[Decision] what 「is empty」 means on a text column and on a multi-value column: null only (the spec's lowering today), or null OR `''` / `[]` — three objectui builders disagree, and a stored sharing rule's rows depend on the answer

[objectstack-fleet]

Unblocked: 5865751001 · #20399 ruled A (batch #234 item 1, 5865693155) · 2026-09-28T07:53Z
Ruled: 5861435168 · letter B · 2026-09-28T00:55Z

Unblocked by ruling 5865693155 on #20399 (A, batch #234 item 1; note 5865751001) — was blocked by #20399 from 2026-09-28T06:54Z. Execution: declare $empty per the ruling.

Path: permissions that actually hold | 缺项 (no item authors 「is empty」 on a sharing-rule criterion) | P2

Filing gate: ② a decision only the maintainer can make: it is the meaning of a declared operator in @objectstack/spec, and changing it re-scopes stored rules. Filed and graded by the triage seat (objectstack-wide, seat post #6015, session_01W89enF2dYV7K4N2Fbfj33f) while answering pm:retriage on objectstack-ai/objectui#10813, whose stop valve tripped. ⛔ Not a claim.

Grade: needs-user-decision · priority:p2 · domain:spec · area:records.

维护者速读

事情:「为空」这个筛选条件,在平台里有三种意思。

• 协议(spec)把它翻成「值是 null 或缺失」。
• objectui 的数据集桥写成「字段不存在」。
• 最常用的那个筛选控件(共享规则、汇总字段的过滤都用它)写成「null 或空字符串」。

同一个条件,在不同地方挑出的记录不一样。

为什么要你拍板:开发实测过,清空的文本框存的是空字符串 '',不是 null。平台底层有一条测试专门钉住「空字符串不能变成 null」。所以如果照协议现在的写法(只认 null),用户眼里空着的文本记录,「为空」反而筛不出来。而改控件的写法,会改变已存共享规则能看到的记录。这是协议语义加存量数据的决定,席位不能自己选。

⭐ 数值、日期、布尔字段里存了 '',是写入环节的缺陷,不需要你裁,已单独立卡 #20308,在写入口统一转成 null。本卡只问文本字段和多值字段。

选项:

• A:「为空」只认 null。写入口把文本的 '' 也转成 null,并迁移存量数据。代价:推翻那条「空字符串不能变成 null」的底层约定,还要做数据迁移。
• B(荐):协议声明「为空」按字段类型展开:文本类 = null 或 '';多值类(多选、标签、多值查找)= null 或 [];其它类型 = null。三个构建器都改成读协议的这一个展开。代价:协议里多一张按类型展开的表,但只写这一处。
• C:不动,把三种意思写进文档。

席位意见:荐 B。

• 用户眼里「空着」就是空。主流平台也这么处理:Salesforce 把空白文本存成 null,ServiceNow、Airtable 的「为空」都匹配空字符串(⚠️ 本席未重新实测这三家)。
• B 不动存储,不需要迁移,也保住了「空字符串不能变成 null」的底层约定。
• 已测到零条存量规则会因为 B 改变结果(objectui、objectstack 都是 0 条;生产环境读不到)。

⚠️ 没测的:生产环境里有没有共享规则用了「为空」,读不到。

你要做的:回一个字母 —— A / B / C。

os-decision-facets

• ① 项目长远合理性:一个操作符在协议里只该有一个意思,并且只写在一处。B 把「为空」按类型的展开写进协议,三个构建器改成读它,以后不会再出现第四种意思。A 也只剩一个意思,但要推翻已钉住的存储约定。C 让分歧永久存在。
• ② 实际业务拉动:共享规则和汇总字段过滤用的正是那个控件。今天「为空」在多值字段上还会直接报错:控件写出 $in: [''],而 SQL 驱动拒绝对 JSON 列做 $in。这是真实的、会报错的路径,不是理论问题。
• ③ 防 AI 写代码/写元数据犯错:三种意思意味着 AI 在不同构建器里写同一个条件,得到不同的记录集。B 让协议自己回答「为空」是什么,构建器只照读。
• ④ 创业阶段不扩散:B 不是新能力,是给已声明的操作符补上它缺的类型展开。A 需要一次数据迁移,是三者里最重的。

