e0d`) | changesets' adding commits | 2 / 2 | both sites are slice changesets' pointers to their own card | | objectstack-ai#6905 (a pull request) | `85f6a6097` | `git log --grep`: its squash subject ends "(objectstack-ai#6905)" | 8 / 2 | it judges gantt dates by TYPE before `new Date`, the "objectstack-ai#6781 type rule" the sentences name | | objectstack-ai#6907 | `7fc5c3c12` | changeset's adding commit, `git log -S` | 7 / 2 | it spells a refused gantt date by a rule, and its changeset records the measured throwing-getter crash | | objectstack-ai#6921 | `4eb665bcf` | changeset's adding commit, `git log --grep` | 2 / 2 | it pins, in the rendered DOM, which body cells schema-level `cellClassName` reaches | | objectstack-ai#6923 | `d3bf4fa6f`; "the 2026-08-31 ruling" | changeset's adding commit, `git log --grep` | 5 / 4 | it gives the wrapper-key list its build-free JSON home "per the 2026-08-31 ruling on objectui#6923" | | objectstack-ai#6924 | `78e98bf44`; PR objectstack-ai#7024 kept | changeset's adding commit, `git log --grep` | 5 / 3 | it adds the one enum-options walk for the top-level reader family | | objectstack-ai#6931 | `8063bcbdc` | changeset's adding commit, `git log -S` | 8 / 5 | it converts the remaining eleven tombstones, `confirm` and the menu divider `type` included, to `retirementTombstone()`. At the `MarkdownSchema` site, which `8063bcbdc` never touched (`446d93d4e` retired those keys), it is cited only as the commit that made that spelling uniform: see **Special cases** 9 | | objectstack-ai#6940 | pointers dropped (landing `0c386dd7f`) | changeset's adding commit | 2 / 2 | the only sites are its own changeset's pointer and a `.describe()` string | | objectstack-ai#6942 | `57f9b077b`; one site re-qualified to objectstack#6942 | `git log --grep`; REST for the sister PR | 6 / 4 | it makes `ui:text` honour the published `variant` enum, with a node that omits the key left without a typography class | | objectstack-ai#6958 | `6a449fc49` | changeset's adding commit | 6 / 5 | it clears a field its own `visibleWhen` hides | | objectstack-ai#6959 | `c0c436db8` | changeset's adding commit | 2 / 1 | it keeps the find-call gate and adds a second one, and records the trap that a mock call is one resolution ahead of its value | | objectstack-ai#6962 | `7b433197d` | changeset's adding commit | 3 / 2 | it removes `publishDraft`'s envelope unwrap after measuring the route at the producer | | objectstack-ai#6965 | `ce986aafc`; PR objectui#10038 kept at three sites | changeset's adding commit, `git log --grep`; its body names the card it settles | 21 / 9 | it routes "publish whole app" through the advisory seam, adds `publishPackageDrafts`, and writes the "Corrected before release" note | | objectstack-ai#7004 | `85b495795` (root Path line); `a5d55472b` (arm selection) | both changesets' adding commits | 7 / 4 | the first gives a root-level issue a Path line; the second prints the arm the document selected, per the 2026-09-02 ruling | | objectstack-ai#7008 | `f08bcd9af` | changeset's adding commit | 15 / 11 | it makes `FieldEditWidget` deliver the declared NON-DOM block through `toHostProps`, and corrects `RequiredFieldsDialog` | | objectstack-ai#7023 | nothing answers: the card named by role | `git log --grep` finds only `d88e20f55`, the objectstack-ai#4895 retirement it dissolved into | 1 / 1 | see **Special cases** 3 | | objectstack-ai#7025 | own-card pointer dropped (landing `14884620e`) | changeset's adding commit | 1 / 1 | the only site is its own changeset's pointer | | objectstack-ai#7036 | `869b876c8` | `git log -S` on the timeline renderer | 11 / 1 | it scopes the speller's totality claim, and refuses a `catch` because it would substitute `an object` for a failure | | objectstack-ai#7068 | `6bca0e4e8`; runtime text `RETIRED (ADR-0049) —` | changeset's adding commit | 8 / 3 | it retires the callback pair as ADR-0049 tombstones and deletes `ActionCallback` / `ActionCallbackSchema` | | objectstack-ai#7069 | `2760075ff` | changeset's adding commit | 3 / 3 | it adds the fifth parity ledger that watches the mirror-wider-than-declared direction | | objectstack-ai#7077 | own-card pointer dropped (landing `00d3f09c5`) | changeset's adding commit | 1 / 1 | the only site is its own changeset's pointer | | objectstack-ai#7079 | own-card pointer dropped (landing `ab7dc31ce`) | changeset's adding commit | 1 / 1 | the only site is its own changeset's pointer | | objectstack-ai#7087 | `c93b4d5f3` | changeset's adding commit | 8 / 6 | it drops the 18 `disabled?: boolean` narrowings, adds the twin-symmetry pin, and keeps the six independent `boolean` declarations | | objectstack-ai#7088 | `c1fe272ad` | changeset's adding commit | 4 / 3 | it corrects `BaseSchema.hidden`'s JSDoc to the single hide path and records the synonymy | ## Special cases (the judgement calls) 1. **A dead card beside its own live landing.** - "objectui#6965 / PR objectui#10038" (`AiChatPage.tsx`, `PendingDraftsBar.tsx`, `StudioDesignSurface.tsx`) becomes "PR objectui#10038". PR objectui#10038 answers 200 and is the landing of `ce986aafc`, so the dead half goes and the live half stays. - `7025-nonoptional-enum-cast.md`: "the same walk objectui#6924 converged the optional-cast family onto in PR objectstack-ai#7024" becomes "the same walk PR objectstack-ai#7024 converged the optional-cast family onto". PR objectstack-ai#7024 answers 200 and is `78e98bf44`'s pull request. 