ValueDataSource refuses a stored is_empty / is_not_empty once @objectstack/spec folds the empty pair onto its own names (objectstack PR #20570)

Filing-gate category: ① a product defect with a named site, reached when objectui adopts the next @objectstack/spec release. Filed bare for triage. ⛔ Not a claim.

Filed by the objectstack domain:spec seat 5 (session_01Sfe5YjBLwB9J3y8fvm2xq1, seat post objectstack-ai/objectstack#19357). It comes from item (h) of the contract review of objectstack-ai/objectstack#20570 (card objectstack-ai/objectstack#20446). The seat re-read the objectui sites below at the sha objectstack pins (.objectui-sha dd3f7e1be3); nothing was run.

What changes upstream

objectstack-ai/objectstack#20570 is an @objectstack/spec minor, so it ships in the next spec release. It makes $empty a filter operator, and it changes canonicalAstOperator:

is_empty / isempty fold to is_empty, no longer to is_null;
is_not_empty / isnotempty fold to is_not_empty, no longer to is_not_null.

What that does here (read at `dd3f7e1be3`)

The switch: packages/core/src/adapters/ValueDataSource.ts:468 switches on canonicalAstOperator(rawOperator). It has is_null (:509) and is_not_null (:511) arms, and no is_empty / is_not_empty arm.
The refusal: the default arm (:602) refuses the node ("filter operator … is not implemented by the in-memory matcher").
The effect: a stored rule with is_empty / is_not_empty is answered today through the is_null arms. Once objectui runs that spec release, it is refused. packages/core/package.json declares @objectstack/spec ^17.3.0.
The test: packages/core/src/adapters/__tests__/ValueDataSource.astFilterVocabulary.test.ts pins the current fold. The reviewer read that it goes red on the new release; this seat did not run it.

finding(fields,app-shell,spec): 「is empty」 has three meanings across objectui and the spec ($null only, $exists, or "null OR empty string"), and the empty-string member may not reach typed SQL columns safely #10813 covers the meanings of 「is empty」 in the three builders. This site is a consumer of the spec's fold, not a builder.
The spec's meaning after the flip, per the #20446 changeset: a text-like field is empty when null or '', a multi-value field when null or [], and every other type when null.

Blocked-by: objectstack-ai/objectstack#20446

Dedupe: the seat grepped the open objectui issues for ValueDataSource, canonicalAstOperator, is_empty and $empty. Only #10813, #9048, #8945 and #7650 matched, and none is about this site.

Dedupe words: ValueDataSource is_empty refused · canonicalAstOperator empty fold · in-memory matcher is_empty

Its blocker landed: objectstack-ai/objectstack#20446 closed via objectstack-ai/objectstack#20570 (f1e921ab8e on objectstack main) · domain:spec seat 5 (session_01Sfe5YjBLwB9J3y8fvm2xq1) · 2026-09-29T12:39Z · ⛔ Not a claim.

The change ships in the next @objectstack/spec release, which is not yet published: npm's latest read 17.5.0 today, and that release predates this change. This card's site starts refusing only when objectui installs a spec release that carries the change. Unlocking waits for that release to be installable here.

Generated by Claude Code

Restart-when: npm view @objectstack/spec version answers a release in which canonicalAstOperator('is_empty') answers is_empty (objectstack PR #20570, f1e921ab8e)

Triage: first grade — `bug` · `priority:p2` · `domain:ui` · `area:records` · `pm:on-hold`. `Blocked-by: objectstack#20446` is re-derived to the release that carries PR #20570. The fix lands in, or before, objectui's bump to that release

Triage: lands in packages/core/src/adapters/ValueDataSource.ts (the canonicalAstOperator switch at :468, whose default arm refuses at :602) ⇒ domain:ui.

Triage seat (objectstack-wide, seat post objectstack-ai/objectstack#6015) · session_01AavokzJ5DndAwitDXvKy4U · 2026-09-29T12:56Z. ⛔ Not a claim, ⛔ not a dispatch.

State.

The body's blocker, filter: flip is_empty / is_not_empty to $empty and add it to FILTER_OPERATORS, once every compile surface answers it (the last step of ruling A on #20399) objectstack#20446, closed via PR #20570 (5890482699). But the change isn't in a published release: npm's latest 17.5.0 predates it.
The site refuses only once objectui installs a release that carries it. So the hold is on npm, with the probe above as the restart condition.

Why p2.

Once reached, every stored is_empty / is_not_empty rule on an in-memory (ValueDataSource) source is refused, where today it is answered through the is_null arms. That breaks a stored view on inline data, which is a road step that runs but fails.
It fails loud, not silent.

Direction at restart.

New arms. ValueDataSource gains is_empty / is_not_empty arms that implement the released meaning: a text-like field is empty when null or '', a multi-value field when null or [], and every other type when null. They go through the spec's own expansion function (keyed on the field's declared type and multiple), ⛔ never a second per-type table.
Without a declared type, the arm judges by value, as the ruling on #20399 allows for the faces that have no field declarations.
⛔ The objectui bump to that release must not merge with this site refusing. Land this first, or in the bump PR itself. ValueDataSource.astFilterVocabulary.test.ts, which pins today's fold, is rewritten to the new fold in the same change.
Pins: is_empty / is_not_empty across text, multi-value and number fields, including '' and [], plus the no-declared-type case.
Not this card: objectui#10813 (the three builders' meanings).

Cross-seat request from objectstack's domain:devx seat 2 (seat post objectstack-ai/objectstack#20163, session_01HRYqpqGcWpJuJkDmbRF75w) · 2026-10-03T03:57Z. ⛔ Not a claim, ⛔ not a release of this hold.

Ruling. objectui#11438 (@objectstack/* 17.6.0 bump) is ruled A′ (record 5965062249: director batch #272, maintainer 「同意」). The bump cannot land green without this card's work, and this card waits for the bump. So:

This card's slice lands as commits pushed onto the trunk branch claude/issue-11438-objectstack-17-6-0 (draft objectui#11531), ⛔ not as a separate PR. objectui CI runs only on PRs based on main, and objectui#11531's CI verifies every push.
The branch is shared: ⛔ no force-push, merge commits only.
This card's own lane seat reviews the slice, at its tier, before it is pushed. The devx seat reviews the whole once green and lands objectui#11531, which closes this card along with the other owner cards.

What on the trunk is this card's: row 4 of objectui#11531's body, section "Still red: not mechanical, reported by row":

$empty joins FILTER_OPERATORS, and is_empty / is_not_empty lower to it (objectstack f1e921ab).
⚠️ This is a runtime regression if the bump lands alone: in ValueDataSource.astFilterVocabulary.test.ts (6), an is_empty rule selects nothing.
Also filter-source-merge.test.ts (1), plus the staged-$empty tripwires in four files (1 each).
objectui#11094 and objectui#10813 share this row. The lane seat decides the split.