Governing text:

• packages/spec/src/data/filter.zod.ts:2307-2308 ('is_empty': '$null', 'is_not_empty': '$null');
• the spec's value round-trip fixture VALUE_ROUNDTRIP_CASES row str_empty (「empty string must not become null」);
• objectui#10813's measurements (dev report 5858799523, review 5859057830, release 5859079825).

After the ruling

• B: domain:spec declares the per-type lowering once (filter.zod.ts). objectui#10813 then converges FilterConditionField and datasetFilterCondition.ts on it, including the multi-value arm that today writes $in: [''] onto JSON columns.
• A: the write-door card (record write door: '' skips every type check, so a number, boolean, date, datetime or time column stores an empty string — normalise it to null at the door (seam from objectui#10813) #20308) extends its normalisation to string columns, with a data migration, and the fixture row changes by ruling.
• C: the three meanings are documented where each builder offers the operator.

Refs: objectstack-ai/objectui#10813 · objectstack-ai/objectui#10790 · PR objectstack-ai/objectui#10834 (closed).

[objectstack-fleet]

Ruling: batch #229 item 1 · letter B · maintainer 「同意」 2026-09-28T00:52Z

Director seat, summon #30 续 2, session_01AsCNgFBs8HCjwhyHQsFbx3. Batch #229 was presented in the seat chat with the full analysis of each item (background, options with real costs, mainstream platforms, four axes, recommendation, confidence gaps, execution; the triage seat's analysis in this card's body re-read, four axes not flipped); the maintainer answered 「同意」 (verbatim, recorded here per the charter).

Ruled: B — @objectstack/spec declares what 「is empty」 means once, per field type: text-like = null or ''; multi-value (multi-select, tags, multi-value lookup) = null or []; every other type = null only. The three objectui builders read that one lowering; none keeps a meaning of its own.

Readings on origin/main ab820016b3 (objectstack) and fab627ff (objectui): packages/spec/src/data/filter.zod.ts:2395-2396 lowers is_empty / is_not_empty to $null; the value round-trip fixture value-roundtrip-conformance.ts:244 (str_empty, 「empty string must not become null」) pins that a cleared text box stores ''; objectui filter-builder.tsx:161 says in its own comment that is_empty / is_not_empty 「also treat '' as empty」, and the dataset bridge spells the operator a third way (the card's measurement, objectui#10813). So the operator the protocol declares has three meanings on the product, and the one the protocol writes today cannot find the records a user sees as empty. On a multi-value field the builder emits $in: [''], which the SQL driver refuses on a JSON column: a live error path, not a theoretical one. Numeric, date and boolean columns storing '' are a write-door defect and are #20308's, not this card's. Prior-ruling sweep over AGENTS.md, docs/adr, packages/spec/src: no ADR gives the operator a per-type meaning.

Mainstream reading the ruling rests on: Salesforce stores a blank text field as null, so its 「is empty」 covers the blank case by construction; ServiceNow's and Airtable's 「is empty」 match the empty string; no mainstream product asks a user to tell null from ''. B reaches the same user-facing meaning without moving storage, and keeps the round-trip fixture true. A reaches one meaning by overturning that fixture and migrating data, the heaviest of the three; C makes the disagreement permanent.