2. **The sister-repo number.** In `UnpublishedAppBar.tsx`, "(objectstack#4829 A1, framework PR objectstack-ai#6942)" becomes "(objectstack#4829 A1, framework PR objectstack#6942)". See **Premise**. 3. **The card named by role.** - **objectstack-ai#7023**, in `4895-retire-block-schema-family.md`: "and objectui#7023 — the narrower validator-only fix — dissolves into this retirement" becomes "and the separate card for the narrower validator-only fix dissolves into this retirement". That card never landed on its own. - **"The card behind SHA"**, where the sentence rests on something that lived only on the card: - `6755-field-diagnostics-i18n.md`: "objectui#6888 carries it" becomes "The card behind `320374d2a` carries it". Carrying a pending item is the card's act. - `data-objectstack` `index.ts`: "Unlike objectui#6889's exotic case" becomes "Unlike the exotic case on the card behind `f75810e7c`". The exotic case is not in the commit. - `toHostProps.ts`: "the specific mistake objectui#7008's ruling fences off" becomes "the specific mistake the ruling on the card behind `f08bcd9af` fences off". The quoted ruling text is the card's, not the commit's. - `zod-wrapper-keys.ts`: "the wall objectui#6923 was filed to get a ruling on" becomes "the wall the card behind `d3bf4fa6f` was filed to get a ruling on". - `base.zod.ts`: "objectui#7069 called this repo's systematic producer" becomes "the card behind `2760075ff` called this repo's systematic producer". The commit's own diff attributes that phrase to the card. - The timeline renderer names objectstack-ai#7036's deferral, triage and "1 of 6" trade. These become "the file surface the card behind `869b876c8` deferred", "the triage of the card behind `869b876c8`" and "the card behind `869b876c8` was stopped from making". None of the three is in the commit. 4. **Rulings by date.** - `union-arm-diagnostics.ts`: "the 2026-09-02 ruling on objectui#7004 asks for" becomes "the 2026-09-02 ruling `a5d55472b` implements asks for". - `validate.ts`: "The ARM-SELECTION half of objectui#7004 landed on the 2026-09-02 maintainer ruling" becomes "The ARM-SELECTION half (`a5d55472b`) landed on the 2026-09-02 maintainer ruling". - `7201-schema-slot-census-gate.md`: "across objectui#6882, the ruling that expired both holds" becomes "across the 2026-08-30 ruling that expired both holds (`bf97b98c8`)". - `7687-combobox-option-disabled.md`: "by the objectui#7087 twin-symmetry ruling" becomes "by the 2026-09-01 twin-symmetry ruling, `c93b4d5f3`". - `defaults-table-scan.ts`: "the one objectui#6923 already ruled for exactly that wall" becomes "the one the 2026-08-31 ruling (`d3bf4fa6f`) already set for exactly that wall". 5. **Own-card pointers that needed a word.** - `6931-tombstone-guidance-remainder.md` and `6888-location-residue-refusal-keyed.md`: the pointer sat on its own line, so the period moved up with it. - `7087-disabled-twin-symmetry.md`: the parenthesis opened on the dead card, so the line break moved with it. - `7004-cli-root-path-line.md`: "(objectui#7004, mechanical half)" is dropped, so the sentence ends "sits at the document root." (patch round 1 took the review's nit), and "left open on objectui#7004 for a maintainer ruling" becomes "left open for a maintainer ruling". - `shared-zod-wrapper-keys-6923.md`: "objectui#6923: the Zod wrapper-key list" becomes "The Zod wrapper-key list". - `vi-mock-inherit-slice2.md` and `slice3.md`: "(objectui#6892 slice 2)" becomes "(slice 2)", and likewise for slice 3. 6. **The totality sequence in the timeline renderer.** "(objectstack-ai#6759 -> objectstack-ai#6905 -> objectstack-ai#6907 -> objectstack-ai#7027)" becomes "(objectstack-ai#6759 -> `85f6a6097` -> `7fc5c3c12` -> objectstack-ai#7027)", and likewise with "-> objectstack-ai#7036" as `869b876c8`. objectstack-ai#6759 and objectstack-ai#7027 answer 200 and stay. objectstack-ai#6905 was the pull request whose squash is `85f6a6097`. "⚠️ objectui#7036 — READ THE TWO PARAGRAPHS" becomes "⚠️ `869b876c8` — READ…", and the "see the objectui#7036 note below" that points at it follows. 7. **The runtime strings.** Only the listed text moves in each. | file | member | before | after | |:--|:--|:--|:--| | `types/src/zod/crud.zod.ts` | `ActionSchema.onSuccess` (tombstone) | "RETIRED (objectui#7068) — `onSuccess` is no longer part of this legacy ActionSchema; ..." | "RETIRED (ADR-0049) — `onSuccess` is no longer part of this legacy ActionSchema; ..." | | `types/src/zod/crud.zod.ts` | `ActionSchema.onFailure` (tombstone) | "RETIRED (objectui#7068) — `onFailure` is no longer part of this legacy ActionSchema; ..." | "RETIRED (ADR-0049) — `onFailure` is no longer part of this legacy ActionSchema; ..." | | `types/src/zod/data-display.zod.ts` | `DataTableSchema.rowActions` `.describe()` | "... mirrors the boolean the renderer truthiness-tests (objectui#6940)" | "... mirrors the boolean the renderer truthiness-tests" | | `types/src/zod/data-display.zod.ts` | `DataTableSchema.cellClassName` `.describe()` | "... so row density has to be set on both (objectui#6882)" | "... so row density has to be set on both" | **The ADR-0049 condition, measured from the retirement itself** (Zone 1 item 4): - the zod comment above the two keys opens "ADR-0049 RETIREMENT TOMBSTONES"; - its landing commit `6bca0e4e8` writes them as tombstones, not deletions, and adds the pending `action-callback-retired-7068.md`, which says "ADR-0049 enforce-or-remove"; - each string already ends "Retired under ADR-0049 enforce-or-remove". Both open `RETIRED (ADR-0049) —`. Their TypeScript twins in `crud.ts` read "RETIRED (`6bca0e4e8`, ADR-0049 enforce-or-remove)", the batch 3 twin form. 8. **What follows the strings, and nothing else in it.