Release. Triage releases this hold by its Restart-when. ⛔ The devx seat does not release it. objectui#11438 carries Blocked-by: objectui#11094.

Generated by Claude Code

Triage: released — `pm:on-hold` → `pm:queue`, raised to p1 under the release rule. This card's slice lands on objectui#11531's trunk (A′)

Triage seat (objectstack-wide, seat post objectstack-ai/objectstack#6015) · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-03T04:03Z. ⛔ Not a claim, ⛔ not a dispatch.

The ruling. objectui#11438 (the @objectstack/* 17.6.0 bump) is ruled A′ (5965062249, director batch #272, maintainer 「同意」).

This card's slice lands as commits pushed onto claude/issue-11438-objectstack-17-6-0 (draft objectui#11531), ⛔ not as a separate PR.
This card's lane seat reviews the slice before the push.
objectui#11531 lands once, green, and closes this card.

The devx seat's cross-seat request is 5965309322. The ruling has triage release by this card's Restart-when.

Restart-when, read now:

A published release in which canonicalAstOperator('is_empty') answers is_empty: PR feat(spec)!: $empty joins FILTER_OPERATORS, and is_empty / is_not_empty lower to it (#20446) objectstack#20570's f1e921ab8e is an ancestor of the 17.6.0 release commit 617f25f8a4. ✓

Why p1: 「发版本优先的都p1」. The 17.6.0 bump cannot land without this card, and it carries Blocked-by: this card.

Generated by Claude Code

Claim: PM loop round 1 (p1 release slice, ruling A′)
Session: session_01YLg8XqWGJ785fwQ5v4pH37
Account: os-elon-musk (the seat's linked user as GET /user answers it; the card's assignee)
Branch: claude/issue-11094-value-datasource-empty-17-6 (the slice's review branch, cut from the trunk tip c9121f92f2)
Worktree: objectui-issue-11094
Domain: domain:ui
Seat: domain:ui#1
Lands on: the trunk claude/issue-11438-objectstack-17-6-0 (draft PR objectui#11531, owned by objectstack's domain:devx seat 2), under ruling A′ (5965062249, director batch #272, maintainer 「同意」), as commits pushed onto that branch. ⛔ No PR of its own. The slice is prepared on this card's own branch (below), cut from the trunk tip, and reviewed by this seat before the push (at-tier contract review, Clause-②: yes). The push onto the trunk is then ordered by this seat: merge commits only, ⛔ no force-push, ⛔ no rebase (AGENTS.md rule 3). objectui#11531's CI verifies it, and objectui#11531 lands once, green, closing this card.
File surface:

packages/core/src/adapters/ValueDataSource.ts: is_empty / is_not_empty arms through the spec's own expansion function, judging by value when no type is declared;
the tests trunk row 4 names (ValueDataSource.astFilterVocabulary.test.ts rewritten to the new fold, filter-source-merge.test.ts, ValueDataSource.dollarFilterVocabulary.test.ts, ValueDataSource.filterLogicConformance-8513.test.ts, FilterConditionField.operators.test.ts, readme-filter-operator-table.test.ts) and the README rows they read, plus new pins;
the staged-$empty tripwire sites those tests name, only to promote what the release now carries;
changesets for each released package that moves.

The dev stops on a breach and explains it in the report. ⛔ Not objectui#11531's body, not its other families' rows, and not another owner card's files.
Container & model: S, mode:subagent, model: opus (default judgment tier; no path-derived mandate for these paths)
Clause-②: yes
Why yes: ValueDataSource's matcher answers is_empty / is_not_empty with the released meaning ('' and [] count as empty for text-like and multi-value fields), where it answered them as null-only, so a stored filter's answer changes.
Thread-read: 5965349134
Serial constraints cleared: read 2026-10-03T04:42Z; the trunk tip is c9121f92f2.

objectui#10813 (same family, row 4's "and"): the three builders' meanings of 「is empty」. Not this slice (triage 5890759774: "Not this card"). It is claimed after this slice is on the trunk.
This seat's in flight: PR objectui#11537 (objectui#11530, plugin-charts ObjectChart.tsx) in the merge queue, plus the three release slices claimed in the same act (objectui#11227, objectui#11094, objectui#11344). Their surfaces are disjoint, and the seat serializes their pushes onto the trunk one at a time.
The trunk's other owners: objectui#11334 (domain:spec, still pm:on-hold) and objectui#11536 (booked on the trunk by the devx seat). Neither touches these files.

Scope: objectui#11531 closes this card when it lands.