Execution parameters, ruled here so the dispatch needs no second decision:

1. Spec (domain:spec): the lowering of is_empty / is_not_empty becomes a per-type table in filter.zod.ts, declared once, with the three arms above and a describe that says which types take which arm; is_not_empty is its exact complement. A pin per arm and per type family; the round-trip fixture is untouched.
2. objectui#10813 then converges FilterConditionField and datasetFilterCondition.ts on the spec's lowering, including the multi-value arm that emits $in: [''] today; no builder keeps a private meaning.
3. Stored sharing rules: re-measured at landing on both repositories (0 rows change result today, per the card); production rules are NOT MEASURED and the changeset says so.
4. Changeset @objectstack/spec minor (a declared operator gains a per-type meaning: a widening of what a filter finds on text and multi-value columns), Clause-②: yes; ADR-0087 registration if the migration gate asks for a semantic entry.
5. Carrier: this card becomes the spec-lane work card (domain:spec · p2 kept); it closes by Fixes with the spec PR, and objectui#10813 unblocks on that landing.

Confidence gap carried: production sharing rules that use 「is empty」 are unreadable from the fleet.

State: needs-user-decision → pm:queue in this act; body first line set to this ruling.

---

Generated by Claude Code

[objectstack-fleet]

Claim: PM loop round 1
Session: session_01B3TqpoQbTAfG7G74GMDWNW
Account: hotlong (the seat's linked user as GET /user answers it; the card's assignee; seated on the maintainer's ruling 「放行，以 hotlong 坐席」, seat post #6017 marker 5864561291)
Branch: claude/issue-20311-is-empty-per-type-lowering
Worktree: objectstack-issue-20311
Domain: domain:spec
Seat: domain:spec#1 (seat post #6017)
Ruling-ref: 5861435168 (batch #229 item 1, letter B)
File surface:

• packages/spec/src/data/filter.zod.ts: the is_empty / is_not_empty lowering (the operator map rows, canonicalAstOperator's null/empty fold, and the array-sugar lowering's null/empty branch) becomes the per-type table the ruling names, declared once, with its describe.
• Its tests in packages/spec/src/data/. ⛔ The value round-trip fixture (value-roundtrip-conformance.ts) is untouched.
• An ADR-0087 semantic entry under packages/spec/src/migrations/entries/semantic/ plus the step-18 list in packages/spec/src/migrations/registry.ts, only if the migration gate asks for one.
• The spec artefacts the repo's generators regenerate, produced by the generators, never by hand.
• .changeset/.
• ⛔ Not objectui (objectui#10813 converges its builders after this lands). Any caller outside packages/spec that must change to honour the table is reported first, not edited silently.
  (stop on breach; explain in the report)
  Container & model: M, mode:subagent, model: opus (dispatch-gates --tier --repo objectstack-ai/objectstack packages/spec/src/data/filter.zod.ts: no path-derived mandate; clause-② suspect surface). Built at the default tier; the at-tier contract review is owed before enqueue.
  Clause-②: yes
  Thread-read: 5861435168
  Serial constraints cleared: read at 2026-09-28T06:28Z.
• No open PR touches filter.zod.ts (open-PR file lists read at this claim).
• packages/spec/src/migrations/registry.ts (step-18 list) is carried by open PRs feat(automation)!: edge-branched decision is exclusive; mode: 'inclusive' takes every branch (#15429) #20344, feat(analytics): enforce analytics_cube.public and default it to visible #20348, feat(spec)!: retire the connector resilience family — health (probe + breaker), status and nested webhooks, sixteen keys nothing read (#20273) #20350, feat(spec)!: retire the list view's own tabs key; named presets are listViews entries #20357 and fix(spec): os migrate meta guidance for the driver-*, kernel-* and system-* migration entries states each lesson in words, not tracker numbers (stage 3) #20384. If this card adds an entry, its landing merges main with scripts/pm/os-regen-merge.sh only, never by hand.
• Sibling in this round: driver-sql on PostgreSQL answers 500 for a non-numeric string against a number field — where { amount: { $gt: "abc" } } is DATABASE_ERROR / 500 over REST, while InMemoryDriver and SQLite answer 200 with no rows #20336 (this seat, a new module in the same directory) shares the data barrel at most; region-disjoint.