** `action-callback-retired-7068.test.ts` pins both guidance strings as full literals (`ON_SUCCESS_GUIDANCE`, `ON_FAILURE_GUIDANCE`), and their openings follow. Its test names and four `@ts-expect-error` comments still read "RETIRED (objectui#7068)". They describe the `?: never` TypeScript face, quote no runtime string, and are out of the card's classes (**Acceptance notes** 3). No test, doc or changeset quotes either `.describe()` string: a whole-tree search for each old string returns only the source line. - Patch round 1: three pins quoted re-pointed COMMENT text, and they follow it, anchor only (ruling `5861900779`, route a). - `layout-default-jsdoc-7361.test.ts`: `toContain('6942')` becomes `toContain('57f9b077b')` (the `TextSchema.variant` docblock). - `cellClassNameCensusProse-6921.test.ts`: `toMatch(/objectui#6882|objectstack-ai#6882/)` becomes `toMatch(/bf97b98c8/)` (the `cellClassName` census entry). - The same file: `toMatch(/REDUNDANT since objectui#6882/)` becomes a match on "REDUNDANT since `bf97b98c8`" (the holds-member docblock). - Their titles, messages and comments stay as they are. 9. **The `MarkdownSchema` docblock (patch round 1, review `5861897723`).** Round 0 wrote "Both refuse BY NAME through `retirementTombstone()` (`8063bcbdc`), with the remedy in the message". `8063bcbdc` never touched `MarkdownSchema`; `446d93d4e` retired `sanitize` / `components`. So the sha did not carry what it was cited for, the objectstack-ai#5738 class. It now reads "through `retirementTombstone()`, the spelling `8063bcbdc` made uniform, with the remedy in the message". That commit's own subject says the remaining tombstones carry their remediation text. The edit is a comment, and the file's C4 print is unchanged by it. 10. **The anchor sweep that closes the gap (patch round 1).** - **Instrument:** every string, template, numeric and regex literal in all 3814 test and script files under `packages/` and `scripts/`, enumerated with `git ls-tree` at this head and parsed with TypeScript. - **Test applied to each literal:** does its occurrence count DROP between `main` and this head in any of the 90 files this PR changes? Raw text and a comment-flattened form are both checked. - **Specific anchors found** (a batch number, a changed phrase, or a pattern keyed to them), after reading every hit against the file it sits in: - the three pins above; - `tombstone-discriminator-agreement-9684`'s /RETIRED \(objectui#/ note search, which reads only `mobile.ts` and is untouched by this PR. - **Everything else:** generic tokens (whitespace, punctuation, `objectui#` spellings in ledger checks over the tests' own data, digit and letter classes), none of them an assertion over text this PR changed. - **Round 0's quote search** covered runtime strings only. That is the gap this sweep closes. ## Held **By the serial rule: nothing.** Round 0's list: objectui#10873, objectstack-ai#10871, objectstack-ai#10870, objectstack-ai#10852, objectstack-ai#10821, objectstack-ai#10780, objectstack-ai#10777, objectstack-ai#10714, objectstack-ai#10278 and objectstack-ai#5400. Since then objectstack-ai#10873, objectstack-ai#10871, objectstack-ai#10870 and objectstack-ai#10821 have merged. objectstack-ai#10870 touched `base.zod.ts` in a region apart from this PR's line, and the merge with it is clean. - Re-read before the round-1 push: 11 open. Besides this PR they are objectui#10880, objectstack-ai#10879, objectstack-ai#10878, objectstack-ai#10876, objectstack-ai#10852, objectstack-ai#10780, objectstack-ai#10777, objectstack-ai#10714, objectstack-ai#10278 and the release PR objectstack-ai#5400. - Four of them hold files this PR edits: - objectui#10879: `types/src/zod/data-display.zod.ts`; - objectui#10878: `data-objectstack/src/index.ts` and `plugin-grid/src/ObjectGrid.tsx`; - objectui#10714: `types/src/data-display.ts`, `types/src/layout.ts` and `types/src/zod/data-display.zod.ts`; - objectui#10278: `plugin-grid/src/ObjectGrid.tsx`. - Each one's hunks were read against its merge-base, with 3 lines of context. **None of the 29 lines this PR changes in those files appears inside or beside any of their hunks.