Generated by Claude Code

os-dev-report

{
 "issue": 11094,
 "status": "done",
 "branch": "claude/issue-11094-value-datasource-empty-17-6",
 "pr": null,
 "head": "8a9cf1b5ed49b47affce85067733a228f11a4a4b",
 "trunk_base": "c9121f92f2e813446fda684e4eb696ba619a3c57",
 "session": "session_01YLg8XqWGJ785fwQ5v4pH37 (the PM's session, carried by this subagent run)",
 "premise_still_valid": true,
 "summary": "ValueDataSource now executes the empty pair on @objectstack/spec 17.6.0. The AST arm has is_empty / is_not_empty and the object arm has $empty, both answered by the spec's own isEmptyFilterValue with no declared row. The adapter holds no field declarations (its config is items + idField, and the spec's value provider declares only items), so it judges by value as ruling A on objectstack#20399 allows: null, absent, '' and [] are empty, and a non-boolean $empty flag is refused. convertFiltersToAST also lowers $empty to is_empty / is_not_empty, the inverse of the spec's parseFilterAST. Without that, the README pin's FILTER_OPERATORS check stayed red and an object filter carrying the canonical operator threw unknown-operator. All 11 trunk-row-4 assertions now pass. The four staged-$empty tripwires are promoted. No assertion was weakened and no other family's row was touched. Not pushed onto the trunk: that waits for the seat's order.",
 "rows_green": [
  "ValueDataSource.astFilterVocabulary.test.ts (6; file rewritten to the new fold): `is_empty` / `isempty` / `is_not_empty` / `isnotempty` (now in section 2b, pinned per text / multi-value / number column), plus VIEW `is_empty` and VIEW `is_not_empty`",
  "filter-source-merge.test.ts (1): 'emits no value slot for a rule that carries none' now expects { a: { $empty: true } }",
  "ValueDataSource.dollarFilterVocabulary.test.ts (1): 'the case table covers FILTER_OPERATORS exactly' ($empty is an executed case)",
  "ValueDataSource.filterLogicConformance-8513.test.ts (1): the expiry row is flipped to 'the staging is over upstream — $empty is in FILTER_OPERATORS', and the 5 $empty conformance cases run in the executed table",
  "FilterConditionField.operators.test.ts (1): flipped to 'the staged $empty exclusion expired: FILTER_OPERATORS admits it'",
  "readme-filter-operator-table.test.ts (1): 'no member of the spec FILTER_OPERATORS is missing from BOTH tables' (README supported row + converter arm)"
 ],
 "tests": "Trunk tip red first (c9121f92f, the six row-4 files, under os-verify-lock): 'Tests 11 failed | 246 passed (257)', VERDICT command-exit 1, the 11 trunk-row-4 assertions exactly; each is_empty row selected [] (AST arm refusal). Slice green at head 8a9cf1b (13 files: the six row-4 files, the new filter-empty-operator-11094, filter-refusal-subject-9050, doc-version-claims, dollar-dialect-alias-census, check-doc-links, check-doc-snippet-types, filter-icontains-alignment-8976): 'Test Files 13 passed (13) / Tests 631 passed (631)'. Full suite (`pnpm exec vitest run` from the worktree root under the lock, at 9786e01): 'Test Files 8 failed | 4329 passed | 7 skipped (4344) / Tests 46 failed | 59192 passed | 356 skipped (59594)', VERDICT command-exit 1, held 1h53m on a shared 4-core box. 44 of the 46 are other families' trunk rows: row 1 = registry-inputs-spec-parity 2 + object-grid-properties-bag-11276 4 = 6 (objectui#11531's body says 7); row 3 = WidgetConfigPanel.measureDoor-8894 21 + dashboard-widget-metric-measure-door-8894 14 + spec-object-refinements-7715 2 = 37; row 5 = actionKeys.pin 1; row 4 = 0. The other 2 were this slice's own: filter-refusal-subject-9050 (13 throw sites vs 12 rows) and doc-version-claims (a version literal in the data-objectstack README). Both fixed in 8a9cf1b and re-run green in the 13-file run above. After 9786e01 only these changed: two test files (type-only casts, runtime-identical), that census row, and two README sentences. Every test reading those files was re-run at 8a9cf1b. Reverse verification, predicted first (direction RED for every leg; predictions file written before the runs), run at 8a9cf1b through objectstack scripts/ablation-replace.mjs in WRAP mode (anchor hit x1 -> x0, blob changed, restore proven blob == HEAD and git diff HEAD empty), with a trap restoring both files from HEAD by absolute path. Leg A deleted the AST is_empty / is_not_empty arms: 23 red (all 20 AST rows of astFilterVocabulary incl. both VIEW rows, the dollar AST-twin row, 2 every-path rows), exactly as predicted. Leg B deleted the $empty object arm: 14 red (dollar 4, conformance 5 $empty cases, astFilterVocabulary 3 object rows, 2 every-path rows); the 4 non-boolean-refusal rows stayed green as predicted (the default arm also refuses and names $empty). Leg C deleted the converter $empty arm: 8 red (converter lower / round-trip / compose, 2 every-path, refusal-subject census, 2 README-pin rows); its refusal rows stayed green as predicted. After the legs: DIFF_LINES=0, and both blobs equal HEAD (ValueDataSource.ts 904c94d0, filter-converter.ts ff63bb7d). No build was needed for the ablation: vitest.config.mts aliases @object-ui/core to packages/core/src, so the subject resolves to source.",
 "mcp_calls": "0",
 "api_writes": "1 — POST /repos/objectstack-ai/objectui/issues/11094/comments (this os-dev-report, through scripts/pm/post-stamped.mjs). Not REST: 5 git pushes of the own branch claude/issue-11094-value-datasource-empty-17-6 (the empty write probe, then 9786e01, then 8a9cf1b). Nothing was pushed onto the trunk. No PR, no label, no assignee, no edit to objectui#11531.",
 "gates": [
  "pnpm turbo run build --filter='./packages/*' --concurrency=2 (head 8a9cf1b, under the lock) -> 'Tasks: 39 successful, 39 total', exit 0",
  "pnpm --filter @object-ui/console build (tsc && vite build && build:sdui-manifest && build:plugin) -> exit 0",
  "pnpm turbo run type-check --filter=@object-ui/core --filter=@object-ui/fields --filter=@object-ui/data-objectstack --concurrency=2 -> 'Tasks: 14 successful, 14 total', exit 0. Each package's tsconfig.test.json program lists the slice's test files (tsc --listFilesOnly). Narrowed from the 81-task run: no exported type or signature changed, so importers are not owed it; objectui#11531's CI runs the full one after the push",
  "pnpm check:control-bytes -> 'check-control-bytes: OK (scanned 10135 tracked text file(s); skipped 85 binary)'",
  "pnpm check:new-line-citations -> 'VERDICT new-cross-file-line-citations: 0 new citation(s), enforcement report-only -> exit 0'",
  "node scripts/check-changeset-presence.mjs -> '43 source file(s) of 15 released package(s) changed, and this change declares 2 changeset(s): .changeset/11094-value-source-empty-operator.md, .changeset/11438-objectstack-17-6-bump.md.' exit 0",
  "pnpm changeset:check -> 'All workspace packages are in the changeset fixed group' / 'No changeset declares a major bump', exit 0",
  "pnpm check:changeset-claims -> exit 0 (report-only). It names 3 pending changesets for files this slice edits: 6213-core-adapters-readme-owns-its-directory, 8568-retire-dollar-dialect-lowercase-aliases, 8976-icontains-lowers-in-converter. I read all three, and this diff falsifies none of them",
  "pnpm check:pending-changeset-literals -> 'No test source names a pending changeset.' exit 0",
  "pnpm check:installed-pin-claims -> OK (none of the slice's lines is a pin-claim site, checked with --json)",
  "pnpm check:doc-fences -> exit 0; pnpm docs:check-links -> 'Links are valid across 17 scan roots.'",
  "eslint over the changed lintable files (ESLint API at head 8a9cf1b): 'files=9 lintable=9 errors=0 warnings=70'. Base c9121f9 on the same files: 70, so delta 0. Narrowing proof: (1) the population is eslint's own: isPathIgnored / calculateConfigForFile on every changed path gives 9 LINTED; the changeset and the two READMEs are ignored (no matching config). (2) The count is read from the lint results (9 files). (3) Invariance: no parserOptions.project / projectService in eslint.config.js, so type-aware linting is off, and no rule in eslint-rules/ reads another file from disk, so this diff cannot move any untouched file's verdict"
 ],
 "line_budget": "n/a (no skills/** in the diff)",
 "deviations": [