[objectstack-fleet]

os-dev-report
{
"issue": 20311,
"status": "needs_decision",
"branch": "claude/issue-20311-is-empty-per-type-lowering",
"pr": null,
"session": "session_01B3TqpoQbTAfG7G74GMDWNW — subagent run; this is the parent seat's harness-stamped id",
"premise_still_valid": false,
"summary": "Stopped before code: ruling B's multi-value arm (null or []) has no spelling in the lowered FilterCondition. An empty list in the equality slot is refused at the shared comparand-shape face and at the schema door by ruling 乙 on #19757 (and $ne: [] by #19886), and no declared operator tests for an empty stored list, so a per-type lowering in filter.zod.ts can carry the text arm and the null arm but not the third. Separately, the order's premise A2 is false at three runtime doors: protocol.ts:10862, engine.ts:1056 and filter-normalizer.ts:2184 lower is_empty without the field types they hold or can reach, and protocol.ts receives is_empty from named production producers. The choice between a type-aware lowering (the dispatch's route, which must reopen #19757 for []) and a declared emptiness operator resolved at the compile surfaces (which leaves #19757 standing and needs no type at any lowering door) decides what the spec half even is, so writing either now would be speculative; open_questions[0] carries the options on the four axes, with a recommendation.",
"premise_evidence": {
"tree": "objectstack origin/main ab6fb02 (worktree ../objectstack-issue-20311, BASE ab6fb02); objectui origin/main 9f0c84a",
"A1_lowering_sites": "confirmed, none sees a field type: AST_OPERATOR_MAP rows is_empty/is_not_empty/isempty/isnotempty lower to $null (filter.zod.ts:2460-2463); canonicalAstOperator folds is_empty onto is_null and is_not_empty onto is_not_null (:2514-2521); convertComparison emits { f: { $null: true|false } } (:2643-2650)",
"A2_sites": [
"protocol.ts:10862 parseFilterAST(parsedFilter) — schema reachable, not passed: the same normalizer reads field declarations through this.resolveQueryFields(object) (assertSortFieldsExist, :9844, gate.fields[head].type). Production input WITH is_empty: objectui packages/core/src/utils/filter-converter.ts viewFilterRuleToNode (:1156; VALUELESS_VIEW_OPERATORS :1146) lowers a stored view rule { operator: 'is_empty' } to ['f','is_empty'] onto the list request; objectstack packages/rest/src/view-filter-rule-lowering.ts lowerViewFilterRules (:87) does the same for a public form picker's static filter. Verdict: needs a change outside packages/spec.",
"engine.ts:1056 parseFilterAST(where, ctx) — lowerWhereFilterArray(object, operation, bag, schema) holds schema and hands it to the three sibling doors on the same seam (assertFilterIsMaterializable, assertTextOperatorTargetsAreStringCapable, assertTemporalComparandsInterpretable), not to the lowering. In-repo authored corpus with is_empty: 0. Verdict: needs a change outside packages/spec.",
"service-analytics filter-normalizer.ts:2184 lowerAnalyticsWhere(query) — takes no schema; its callers hold the cube (objectql-strategy.ts:200/219/255). In-repo corpus: 0; objectui's dataset builder writes the object form ({ $exists: false }, datasetFilterCondition.ts:190), not the sugar. Verdict: needs a change outside packages/spec.",
"driver-memory memory-driver.ts:1348 canonicalAstOperator — node form { type: 'comparison' }; producers of that node outside driver-memory and its tests: 0 (grep over packages/). Verdict: no production input; its case arms for is_empty/isempty/is_not_empty/isnotempty (:1415/:1418) are unreachable after the fold.",
"lint validate-preset-comparands.ts:676 canonicalAstOperator — a classifier for date-preset findings; is_empty folds to is_null, which no arm reads; lowers nothing. Verdict: unaffected.",
"spec data-engine.zod.ts:1058 parseFilterAST(bag.where) — a transport-schema transform with no object schema by construction, reached through FindDataRequestSchema.query, which no runtime package parses (0 non-test hits outside packages/spec; plugin-rest-api.zod.ts:963 names it as a string). spec conversions/registry.ts:10811 never sees is_empty (it folds to a view-rule operator first, :10806)."