** - Trial merges of each open PR head with this head (`git merge-tree --write-tree`) are clean for objectstack-ai#10880, objectstack-ai#10879, objectstack-ai#10878, objectstack-ai#10876, objectstack-ai#10852, objectstack-ai#10780, objectstack-ai#10777 and objectstack-ai#10714. - objectstack-ai#10278 (`eab4c8e52`) conflicts in `ObjectGrid.tsx`, but it conflicts identically against `main` alone: one conflict hunk in its own page-size region either way. - PR objectstack-ai#5400 (Version Packages) regenerates and is not a hold. ## Changesets - `.changeset/10803-dead-citation-sweep-fourth-batch.md`, EMPTY frontmatter: the comment-only edits in 12 released packages and the private `@object-ui/test-support`. No published behaviour changes through them. It points at the second file for the runtime text. - `.changeset/10803-fourth-batch-runtime-strings.md`, `'@object-ui/types': patch`: the two tombstones open `RETIRED (ADR-0049) —` with everything after the dash unchanged, and the two describe strings lose their pointer. No key, path, issue code, accept set, refusal or severity moves. ## Proof of prose-only (C4), against the merged `main` `de1b879a6` - **Source.** Each of the 48 touched `.ts` / `.tsx` files was parsed at `de1b879a6` and at this head with TypeScript 6.0.3's `createSourceFile` and re-printed by `createPrinter({ removeComments: true })`. The 48 are the 45 non-test sources and three tests: the one pinning the strings, and the two whose comment-text anchors follow. - 43 of 48 prints are identical. That includes all 45 non-test sources except `crud.zod.ts` and `data-display.zod.ts`, whose `MarkdownSchema` comment edit moves nothing. - The other 5 are equal once exactly the 9 listed substitutions are applied to the `main` print, each matched once: - `crud.zod.ts`: the two tombstone strings; - `data-display.zod.ts`: the two describe-string deletions; - `action-callback-retired-7068.test.ts`: the same two tombstone strings; - `layout-default-jsdoc-7361.test.ts`: the one anchor; - `cellClassNameCensusProse-6921.test.ts`: the two anchors. - 0 parse diagnostics. - Lit controls on the same instrument: dropping "(ADR-0049)" from the `onSuccess` string moves the print; re-spacing the "ADR-0049 RETIREMENT TOMBSTONES" comment does not. - **Changesets.** The frontmatter block of every one of the 40 edited changesets is byte-identical at `de1b879a6` and this head (40 of 40, by md5). The overwrite gate below agrees. - **Scope of the diff against the merged `main`:** 90 files, +227 / −184: 40 edited and 2 new changesets, 45 non-test source files, and 3 tests. ## Gates, on this head `ae0b81089` Each line is the gate's own verdict and exit code, captured by redirect-then-`$?`. - `node scripts/check-changeset-presence.mjs`, exit 0: "44 source file(s) of 12 released package(s) changed, and this change declares 2 changeset(s): .changeset/10803-dead-citation-sweep-fourth-batch.md, .changeset/10803-fourth-batch-runtime-strings.md." (44 = the 41 released sources and the 3 tests under released packages' `src/`; the four `test-support` files sit under a package changesets ignores.) - `pnpm changeset:check`, exit 0: "All workspace packages are in the changeset fixed group." / "No changeset declares a `major` bump." - `node scripts/check-changeset-overwrite.mjs` (report-only), exit 0: "2 changeset(s) added, 40 modified, 0 deleted". `declared at base` equals `declares now` for 40 of 40. - `pnpm check:changeset-claims` (report-only), exit 0: - born-false: "Every one of those 1 address(es) either names the tree it was read from, or points at a line this change does not move"; - self-contradiction: "Every package declared across those 30 body(ies) is either not negated in its own prose, or negated only in an ASPECT of it that may legitimately hold still"; - the standing notice "83 pending changeset(s) describe a file this change touches". - Read against the diff: a pending changeset quoting a replaced pointer would itself carry the dead number and so sit in the census. No pending changeset quotes either tombstone string or either describe string; the one textual hit, "truthiness-tests" in `6940-rowactions-boolean-mirror.md`, is prose, not a quote. - `pnpm check:control-bytes`, exit 0: "check-control-bytes: OK (scanned 9125 tracked text file(s); skipped 85 binary)." - `pnpm check:new-line-citations`, exit 0: "VERDICT new-cross-file-line-citations: 0 new citation(s), enforcement report-only -> exit 0". - Also run: the governed-surface predicate over the 90 paths, exit 0: "NOT GOVERNED — 90 path(s) checked against 5 governed surface(s); none matched." (lit control `AGENTS.md`: exit 3). **Tests**, through the shared verify lock. - **Red, reproduced first.** - `layout-default-jsdoc-7361` on `d21aeb0f6` itself gives `VERDICT command-exit 1`, `Tests 1 failed | 74 passed (75)`: "expected '/**\n * Text variant/style. …' to contain '6942'". - `cellClassNameCensusProse-6921`: its `d21aeb0f6` copy was checked out over this tree and run. `ObjectGrid.tsx` and the census changesets it reads are identical at `d21aeb0f6` and here. Result: `Tests 2 failed | 3 passed (5)`, on `/objectui#6882|objectstack-ai#6882/` and `/REDUNDANT since objectui#6882/`. The file was then restored with `git checkout HEAD --`, checked by blob hash against `HEAD`, and `git diff HEAD` came back empty. - **Green, on this head `ae0b81089`:** `layout-default-jsdoc-7361`, `cellClassNameCensusProse-6921`, `action-callback-retired-7068`, `data-table-declared-keys-6882`, `tombstone-discriminator-agreement-9684` and `zod-mirror-parity` give `Test Files 6 passed (6)`, `Tests 181 passed (181)`. - The run was the last part of the locked batch, so its exit is the one the lock's `VERDICT batch-last-exit 0` reports. **Declared narrowing (per the dispatch: the named checks, not wider sweeps).