  "File surface, bounded in-place fix: the claim names ValueDataSource.ts, the row-4 tests, 'the README rows they read' and the tripwire sites. Promoting the README pin as its own message instructs ('if the code implements neither, that is the defect, not this pin') put packages/core/src/utils/filter-converter.ts (the $empty lowering arm, plus $empty in the unknown-operator enumeration) in the diff. That in turn needed filter-refusal-subject-9050.test.ts (its throw-site census row) and the core adapters README. All four in-place conditions hold. (1) Same defect class: a consumer of the release's $empty that refuses it, the objectui#8976 $icontains shape one operator over. (2) The fix is mechanical and its shape pinned: a mirror of the $null arm, the inverse of the spec's parseFilterAST. (3) No other claim holds filter-converter.ts. (4) The same gate family, with no new verification surface. The seat owes the claim's surface an addition for these two files; the alternative is option B of open question 3",
  "FilterConditionField.operators: the tripwire message said '$empty leaves KNOWN_UNREACHABLE the day FILTER_OPERATORS admits it'. Removing it now requires a builder row that emits $empty, which is objectui#10813's decision ('Not this card'). So the expiry row is flipped to the released fact, the entry stays with its reason restated, and a new expiry row reddens the day a builder operator emits $empty. The set's membership and the reachability assertion are byte-for-byte as strong as before. Seat decision: open question 2",
  "The full suite ran at 9786e01, not at the final head 8a9cf1b. The two later commits are listed under tests, and every test reading their files was re-run at 8a9cf1b",
  "type-check is narrowed to the affected packages (core, fields, data-objectstack) rather than the 81-task run, on the ground that no exported type changed",
  "Mechanism wording in the brief's Zone 2 item 1: the AST-arm reds are not caused by 'the lowering now emits $empty'. ValueDataSource's array arm never lowers to $. The node stays ['ts', 'is_empty'], and canonicalAstOperator now answers 'is_empty' (no longer 'is_null'), which had no arm, so the default arm refused it. The $empty lowering is what moved filter-source-merge (parseFilterAST)",
  "The worktree /home/user/objectui-issue-11094 is left in place, with node_modules, for the ordered trunk push and the re-run of gates after a merge. No process this run started is still alive"
 ],
 "files_changed": [
  ".changeset/11094-value-source-empty-operator.md (new; '@object-ui/core': minor)",
  "packages/core/src/adapters/ValueDataSource.ts",
  "packages/core/src/adapters/README.md",
  "packages/core/src/utils/filter-converter.ts",
  "packages/data-objectstack/README.md",
  "packages/core/src/adapters/__tests__/ValueDataSource.astFilterVocabulary.test.ts",
  "packages/core/src/adapters/__tests__/ValueDataSource.dollarFilterVocabulary.test.ts",
  "packages/core/src/adapters/__tests__/ValueDataSource.filterLogicConformance-8513.test.ts",
  "packages/core/src/utils/__tests__/filter-source-merge.test.ts",
  "packages/core/src/utils/__tests__/filter-empty-operator-11094.test.ts (new)",
  "packages/core/src/utils/__tests__/filter-refusal-subject-9050.test.ts",
  "packages/fields/src/widgets/__tests__/FilterConditionField.operators.test.ts"
 ],
 "zone2_readings": [
  "1. Measured at c9121f92f: the 11 assertions, exactly as listed (6 + 1 + 1 + 1 + 1 + 1). Mechanism corrected under deviations.",
  "2. The spec's expansion: expandEmptyOperator(field: ValueShapeFieldDef) returns EMPTY_OPERATOR_ARMS.text / multi_value / null_only, keyed on isMultiValueField (type + multiple) and STRING_VALUE_TYPES. Its value-level half is isEmptyFilterValue(value, expansion?), both exported from @objectstack/spec/data 17.6.0. Both new arms and $empty call isEmptyFilterValue with no expansion, because this face holds no declaration. There is no hand-written per-type table. @objectstack/formula's matcher makes the same call.",
  "3. Each tripwire is promoted as its message instructs, except the FilterConditionField entry (deviations, open question 2). None was deleted or weakened. The conformance file's staged it.each (asserting a refusal) was replaced by those cases running in the executed it.each (asserting the table's rows and no warning), which is stronger.",
  "4. filter-source-merge expected $null because, through 17.5.0, the spec's parseFilterAST lowered is_empty to $null. The assertion exists to show that the server reads the valueless 2-tuple as the rule's own operator-keyed predicate, not as an equality on an invented null. On 17.6.0 that predicate is $empty: true, which is the right answer for that merge: toFilterNode emits the same 2-tuple either way, and the release made is_empty mean emptiness. Moved, and the comment says why.",
  "5. No builder was touched (objectui#10813)."
 ],
 "open_questions": [
  {
   "question": "Should ValueDataSource gain a field-declaration input, so the empty pair can take the declared row (expandEmptyOperator) instead of the by-value reading? The triage text says 'keyed on the field's declared type and multiple', 'without a declared type, by value'. Today no declaration can reach this adapter.",
   "options": [
    "A: No input (as built). Every call is by value. That is the branch the ruling gives a face with no declarations, the same reading @objectstack/formula uses. Pull measured: zero. resolveDataSource, ObjectCalendar / ObjectMap / ObjectTree's inline sources, FieldDesigner and ObjectManager all construct it with items only, and the spec's value ViewData declares only items.",
    "B: Add an optional `fields` map of ValueShapeFieldDef to ValueDataSourceConfig and route through expandEmptyOperator when present. That is a new public config key that no producer fills: declared surface without pull, and a second place an author could put field types."
   ],
   "recommendation": "A. Real need: no producer holds declarations, so B is speculative. Long-term soundness: B adds an input nobody feeds. Preventing AI mistakes: A has one reading, B has a key that silently changes answers when present. Startup focus: no new surface without pull. The readings differ only on '' or [] stored in a scalar column, which the spec calls a write-door defect. If a producer with declarations appears, B becomes a contract card of its own."
  },
  {
   "question": "FilterConditionField.operators: keep $empty in KNOWN_UNREACHABLE, with its expiry moved to 'a builder operator emits $empty', or execute the old tripwire's instruction now?",
   "options": [
    "A: Keep it (as built). The reason is restated: which builder row authors $empty is objectui#10813's decision. The new expiry row reddens when that card moves a row onto $empty.",
    "B: Remove the entry now. That forces a builder operator to emit $empty in this slice, which changes FilterConditionField's stored-rule meaning: the ruled 'Not this card' work, re-scoping stored sharing rules and roll-up filters."
   ],
   "recommendation": "A, because B is objectui#10813's re-scoping decision and the ruling keeps builders out of this card. A keeps the reachability assertion exactly as strong and makes the remaining reason mechanical."
  },
  {
   "question": "Keep the convertFiltersToAST $empty arm in this slice, or move it out?",
   "options": [
    "A: Keep it (as built). The README pin's FILTER_OPERATORS completeness is green because the code implements the operator. ObjectStackAdapter.find with an object $filter (translateFilterToAST, then convertFiltersToAST) lowers $empty instead of throwing unknown-operator. The converter and ValueDataSource give one answer for one authored filter.",
    "B: Drop commit 158204d79 and document $empty in the README's refused table instead. The row would be true of the code, but it records a canonical operator as refused by the generic arm. The README pin's own message calls that state the defect, and the converter and the in-memory matcher would then disagree on one filter, the objectui#8568 / objectui#8976 split."
   ],
   "recommendation": "A. It is the objectui#8976 precedent exactly, and it is mechanical. Under B the trunk would land a known face that refuses the release's canonical operator."
  }
 ],
 "out_of_scope_findings": [
  "carrier: objectui#10813 · FilterConditionField's kvToCondition reads stored criteria back into builder rows and has a $null arm but no $empty arm. Neither FilterConditionField nor the components filter-builder nor ListView's is_empty lowering ([field, '=', null]) emits or reads $empty. These are the three builders' meanings that card reconciles. Noted, not filed."
 ]
}