],
"A2_corpus_control": "is_empty spellings in examples/: 0; control is_null|is_not_null|not_equals|contains in examples/: 37 hits in 12 files. Whole-tree is_empty hits: spec, drivers, formula (CEL isEmpty, a different thing), rest docblock, docs, changesets; no platform-objects, template or example authors it.",
"A3_type_sets": "text-like = STRING_VALUE_TYPES (field-value.zod.ts:57, 14 types); multi-value = isMultiValueField(def) (:355) = MULTI_OPTION_TYPES (multiselect, checkboxes, tags) or a MULTI_CAPABLE_TYPES member (select, radio, lookup, user, file, image) with multiple: true; so a multi-value lookup is lookup or user with multiple: true. Consequence: the table keys on the field DEFINITION (type plus multiple), so the zone-3 signature (fieldType, negated) cannot tell a multi-value lookup from a single one.",
"multi_value_arm_probe": "tsx probe over packages/spec/src at ab6fb02, EXIT=0. Text arm { $or: [ { name: { $null: true } }, { name: { $eq: '' } } ] }: shape face passed, FilterConditionSchema true; complement { name: { $null: false, $ne: '' } }: passed, true. Multi-value: parseFilterAST(['tags','=',[]]), { tags: [] }, { tags: { $eq: [] } }, { tags: { $ne: [] } }, { tags: { $in: [null] } } and the $or arm with $eq: [] all THROW code=INVALID_FILTER status=400; FilterConditionSchema and FieldOperatorsSchema refuse $eq: [] at parse. Control: { tags: { $in: [] } } passes the face, FieldOperatorsSchema { $in: [] } true. Governing text: filter-comparand-shape.ts:237-262 (ruling 乙 on #19757, record 5793368540: 'An EMPTY array is still an array: { tags: [] } is refused too'); pre-face readings in migrations entry 18.filter-equality-array-comparand-refused.ts (SQL family 400 top level / 500 nested, driver-memory 400, formula no row, driver-mongodb answered) — quoted, not re-measured. [] is a real stored state: record-validator.ts:263-279 (#9447 ruling: the empty set reads back as [], never null).",
"A4_compile_surfaces": "not reached — the lowered shape is the open decision. What each option costs per surface: the text arm uses only declared operators ($or, $null, scalar $eq/$ne), per-surface execution NOT MEASURED this run; the [] arm is refused at the shared face before any surface runs (probe above). Declared-type access for option A, by grep of each compile file for NON_TEXT_STORED_VALUE_TYPES / isMultiValueField / STRING_VALUE_TYPES / field-def reads: driver-sql sql-driver.ts 10 and driver-turso remote-transport.ts 3 read declared types in the compile file; service-analytics read-scope-sql.ts takes a caller-supplied classifier (nonTextColumn, :653); objectql having-filter.ts 1 (a comment; it filters aggregated rows with no declarations); 0 in formula matches-filter.ts, driver-memory memory-matcher.ts, driver-mongodb mongodb-filter.ts, service-analytics filter-normalizer.ts.",
"A5_stored_rules": "objectstack ab6fb02: criteria sharing rules in examples/ = 2 (app-showcase src/security/sharing-rules.ts:76 record.status == 'new', :110 record.stage == 'qualified' && record.company == 'Northwind'); using emptiness = 0; app-crm declares none (objectstack.config.ts:102). objectui 9f0c84a: criteria rule fixtures = 1 (app-shell clientValidation.optOuts.test.ts:52, region == "west"); using emptiness = 0. The widget's stored shape ($in: [''] beside $null) in objectstack outside spec tests and one changeset: 0. Production sharing rules: NOT MEASURED (unreadable from the fleet)."
},
"tests": "No code, so no test run is owed. Readings taken: (1) tsx scratchpad probe against packages/spec/src at ab6fb02, EXIT=0, 16 result lines quoted in premise_evidence.multi_value_arm_probe (refusals print code=INVALID_FILTER status=400; controls pass). (2) node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack: EXIT=2, 'this branch changes nothing against origin/main (merge base ab6fb02) — nothing to derive'. (3) Corpus greps with positive controls as listed in premise_evidence (A2_corpus_control, A5_stored_rules). No build, no ablation: nothing was changed.",