** NOT MEASURED locally: the full suites and type-check of the 12 touched packages, and eslint. Reason: the comment-stripped syntax tree of 43 of the 48 touched files is identical to `main`, and the other 5 differ only by the listed literal edits. Every test that anchors on a changed string or comment was found by the anchor sweep (**Special cases** 10) and run above. CI runs the full farm. ## Acceptance notes 1. **The remainder rides this card, unchanged (seat amendment `5860244997`, Q2 = A).** Of the 83 numbers PR objectui#10869's Acceptance notes item 1 lists, this PR carries the first 30 in ascending order. The other **53** are: objectstack-ai#7091 objectstack-ai#7097 objectstack-ai#7108 objectstack-ai#7177 objectstack-ai#7612 objectstack-ai#7620 objectstack-ai#7623 objectstack-ai#7658 objectstack-ai#7666 objectstack-ai#7667 objectstack-ai#7669 objectstack-ai#7678 objectstack-ai#7681 objectstack-ai#7682 objectstack-ai#7703 objectstack-ai#7704 objectstack-ai#7708 objectstack-ai#7804 objectstack-ai#7844 objectstack-ai#7853 objectstack-ai#7869 objectstack-ai#7874 objectstack-ai#7877 objectstack-ai#7926 objectstack-ai#7959 objectstack-ai#7967 objectstack-ai#7979 objectstack-ai#7980 objectstack-ai#8058 objectstack-ai#8060 objectstack-ai#8072 objectstack-ai#8127 objectstack-ai#8137 objectstack-ai#8204 objectstack-ai#8229 objectstack-ai#8248 objectstack-ai#8307 objectstack-ai#8408 objectstack-ai#9231 objectstack-ai#9241 objectstack-ai#9244 objectstack-ai#9365 objectstack-ai#9373 objectstack-ai#9375 objectstack-ai#9542 objectstack-ai#9553 objectstack-ai#9585 objectstack-ai#10117 objectstack-ai#10119 objectstack-ai#10120 objectstack-ai#10129 objectstack-ai#10132 #14026. None of them occurs in any hunk of this diff. 2. **The C0 bare-number 404s, for later batches** (sites at `e32dae160`, lines / files): - **New to the family:** objectstack-ai#3720 (5 / 3), objectstack-ai#5420 (1 / 1), objectstack-ai#5503 (1 / 1), objectstack-ai#5506 (1 / 1), objectstack-ai#5737 (4 / 4), objectstack-ai#6467 (1 / 1), objectstack-ai#6936 (1 / 1), objectstack-ai#6945 (1 / 1), objectstack-ai#7622 (1 / 1), objectstack-ai#7662 (1 / 1), objectstack-ai#7684 (2 / 2). - objectstack-ai#5737, objectstack-ai#6945, objectstack-ai#7622, objectstack-ai#7662 and objectstack-ai#7684 are written "PR #N". A later batch can look for each one's squash commit, as batch 3 did for objectstack-ai#5505 and objectstack-ai#5758. - The objectstack-ai#3720 sites are in `ObjectGrid.tsx`, `rowCrudAffordances.ts` and `ListView.tsx`. `ObjectGrid.tsx` is also held by open PR objectui#10278. - **Already on the family list:** objectstack-ai#7620, objectstack-ai#7678 and objectstack-ai#7853. They have bare sites too, which the printed instrument counts. - **Not read:** the 22 distinct bare numbers above 10900 (budget). Nothing here re-derives whether any of them is dead. 3. **Published or shipped text outside the two in-scope classes still names some of these numbers.** It is untouched by scope, as another class the triage split did not scope here: - `content/docs`: `enhanced-actions.mdx` and `schema-overview.md` say the callbacks were "RETIRED (objectui#7068)" in prose, not as verbatim quotes of the runtime string; `plugin-form.mdx` names objectui#6958. - The private `@object-ui/test-support` README names objectui#6923 and objectstack-ai#6924. - Published `CHANGELOG.md` history, which is never re-addressed. It includes a bare "framework PR objectstack-ai#6942" in `@object-ui/app-shell`'s. - In tests: `action-callback-retired-7068.test.ts`'s test names and four `@ts-expect-error` comments reading "RETIRED (objectui#7068)"; `layout-default-jsdoc-7361.test.ts`'s header-table comment naming objectui#6942; and the test names and comments in `cellClassNameCensusProse-6921.test.ts` that name objectstack-ai#6882. Only anchors moved (ruling `5861900779`). - Carrier: none. 4. **Filenames are not citations.** Pending changeset and test FILENAMES carry several of these numbers. They stay, as in PRs objectui#10707, objectstack-ai#10797, objectstack-ai#10854 and objectstack-ai#10869. --- _Generated by [Claude Code](https://claude.ai/code/session_014mXUNuFomfj24w7s1pZzhN)_ --------- Co-authored-by: Claude <noreply@anthropic.com>
…imports (^17.4.0) (objectstack-ai#10882) Fixes objectstack-ai#10864 Clause-②: no ## What changed - Raised `packages/app-shell/package.json`'s declared `@objectstack/spec` range from `^17.3.0` to `^17.4.0` — the first published spec version that exports the four symbols two shipped app-shell sources import at runtime (`predicateSlotRefusal`, `structuralConditionRefusal` from `@objectstack/spec/automation`; `EVALUATED_EXPRESSION_SOURCE_REQUIRED`, `EvaluatedExpressionSchema` from `@objectstack/spec/shared`), matching the range `@object-ui/types` already declares. - Regenerated `pnpm-lock.yaml` with `pnpm install --lockfile-only` — moves only the app-shell `@objectstack/spec` specifier line. - Added `.changeset/10864-app-shell-spec-floor.md` (`patch`, `@object-ui/app-shell`). - Amended the pending `.changeset/5920-flow-runner-translation-overlay.md` with a dated note: its closing sentence claimed the package "now declares `^17.3.0`", which this change makes false. Frontmatter untouched — only the body gained the note, following the objectui#10533 precedent. - No source files changed, and no revert of anything to lower a floor. ## M1 to M4 - **M1** (reproduce red on `main`): full workspace build (`turbo run build --filter='!