Generated by Claude Code

Seat rulings on dev report `5966942957`, a claim amendment, and ✅ ACCEPT of the slice (branch head `8a9cf1b`, cut from trunk `c9121f92f`); the at-tier contract review is owed before the push

domain:ui seat 1 · session_01YLg8XqWGJ785fwQ5v4pH37 · 2026-10-03T07:56Z.

Rulings (each inside triage `5890759774`; none is a new decision)

No field-declaration input: A. The ruling gives a face with no declarations the by-value branch. No producer constructs ValueDataSource with field types (the dev measured resolveDataSource, the inline sources, FieldDesigner and ObjectManager, and the spec's value ViewData declares only items). A fields key would be declared surface with no pull. If a producer with declarations appears, it is its own contract card.
$empty stays in FilterConditionField's KNOWN_UNREACHABLE, with its expiry moved to "a builder operator emits $empty": A. Which builder authors $empty is objectui#10813's decision ("Not this card"). The reachability assertion keeps its strength, and the new expiry row turns red when objectui#10813 moves a row.
The convertFiltersToAST $empty case stays in this slice: A. The README pin calls an unhandled canonical operator the defect. Under B, the converter and the in-memory matcher would disagree on one filter, the split objectui#8568 and objectui#8976 closed.

Claim amendment

Claim 5965621810's file surface gains packages/core/src/utils/filter-converter.ts (the $empty lowering and the unknown-operator list), packages/core/src/utils/__tests__/filter-refusal-subject-9050.test.ts (one census row), packages/core/src/adapters/README.md and packages/data-objectstack/README.md (the sentences the change made false). Same defect class, mechanical, not held by another claim.

ACCEPT

The matcher: ValueDataSource answers is_empty / is_not_empty and $empty through the spec's own isEmptyFilterValue, ⛔ no per-type table. Null, absent, '' and [] count as empty. A non-boolean $empty is refused like every refusal in this adapter.
Trunk row 4: all 11 assertions pass. astFilterVocabulary is rewritten to the new fold, pinned per text, multi-value and number column. filter-source-merge's expectation moves to $empty: true, with the reason written, and the four staged-$empty tripwires are promoted. ⛔ No assertion weakened: the conformance file's staged cases now run in the executed table.
Reverse verification:
- the trunk tip: 11 failed (exactly row 4);
- ablations A (the AST cases) 23 red, B (the $empty case) 14, C (the converter case) 8, each as predicted, restores proven.
The full suite at 9786e01: 46 failed, of which 44 are other families' trunk rows (row 1: 6, row 3: 37, row 5: 1) and 2 were this slice's, fixed in 8a9cf1b. Every test reading the later commits' files was re-run green.
The changeset: @object-ui/core minor, saying what moves for a stored rule ('' and [] now count as empty).
Builders: no builder was touched (objectui#10813).

Contract review

Served-tier: CONTRACT_REVIEW_TIER
Head-sha: 8a9cf1b5ed49b47affce85067733a228f11a4a4b
Local-runs: none

Trunk slice of card objectui#11094 under ruling A′ (5965062249 on objectui#11438): branch claude/issue-11094-value-datasource-empty-17-6, five commits acd62a3a1..8a9cf1b5e, cut from and diffed against the trunk tip c9121f92f2e813446fda684e4eb696ba619a3c57 (12 files, +571 / −73; merge-base = trunk tip; the remote trunk tip and draft objectui#11531's head still read c9121f92f2 at this review). Not a PR review: the slice has no check-runs, so the dev report 5966942957 and the seat's ACCEPT 5966957488 were read as claims and tested against the diff, the trunk tip and the installed @objectstack/spec 17.6.0 (filter.zod.ts, filter-empty-operator, view.zod.ts) plus @objectstack/formula 17.6.0. Read 2026-10-03T08:10Z.

① Derived judgments

The matcher, AST arm — right. canonicalAstOperator in 17.6.0 answers is_empty / is_not_empty for the four empty spellings (it folded them onto the null pair before), and the trunk-tip switch had arms only for is_null / is_not_null, so the default arm refused every stored empty rule. The slice adds case 'is_empty' / case 'is_not_empty' answering through isEmptyStoredValue, a one-line delegate to the spec's isEmptyFilterValue(value) with no expansion. That IS the ruling's by-value branch and not a second table: the spec's own function reads null, undefined, '' and [] as empty when called without an expansion (verified in the installed 17.6.0 source), the spec's FILTER_OPERATORS docblock assigns exactly this reading to the faces that hold no declarations, and no declaration can reach this adapter: ValueDataSourceConfig carries rows and an id field, the spec's provider: 'value' data source is a strictObject declaring only items. No per-type table is written and none is inferred from the rows; expandEmptyOperator is correctly not called because there is no field to key it on. @objectstack/formula 17.6.0's matchesFilterCondition makes the identical call (isEmptyFilterValue(actual) for true, its negation for false), as the code comment and changeset claim.

AST_NO_COMPARAND_OPERATORS gains the empty pair — right. The spec's parseFilterAST discards the value slot on all four unary operators, so an array in slot three of an is_empty node is not an array comparand. This restores the 17.5.0 accept for that shape (where the fold onto is_null put it in the set already); the array-comparand refusal still governs every other operator.

The matcher, object arm $empty — right. Same function, target === true / === false strict, direction from the flag; it is the twin of the new AST arms and sits beside $null / $exists. The parity guard that holds the dollar case table equal to FILTER_OPERATORS is what forces the arm to exist, and REFUSED_OPERATORS stays empty.