"gates": "not run: files_changed is empty (branch head = origin/main ab6fb02); the derivation above has nothing to derive.",
"files_changed": [],
"line_budget": "n/a — no diff",
"deviations": [
"The order's A2 fallback ('implement the spec half … write Part of #20311') was not taken: what the spec half IS depends on open_questions[0] (a type-aware lowering entry under B, a declared operator plus the table under A), so writing it now is speculative code. The role file's stop rule (a public-contract shape with two readings that lead to two architectures) and the order disagree here; the role file wins by its own precedence clause, and this line names the conflict rather than choosing silently.",
"The zone-3 signature (fieldType, negated) is falsified by A3: the multi-value arm needs the field definition (type plus multiple).",
"Writes: git push of the empty branch (the route probe) and this one comment; no pr_create, no label-write, PR assignee not set (no PR).",
"Worktree ../objectstack-issue-20311 was clean with zero commits ahead and is removed (node_modules first, then git worktree remove without --force); the remote branch stays at ab6fb02 for the re-dispatch after the ruling."
],
"mcp_calls": "0",
"api_writes": "1 — POST /repos//issues/20311/comments (this report), through scripts/pm/post-stamped.mjs on the fleet-write relay. git push of claude/issue-20311-is-empty-per-type-lowering is not a REST write.",
"open_questions": [
{
"question": "How does the lowered filter spell ruling B's multi-value arm (null or []), and so where is the per-type table applied: at the lowering, or at the compile surfaces?",
"options": [
"A — a declared emptiness operator. Add $empty: boolean to the FilterCondition vocabulary (FILTER_OPERATORS, FieldOperatorsSchema); its describe IS the ruled table (text-like = STRING_VALUE_TYPES: null or ''; multi-value = isMultiValueField: null or []; every other type: null); is_empty / is_not_empty lower TYPELESSLY to { f: { $empty: true|false } }, and each compile surface expands it by the field's declared type through one spec function. Cost: one new operator; all 8 compile surfaces of A4 implement one arm, in a staged landing (surfaces answer first, the lowering flips last, or every stored is_empty is refused as an unknown operator in between); the faces with no declarations (formula matcher, having) read by value, a named divergence only on a non-text column holding '' (the write-door class #20308 carries). Gain: no lowering door needs a field type (A2's failure dissolves), #19757 stands untouched, a stored filter keeps its meaning if its field's type changes, and an author of the canonical object form gets one spelling for 「is empty」.",
"B — type at lowering time (the dispatch's suggested route). parseFilterAST takes a field-definition resolver and emits per-type fragments. Cost: the multi-value fragment needs [] as an equality comparand, so ruling 乙 on #19757 (and its save-door twin on #19889) must be reopened for the empty list, and the SQL family, driver-memory and the formula matcher must learn to execute it on a JSON column (each refuses or answers no row today); the three runtime doors of A2 must pass resolvers; the transport-schema door (data-engine.zod.ts:1058) has no object schema at all and needs its own answer (open_questions[1]); and a lowered filter stored in sharing-rule criteria freezes the field's type at authoring time.",
"C — B for the text and null arms now, the multi-value arm deferred to its own ruling. Cost: ruling B's third arm undelivered and the table ships two of three arms; if the deferred arm later lands as A, the B half is rework."
],