@object-ui/site' --concurrency=2`, 43/43 tasks successful), then `pnpm check:spec-floors` exited 1 with exactly 4 `floor-too-low` findings, all in `@object-ui/app-shell`, matching the issue: `predicateSlotRefusal` and `structuralConditionRefusal` from `packages/app-shell/dist/views/metadata-admin/previews/screen-spec.js` / `flow-sim-validate.js`, plus `EVALUATED_EXPRESSION_SOURCE_REQUIRED` and `EvaluatedExpressionSchema` from `flow-sim-validate.js`. No other package surfaced a finding. - **M2** (confirm the four symbols): `npm pack @objectstack/spec@17.3.0` and `@17.4.0`, unpacked and grepped `dist/`. All four names: 0 hits anywhere under 17.3.0's `dist/`; present under 17.4.0's `dist/automation` and `dist/shared` (`.js` / `.mjs` / `.d.ts` / `.d.mts`). - **M3** (regenerate lockfile with pnpm): `pnpm-lock.yaml` diff is `1 file changed, 1 insertion(+), 1 deletion(-)` — only the `@object-ui/app-shell` importer's `@objectstack/spec` `specifier:` line moves, from `^17.3.0` to `^17.4.0`. The resolved `version:` on that line was already `17.4.0` before this change (normal resolution already picked the newest 17.x; `pnpm install` after the edit reported "Already up to date"). - **M4** (after the change): `pnpm check:spec-floors` exits 0 — "✅ Every consumer-facing `@objectstack/spec` floor carries the symbols its package's artifact references" (19 of 39 published packages inspected, 341 (subpath, symbol) pairs judged). `pnpm --filter @object-ui/app-shell type-check` exits 0 (`tsc --noEmit && tsc -p tsconfig.test.json`). ## Gates - `pnpm check:spec-floors` — red before (4 `floor-too-low` findings) → green after (0 findings; see M1/M4). - `pnpm --filter @object-ui/app-shell type-check` — exit 0. - `pnpm --filter @object-ui/app-shell test` — exit 0: `Test Files 841 passed | 1 skipped (842)`, `Tests 8618 passed | 9 skipped (8627)` (root-form invocation, `vitest run --root ../.. packages/app-shell/`, no filtering trap). - `node scripts/check-control-bytes.mjs` — exit 0 (9120 tracked text files scanned, 85 binary skipped). - `node scripts/check-changeset-presence.mjs` — exit 0 (1 changeset added; the `dependencies` version bump itself is not one of the eight published-contract fields the gate tracks, so it reported "no changeset is owed" independent of the one added). - `node scripts/check-changeset-fixed.mjs` / `check-changeset-no-major.mjs` — both exit 0. - `node scripts/check-lockfile-integrity.mjs` — VERDICT clean (no `@objectstack/*` identity moved backward, no package gained a copy). - `node scripts/check-lockfile-dedupe.mjs` — VERDICT deduped (`pnpm dedupe` would collapse nothing). - `node scripts/check-installed-spec-pin-claims.mjs` — OK (`@objectstack/spec = 17.4.0`, lockfile and resolved tree agree). - `node scripts/check-changeset-claims.mjs` (report-only) — flagged 4 pending changesets naming a file this PR touches (`pnpm-lock.yaml` or `packages/app-shell/package.json`). Read each against this diff: `5793-spec-range-floors.md` and `6361-spec-floor-17-2-0.md` describe a different importer's edge (their own packages resolving `17.2.0`, an edge this one-line specifier bump does not move); `6776-metadata-admin-lazy-registration.md` describes the `sideEffects` array, which this diff does not touch; `7122-objectstack-family-17-3-0.md` describes the `@objectstack/client` / `core` / `formula` / `lint` family pins, also untouched. None falsified — this is the fifth pending changeset the gate named, `5920-flow-runner-translation-overlay.md`, and that one *was* falsified and is amended above. - `node scripts/check-changeset-overwrite.mjs` (report-only) — reports the one changeset this PR modifies (`5920-…`); case 2, "correcting a declaration on purpose" — same package and bump level declared at base and now, only the body gained a dated note. - `node scripts/check-new-cross-file-line-citations.mjs` — 0 new citations. - `node scripts/check-pending-changeset-literals.mjs` — exit 0. ## Serial Re-checked open PRs against this branch's file surface: 11 open (`objectstack-ai#10880, objectstack-ai#10879, objectstack-ai#10878, objectstack-ai#10876, objectstack-ai#10875, objectstack-ai#10852, objectstack-ai#10780, objectstack-ai#10777, objectstack-ai#10714, objectstack-ai#10278`) plus the release PR `objectstack-ai#5400`. Only `objectstack-ai#10878` touches `pnpm-lock.yaml`, and only at the `@object-ui/data-objectstack` and `@object-ui/plugin-grid` importer blocks (each raising its own `@objectstack/spec` floor to `^17.4.0` independently) — not the `@object-ui/app-shell` block this PR moves. No open PR touches `packages/app-shell/package.json` or either changeset file. The release PR regenerates and is not a hold. --- _Generated by [Claude Code](https://claude.ai/code/session_014mXUNuFomfj24w7s1pZzhN)_ --------- Co-authored-by: Claude <noreply@anthropic.com>