The stored-rule semantic move — right, and declared. On this adapter is_empty / $empty: true now select rows holding '' or [] beside null / absent, and is_not_empty / $empty: false no longer do; a value, 0, false, null and a missing key answer as before. That is the released meaning (ruling B on objectstack#20311 as the spec states it, by value for a face with no declaration), and the changeset says exactly this under "What moves for a stored rule". Relative to the trunk tip (17.6.0 installed) it is a refusal becoming an answer; relative to 17.5.0 it is a null test becoming an emptiness test. Pinned in astFilterVocabulary §2b over a text, a multi-value and a number column, '' and [] included, plus the number-column by-value case controlled by the spec's own function called both ways (EMPTY_OPERATOR_ARMS.null_only leaves '' a value; no expansion makes it empty).

The non-boolean $empty refusal — right. SpecialOperatorSchema declares the flag z.boolean(). The adapter excludes the row and logs once per find() through refuseFilterNode, naming $empty and the field; 'yes', 1, 0 and null are pinned refused rather than read for truthiness (an array flag is caught one guard earlier by the array-comparand refusal, also naming $empty). The converter throws FilterOperatorError (code: 'INVALID_FILTER', httpStatus: 400, operator: '$empty', field) for 'yes', 1, 0, null and []; before the slice the same input reached the unknown-operator throw, so the refusal narrows in message only, and filterRefusalSubject still names $empty (the census row added to filter-refusal-subject-9050 is that fact). No runtime string carries a tracker number.

convertFiltersToAST lowering $empty — right, and it is the inverse of parseFilterAST. { f: { $empty: true } } lowers to ['f', 'is_empty', true] and false to ['f', 'is_not_empty', true], written as the $null arm is (direction from the flag, true placeholder), placed before convertOperatorToAST so the unknown-operator throw is no longer reached; parseFilterAST reads those nodes back as { f: { $empty: flag } } (pinned, both directions). $empty was in neither operatorMap nor RETIRED_OPERATOR_ALIASES, so the trunk tip really did throw "Unknown filter operator" for a canonical member of FILTER_OPERATORS. For @object-ui/data-objectstack's find() both routes lower an object $filter through this function (translateFilterToAST on the raw GET, convertFiltersToAST on the client route), so an object filter carrying $empty now reaches the wire as the spec's own AST spelling instead of throwing in the browser; what the server then does with it is the spec's declared-row table (and the spec's own refusal on an undeclared column), not objectui's. The unknown-operator enumeration gains $empty, which is what the README pin reads as the converter's supported set.

Trunk row 4, the 11 assertions — each green by a right reading, none by weakening.

ValueDataSource.astFilterVocabulary.test.ts (6): at the trunk tip the 8-spelling null it.each pairs lost is_empty / isempty and is_not_empty / isnotempty (4) and the VIEW is_empty / is_not_empty rows (2) to the refusal arm; that is 6, and the dev's "selected []" mechanism is the right one (below). The rewrite keeps every null-pair row, reads the fold live from canonicalAstOperator and parseFilterAST (so the next move of the fold reddens beside its reason), and adds §2b. The VIEW rows are untouched and are green because VIEW_ROWS hold no '' / [], where the two readings agree.
filter-source-merge.test.ts (1): the expectation moves from { a: { $null: true } } to { a: { $empty: true } }, which is what 17.6.0's parseFilterAST answers for [['a', 'is_empty']]; toFilterNode still emits the valueless 2-tuple, so the assertion's purpose (the server reads the operator's own predicate, not an equality on an invented null) holds. Moved, with the reason written.
ValueDataSource.dollarFilterVocabulary.test.ts (1): $empty enters EXECUTED_CASES with its AST twin, so the exact-cover guard against FILTER_OPERATORS is green by execution; the fixture's $empty: true equals $null: true on those rows, which the added '' / [] row set disambiguates.
ValueDataSource.filterLogicConformance-8513.test.ts (1): the expiry row flips to toContain('$empty'), and the five staged cases now run in §1 asserting the spec's expected rows and no warning, which is stronger than the refusal assertion they replace; the by-name list is kept over the same five names. The spec's FILTER_LOGIC_ROWS hold d as 'v1' / 'v2' / null / null, so the by-value reading meets the table's expectations.
FilterConditionField.operators.test.ts (1): the expiry row flips; the reachability sweep's KNOWN_UNREACHABLE membership is byte-identical, so its strength is unchanged; the new row holds the remaining reason mechanically (no FILTER_BUILDER_OPERATORS id makes condToMongo emit $empty, with $null as the lit control). See ③ for the ruling.
readme-filter-operator-table.test.ts (1): the pin's third population now finds $empty in the supported table; its probes [true, false, 'probe'] lower to exactly is_empty / is_not_empty as the row's second column says (the string probe is refused inside the INVALID_FILTER / 400 envelope the pin expects); the worked example { email: { $empty: true } } lowers to ['email', 'is_empty', true] byte-for-byte; the "every operator the error calls supported has a row" case is why $empty had to join the message. Green by implementing the operator, which is what the pin's own failure text asks for, not by documenting a canonical operator as refused.

The dev's corrected mechanism — right. The AST-arm reds are not caused by a lowering that "now emits $empty": find()'s array arm never lowers to the $ dialect, the node stays ['ts', 'is_empty'], and the refusal came from canonicalAstOperator's new answer meeting a switch with no arm. Only filter-source-merge reads parseFilterAST. objectui#11531's row-4 sentence is imprecise on this point; a devx-side wording fix, not this slice's.

Builders — untouched. FilterConditionField.tsx, components/custom/filter-builder.tsx and ListView's is_empty lowering are not in the diff (objectui#10813).

② Semver level

@object-ui/core minor — right. Clause-②: yes is the right claim: the converter's accept set widens ($empty with a boolean, previously an unknown-operator throw), the adapter answers two AST operators and one object operator it refused on 17.6.0, the empty pair's value slot is no longer judged as an array comparand, and a stored rule's answer changes ('' / []). yes takes at least minor; objectui's own rule keeps breaking semantics at minor with the meaning written in the body, which this body does. No arm is declared and none is owed ("at most one").

Prose, sentence by sentence against the code and the installed 17.6.0: $empty admitted to FILTER_OPERATORS — true; the fold off is_null / is_not_null and the lowering to $empty: true / false — true (canonicalAstOperator, parseFilterAST); "had arms only for the null pair" and "reached the refusal arm: selected no row and logged one refusal per find()" — true (refusals is a Set drained once after the pass); the After paragraph — true, including the formula parallel; the stored-rule paragraph — true; the converter paragraph, the envelope, "before this, the unknown-operator throw", the message now listing $empty, and data-objectstack's find() — all true. The ValueDataSource docblock's sentence on driver-memory's reference matcher is as the spec's FILTER_OPERATORS table states it (that driver is not installed here; verified against the spec's text, not the driver's code). The core adapters README's "'' stored in a number column" divergence sentence matches isEmptyFilterValue under null_only. The data-objectstack README row and its two paragraphs are true of the converter and the spec; neither edited README region carries a version literal.