2 participants
Fixes #7189
Clause-②: yes — measured on this branch, where the claim reads
no. No metadata or schema key moves (GroupingConfigis unchanged), but published TypeScript surface does:DataSourcein@object-ui/typesgains the optional memberqueryGroupHeaders,ObjectStackAdapterimplements it, and@object-ui/plugin-grid's exportedGroupEntry(newcount,keyValues),AggregationResult.value(nownumber | null) anduseGroupedData(optional fifth argument) change shape. The seat corrects the claim line.What this does
Maintainer ruling A on this card: grouping on a list view is server-side. The set of groups and every number in a group header are properties of the query, not of the fetched page; rows inside a group are paged. This PR is the objectui half, on top of the spec half (objectstack#14556) and the platform half (objectstack#15330).
@object-ui/types,@object-ui/data-objectstack).DataSource.queryGroupHeaders(resource, query)answers the header querycompileListViewGroupQuerycompiles. Presence is the capability. The ObjectStack adapter posts it verbatim to the existingPOST /data/:object/querydoor and answers itsrecords; it never degrades (a refusal throws with the server's code and status; a body withoutrecordsis refused, not read as zero groups). The adapter already had an undeclared spec-shape branch onaggregate(); it was not reused becauseAggregateParamsdoes not declare that shape and the other adapters would answer it with garbage, so a new optional member is the smallest honest addition.@object-ui/plugin-grid). A grouped grid that fetches its own rows from such a source no longer fetches a flat window. It asks one compiled header query per grouping level (outer headers are their own level's numbers;avg/count_distinctdo not fold) and pages each open group's rows with that group's compiled row query (compileListViewGroupRowsQuery: view filter AND group key,limit/offsetper group). Each group larger than the page gets its own pager. A collapsed group costs no row query.aggregationsride the header query (mapped onto the spec's column summaries and read back withderiveColumnSummary). A reference-typed grouping key (a raw id on the wire) is labelled from the referenced record. The view filter is lowered once to aFilterCondition(parseFilterAST) and handed to both compilers, so header numbers and rows answer the same question; a filter that does not lower is refused, never dropped. Grouping keys the principal may not read are not sent as keys.useServerPagination. The grouped exclusion goes for a server-grouped grid: its rows are paged by the server (per group) and its sort is the server's.useGroupedData. With the server's header rows it builds the group tree from them and never bucketsdata. Without them it still groups rows the grid holds.Partialmarker. Retired where counts are server-true: a server-grouped grid never renders it. See the open question below for the one shape that still reaches it.@object-ui/plugin-list). The console mounts the grid throughListView, which fetched one window and handed it down asdata, so the grid could only group a page (and the feat(grid,i18n): a grouped grid states, beside the group counts, that it grouped a page #7226 marker never fired on that path, because no total was handed down for grouped grids). When the data source can answer the header query,ListViewnow hands a grouped grid NO rows, passes it the effective filter (authored + toolbar + chips, the same value the self-querying views get), and draws no record-count bar over it. While a toolbar search is active it keeps hosting the window as before (see the out-of-scope finding: the aggregate door dropssearch). This is outside the claim's named file surface; it is where the page-scoped rows actually came from in the console.Step zero, measured before any grid code (G0)
The installed
@objectstack/spec17.4.0 (read from this worktree'snode_modules) compiled the header query for the 186-row, five-unit fixture, posted through objectstack's realRestServerroute overObjectStackProtocolImplementation+ObjectQL+ in-memory SQLite, from a throwaway objectstack worktree atorigin/maind3958bac(removed afterwards):{"groupBy":["business_unit"],"aggregations":[{"function":"count","alias":"count"},...]}answered five rows 86 / 61 / 31 / 7 / 1 in both the contiguous and the interleaved object,total: 5;total: 86on both pages), 86 distinct ids;Then end to end with this branch's BUILT adapter and the installed
@objectstack/client, wired by a fetch shim into the same real routes:queryGroupHeadersanswered 86 / 61 / 31 / 7 / 1, and paging every group throughadapter.findwith the compiled row query reached all 186 rows (86 as 50 + 36). The empty group (nine rows with a NULL key) was reached through theis_nullpredicate, and astatus = doneview filter moved header numbers and rows together (28 / 20 / 10 / 2, one unit gone).Pins
packages/plugin-grid/src/__tests__/serverGrouping-7189.test.tsx(11): five headers 86 / 61 / 31 / 7 / 1 and no marker in both row orders; the header query equals the spec compiler's output and no flat window is fetched; the 86-row group shows 86 rows under a 100-row window; under a 50-row window it pages 50 then 36 (union 86); all 186 rows reachable through the group pagers; a per-groupsumis the query's; a collapsed group costs no row query; two-level headers come from one query per depth; a lookup key is labelled from the referenced record; CONTROL: a source without the header query still renders 86 / 14 with the marker.packages/data-objectstack/src/queryGroupHeaders-7189.test.ts(4): verbatim body to the door, records answered, refusal withoutrecords, no fallback on a 400.packages/plugin-list/src/__tests__/ListView.groupedGridOwnsFetch-7189.test.tsx(4): grouped grid gets no rows and the effective filter, no record-count bar; controls for a source without the header query and for an ungrouped grid.groupedPartialDisclosure-7189.test.tsxis NOT retired: its data sources declare noqueryGroupHeaders, so it now pins the fallback (header updated to say so).groupedPagination.test.tsxneeded no rewrite (inline rows).Ablation (prediction written before the run): forcing
serverGroupedFetchto false inObjectGrid.tsxturnedserverGrouping-7189.test.tsxto exactlyTests 10 failed | 1 passed (11)as predicted, the one pass being the control. Mutation landed on disk throughablation-replace.mjs(anchor 1 to 0, blob957bad075ca7to6324d33743e9), restored and proven by blob equal to HEAD and emptygit diff HEAD. No build stands between edit and run: the test imports../ObjectGridand the root vitest config aliases@object-ui/plugin-gridtosrc. A first attempt was refused by the tool before running (the replacement contained the anchor), so it measured nothing.Tests and gates, at
8427ae3turbo run build type-checkfor types, data-objectstack, plugin-grid, plugin-list:Tasks: 20 successful, 20 total. The new test files are inside the type-check programs (--listFiles, one hit each).vitest run packages/plugin-grid/:Test Files 164 passed (164),Tests 1536 passed (1536).vitest run packages/plugin-list/ packages/data-objectstack/ packages/types/:Test Files 432 passed (432),Tests 7594 passed (7594).fab627f, clean), lockfile-dedupe.check:spec-floors: zero findings against the four touched packages (all built); it exits 1 only onno-artifactfor ten untouched packages not built locally. CI builds them..ts/.tsxfiles (count from--format json), 0 errors;eslint.config.jssets noparserOptions.project, so type-aware linting is off and this diff cannot move a verdict in an untouched file. The repo-wide run is CI's.check:readme-exports,check:doc-snippets,check:published-dist,check:node-esm-load(each needs the whole workspace built; this diff adds no package export and no fenced block to any document). A live browser run of the console against a running backend was not made; the in-process route probe above stands in for it.Semver
minoron types, data-objectstack, plugin-grid and plugin-list (fixed group, nomajorper this repo's policy); the breaking semantics are stated in the changeset. The floors of plugin-grid and data-objectstack move to@objectstack/spec^17.4.0, the first release carrying the compilers. The pending.changeset/7189-grouped-grid-partial-disclosure.mdcarries two sentences this makes false: "That is a correct implementation of client-side grouping and is unchanged here" and "Server-side grouping — the durable fix — is deliberately NOT part of this ... nothing here builds toward it or changes the fetch." Following this repo's precedent for a pending changeset a later change makes false, both halves are done: the new changeset says it supersedes that paragraph, and a dated note (2026-09-28, naming objectui#7189 and this PR) is appended at the end of.changeset/7189-grouped-grid-partial-disclosure.mdcorrecting those sentences and the one before them (thatuseGroupedDatabuckets the rows the browser already holds). The existing paragraphs and the frontmatter are byte-identical;check-changeset-overwritereports the edit as a touched pre-existing changeset declaring the same packages, a deliberate correction.Open question for the seat
A data source that declares no
queryGroupHeaders(the genericApiDataSource,ValueDataSource, any host adapter) cannot give server-true counts. This PR keeps its old behaviour, grouping the fetched page, and keeps thePartialmarker there because only there are the counts page slices, so thegrid.grouping.partial*i18n rows are retained, not retired. The alternative is to refuse grouping loudly on such a source. Recommendation: keep, as done here. It is honest, it adds no code, and a refusal would break every grouped grid on a custom adapter with no server able to do better.Acceptance notes
POST /data/:object/querywithgroupBy+aggregations+searchanswers the same five full-count groups as withoutsearch, while the samesearchon a flat query narrows to one row. The aggregate branch offindDatacomposeswhere/groupBy/aggregations/havingonly, so the search is dropped silently. That is whyListViewkeeps hosting a searched grouped grid. Reported for filing in objectstack.ListViewstill issues its own window fetch for a grouped grid it hands over (it feeds the filter chips' option counts and the client export), so a grouped console view makes one extra request.Written by the dispatched
os-devagent, sessionhttps://claude.ai/code/session_01MCg3z5cGrV51xEddpGcxEY.Generated by Claude Code