Note, not a defect: only @object-ui/core is bumped, while @object-ui/data-objectstack's find() behaviour moves transitively through the import; the fixed group versions it regardless and the changeset body names the move, but that package's own CHANGELOG will not carry it.

③ Boundary flags

Open question 1 (no field-declaration input), ruled A — right. By value is the branch the spec itself gives a face with no declarations; a fields map would be declared surface no producer fills (resolveDataSource, the inline sources, FieldDesigner, ObjectManager construct with items; the spec's value provider admits only items). A producer that holds declarations is a contract card of its own.
Open question 2 ($empty stays in KNOWN_UNREACHABLE), ruled A — right, within triage's "Not this card". The old tripwire's literal instruction (the entry leaves the set the day FILTER_OPERATORS admits $empty) would force a builder row onto $empty in this slice, which triage 5890759774 reserves for objectui#10813. The reachability assertion keeps its strength; the reason is restated and held by a mechanical row. Flag: the new expiry can only redden when a builder changes, so the entry's life is now tied to objectui#10813's decision; if that card rules that no builder row emits $empty, the entry needs its reason rewritten, not just left standing.
Open question 3 (keep the converter $empty arm), ruled A — right. The alternative documents a canonical operator as refused by the generic arm, which the README pin's own failure text calls the defect, and splits the converter from the in-memory matcher on one authored filter (the objectui#8568 / objectui#8976 shape).
The claim amendment (filter-converter.ts, filter-refusal-subject-9050.test.ts, the two READMEs): the four files are in the diff and nowhere else is; with the claim's named surface, the new pin file and the changeset, the 12 files are fully accounted for. Same defect class, mechanical, held by no other claim.
Full suite at 9786e01 rather than the head: the two later commits are ed622b48f (two casts in test files typed instead of as any; runtime-identical, read in the diff) and 8a9cf1b5e (one census row, two README prose edits); the dev re-ran the 13 files that read those files at the head. Accepted as the read; objectui#11531's CI after the push is the gate reading for this head.
Other trunk families: every file in the diff is row 4, new, or a README; the trunk tip has not moved since the cut. At 9786e01 row 4 reads 0 and the other 44 reds are families 1, 3 and 5. The dev's family-1 count of 6 against the PR body's 7 is a difference between the body's head 40d9b5a and c9121f92 (the row-2 booking commits touched console tests), not this slice; for the devx seat's whole-trunk read.
Governed surfaces: none touched. No model identifier in the diff, the changeset or the commit trailers.
Out of scope, carried by objectui#10813: FilterConditionField's kvToCondition has a $null arm and no $empty arm, so a stored $empty criterion would not read back into a builder row; ListView lowers is_empty to [field, '=', null]. Noted, not filed.

Implemented-by: claude/issue-11094-value-datasource-empty-17-6
Reviewed-by: session_01YLg8XqWGJ785fwQ5v4pH37

VERDICT: PASS

Generated by Claude Code

os-dev-report

{
 "issue": 11094,
 "status": "done",
 "branch": "claude/issue-11094-value-datasource-empty-17-6",
 "pr": null,
 "session": "session_01YLg8XqWGJ785fwQ5v4pH37 (the PM's session, carried by this subagent run)",
 "premise_still_valid": true,
 "pushed_trunk_sha": "9187c5c407060643860f4e3783fe6812dcaaa99b",
 "head": "9187c5c407060643860f4e3783fe6812dcaaa99b",
 "trunk_base": "da4c629f1d02a7b3aca11cd86938568381fd0e72",
 "summary": "Push order carried out. The trunk tip da4c629f1 was merged into the slice (8a9cf1b) as merge commit 9187c5c40, with no rebase and no force. The merge had no conflict and touched neither filter-converter.ts nor ValueDataSource.ts. The re-run under the verify lock is green: install, 97 test files, and type-check across four packages. 9187c5c40 was then pushed onto claude/issue-11438-objectstack-17-6-0 as a fast-forward (da4c629f1..9187c5c40), and the slice branch was pushed to the same sha.",
 "merge_findings": "No conflict. git diff --stat 8a9cf1b 9187c5c: 49 files, +2445 / -241, all from the trunk (objectui#11227's slice, plus main 6f5719e1c with objectui#11533's ListView / object-grid stored-rule dialect change, objectui#11525, objectui#11526, objectui#11530 and others). No slice file is in the delta. filter-converter.ts and ValueDataSource.ts are byte-identical to 8a9cf1b.",
 "tests": "One lock hold at 9187c5c (waited behind objectui#11344's full suite; held 300s). (1) pnpm install -> exit 0, 'Lockfile is up to date, resolution step is skipped' / 'Already up to date'. (2) pnpm exec vitest run over 97 files -> exit 0, 'Test Files 97 passed (97) / Tests 1740 passed (1740)'. The 97 are: trunk row 4's six files; the new pins filter-empty-operator-11094 and filter-refusal-subject-9050; FilterConditionField.operators and doc-version-claims; every test that git grep -l finds naming filter-converter / ValueDataSource / convertFiltersToAST / toFilterNode / mergeFilterNodes / toFilterNodeSafely; and the dialect tests the merge brought: ListView.storedRuleDialect-11533, grid-list-view-conditional-formatting-11533, ObjectMetricWidget.ruleFilterSpecShape-11526, plus filter-icontains-alignment-8976. Row 4 stays fully green (all 11 assertions), and no test in the set is red. (3) pnpm turbo run type-check --filter=@object-ui/core --filter=@object-ui/fields --filter=@object-ui/data-objectstack --filter=@object-ui/plugin-list --concurrency=2 -> exit 0, 'Tasks: 17 successful, 17 total' (all four ran with a cache miss, including the tsconfig.test.json programs). After the hold, git status shows no tracked change. Light gates at 9187c5c, outside the lock: check:control-bytes OK; check:new-line-citations 0 new; check-changeset-presence exit 0 (3 changesets declared); changeset:check exit 0; check:pending-changeset-literals exit 0.",
 "readback": "Trunk ref claude/issue-11438-objectstack-17-6-0 reads 9187c5c407060643860f4e3783fe6812dcaaa99b, and so does the slice branch. objectui#11531: head 9187c5c4, draft true, mergeable true, mergeable_state 'blocked'. Check-runs on 9187c5c4 at the time of reading: total_count 44, of which 41 in_progress and 3 completed (skipped). The CI conclusion is NOT MEASURED, because the runs are still in flight.",
 "mcp_calls": "0",
 "api_writes": "1 — POST /repos/objectstack-ai/objectui/issues/11094/comments (this report, through scripts/pm/post-stamped.mjs). Not REST: git push HEAD:claude/issue-11438-objectstack-17-6-0 (fast-forward da4c629f1..9187c5c40), and git push of the slice branch (8a9cf1b..9187c5c40). No force anywhere, no objectui#11531 body edit, no builder touched (objectui#10813).",
 "deviations": [],
 "open_questions": [],
 "out_of_scope_findings": []
}

Generated by Claude Code

Claim: PM loop round 1 (takeover of the p1 release slice, ruling A″)
Session: session_01FjqrwXPfSMkSfkKYDSRkN2
Account: os-zhuang (the seat's linked user as GET /user answers it; the card's assignee after this act)
Branch: claude/issue-11094-value-datasource-empty-17-6 (continued, not re-cut; head 9187c5c40)
Worktree: none in this container. The slice is already on the trunk, so no dev is dispatched.
Domain: domain:ui
Seat: domain:ui#1
File surface: unchanged from claim 5965621810. This takeover adds no file.
Container & model: S, mode:subagent, model: opus (default judgment tier; dispatch-gates --tier --repo objectstack-ai/objectui printed "no path-derived mandate" for packages/core/src/adapters/ValueDataSource.ts)
Clause-②: yes
Why yes: carried from 5965621810. The matcher's answer to a stored is_empty / is_not_empty changes; the at-tier record is PASS 5967059614.
Thread-read: 5968152020
Serial constraints cleared: the slice is on the trunk claude/issue-11438-objectstack-17-6-0. objectui#10813 (the director's, 5968139578) pushes after it. Nothing else of this seat is in flight.

Release: claim 5965621810, session_01YLg8XqWGJ785fwQ5v4pH37 (os-elon-musk). Reason: that session is out of tokens. Provenance: the maintainer, in this session's chat (session_01FjqrwXPfSMkSfkKYDSRkN2), verbatim 「前任没有 token了，你继续。」. Destination: the same seat, this session.

Handover record: the old branch's last pushed sha is 9187c5c407060643860f4e3783fe6812dcaaa99b. The trunk ref read the same sha at the 10:13Z push (report 5968152020); the trunk has since moved on to 13f07dd0f. Status: ACCEPT 5966957488, contract PASS 5967059614, slice pushed onto the trunk. What remains is to follow objectui#11531 to MERGED, verify by content on main, and close this card.

Generated by Claude Code

Landed and verified by content: closed completed by objectui#11531's merge · domain:ui seat 1 · session_01FjqrwXPfSMkSfkKYDSRkN2 · 2026-10-03T13:27Z.

The merge: PR objectui#11531 → 6158e4c93 on main, at 2026-10-03T13:14Z. Its closing lines name this card (Fixes #11094).
Content check (git grep on objectui origin/main 6158e4c93, with its first parent 6f5719e1c as the control): packages/core/src/adapters/ValueDataSource.ts calls isEmptyFilterValue 4 times on main and 0 on the parent. The file's own class ValueDataSource line reads 1 on both, so the zero is a real absence and not a missing file.
Trail: the slice 8a9cf1b (ACCEPT 5966957488, contract PASS 5967059614), merged onto the trunk as 9187c5c40.
pm:dispatched is removed in the same act. domain:ui, area:records, the type and the priority stay.

Generated by Claude Code

chore(deps): resolve @objectstack/* 17.6.0 in pnpm-lock.yaml and foll…

Navigation Menu

Repository navigation

ValueDataSource refuses a stored is_empty / is_not_empty once @objectstack/spec folds the empty pair onto its own names (objectstack PR #20570) #11094

What changes upstream

What that does here (read at `dd3f7e1be3`)

Related

Triage: first grade — `bug` · `priority:p2` · `domain:ui` · `area:records` · `pm:on-hold`. `Blocked-by: objectstack#20446` is re-derived to the release that carries PR #20570. The fix lands in, or before, objectui's bump to that release

Triage: released — `pm:on-hold` → `pm:queue`, raised to p1 under the release rule. This card's slice lands on objectui#11531's trunk (A′)

Seat rulings on dev report `5966942957`, a claim amendment, and ✅ ACCEPT of the slice (branch head `8a9cf1b`, cut from trunk `c9121f92f`); the at-tier contract review is owed before the push

Rulings (each inside triage `5890759774`; none is a new decision)

Claim amendment

ACCEPT

Next

Contract review

① Derived judgments

② Semver level

③ Boundary flags

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

ValueDataSource refuses a stored is_empty / is_not_empty once @objectstack/spec folds the empty pair onto its own names (objectstack PR #20570) #11094

Description

What changes upstream

What that does here (read at dd3f7e1be3)

Related

Activity

objectstack-fleet commented on Sep 29, 2026

objectstack-fleet commented on Sep 29, 2026

Triage: first grade — bug · priority:p2 · domain:ui · area:records · pm:on-hold. Blocked-by: objectstack#20446 is re-derived to the release that carries PR #20570. The fix lands in, or before, objectui's bump to that release

objectstack-fleet commented on Oct 3, 2026

objectstack-fleet commented on Oct 3, 2026

Triage: released — pm:on-hold → pm:queue, raised to p1 under the release rule. This card's slice lands on objectui#11531's trunk (A′)

objectstack-fleet commented on Oct 3, 2026

objectstack-fleet commented on Oct 3, 2026

objectstack-fleet commented on Oct 3, 2026

Seat rulings on dev report 5966942957, a claim amendment, and ✅ ACCEPT of the slice (branch head 8a9cf1b, cut from trunk c9121f92f); the at-tier contract review is owed before the push

Rulings (each inside triage 5890759774; none is a new decision)

Claim amendment

ACCEPT

Next

objectstack-fleet commented on Oct 3, 2026

Contract review

① Derived judgments

② Semver level

③ Boundary flags

objectstack-fleet commented on Oct 3, 2026

objectstack-fleet commented on Oct 3, 2026

objectstack-fleet commented on Oct 3, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

What that does here (read at `dd3f7e1be3`)

Triage: first grade — `bug` · `priority:p2` · `domain:ui` · `area:records` · `pm:on-hold`. `Blocked-by: objectstack#20446` is re-derived to the release that carries PR #20570. The fix lands in, or before, objectui's bump to that release

Triage: released — `pm:on-hold` → `pm:queue`, raised to p1 under the release rule. This card's slice lands on objectui#11531's trunk (A′)

Seat rulings on dev report `5966942957`, a claim amendment, and ✅ ACCEPT of the slice (branch head `8a9cf1b`, cut from trunk `c9121f92f`); the at-tier contract review is owed before the push

Rulings (each inside triage `5890759774`; none is a new decision)