"recommendation": "A, on the four axes. Real need: the measured production path for 「is empty」 is a stored view rule that objectui lowers to ['f','is_empty'] and protocol.ts:10862 lowers again without types; A serves it with no caller change, B needs three doors changed first. Long-term: one predicate and one table, applied where the declared type is known — the pattern NON_TEXT_STORED_VALUE_TYPES already uses for the text operators; B bakes a type-dependent fragment into stored criteria. AI-error: the canonical object form gets a spelling; under B an AI writing FilterCondition directly must hand-build the per-type $or, which is how '' reached a number column (#20336). Startup scope: A adds one operator but no capability (it is the canonical spelling of an operator already declared); B reopens a five-day-old ruling. The trade-off to weigh: A touches all 8 compile surfaces in one staged landing; B touches fewer compile surfaces but more doors, plus a reopened ruling."
},
{
"question": "Only under B or C: what does a lowering door that holds no field types do with is_empty?",
"options": [
"A — refuse with INVALID_FILTER / 400 naming the missing type. Loud, but every such door must be converted before the lowering flips, or saved views with 「is empty」 answer 400.",
"B — keep $null at such a door. This is the typeless path keeping $null as a meaning of its own, which the order forbids."
],
"recommendation": "A under B or C; the question does not arise under option A of open_questions[0]."
}
],
"out_of_scope_findings": [
"carrier: sibling card, filed by the seat (needed under options B and C only) — metadata-protocol protocol.ts:10862 lowers is_empty without the field definitions it already reads through resolveQueryFields (:9844); reached in production by objectui filter-converter.ts viewFilterRuleToNode and objectstack rest lowerViewFilterRules · dedupe words: is_empty typeless lowering protocol, parseFilterAST field type resolver, view filter is_empty null only",
"carrier: sibling card, filed by the seat (options B and C only) — objectql engine.ts:1056 holds schema on the lowering seam and does not pass it to parseFilterAST · dedupe words: lowerWhereFilterArray schema parseFilterAST, is_empty engine lowering type",
"carrier: sibling card, filed by the seat (options B and C only) — service-analytics filter-normalizer.ts:2184 lowerAnalyticsWhere takes no schema; its callers hold the cube · dedupe words: lowerAnalyticsWhere field type, analytics is_empty lowering",
"carrier: sibling card(s), filed by the seat (every option) — the 8 compile surfaces of A4: under A each implements $empty from the spec table; under B the SQL family, driver-memory, the formula matcher and having must execute an empty-list equality on a JSON column · dedupe words: is_empty compile surface, empty list equality JSON column, $empty operator",
"carrier: the PR that implements the ruling (every option) — packages/spec/src/ui/view-grouping-query.ts:539-546 says the empty-group predicate ($null) and the view filter's is_empty agree on what 「empty」 means; under ruling B that stops being true on text and multi-value columns · noted, not filed",
"carrier: the driver-memory compile-surface card above — memory-driver.ts:1415/:1418 case arms for is_empty/isempty/is_not_empty/isnotempty are unreachable because canonicalAstOperator (:1348) folds them first · dead code, noted, not filed"
]
}

---

Generated by Claude Code

[objectstack-fleet]

Release: session_01B3TqpoQbTAfG7G74GMDWNW · cause: premise falsified (ruling B's multi-value arm has no spelling in the lowered filter) · destination: pm:blocked, Blocked-by: #20399 · domain:spec seat 1 · 2026-09-28T06:53Z

Seat review of the dev report 5864899591: ACCEPT as premise evidence, no PR

The dev stopped before code with premise_still_valid: false. The seat checked the three load-bearing readings on origin/main itself; it did not take them from the report.

• No operator tests for an empty list. FILTER_OPERATORS in packages/spec/src/data/filter.zod.ts ends at '$null', '$exists'; there is no emptiness or size member.
• [] in the equality slot is refused by ruling. Ruling 乙 on [finding] the comparand-SHAPE face declares it closes the door "for every driver at once", but an array in the IMPLICIT-EQUALITY slot passes it — and driver-mongodb alone answers it, as an exact-array match #19757 (5793368540), in filter-comparand-shape.ts, verbatim: 「An EMPTY array is still an array: { tags: [] } is refused too.」
• An emptied multi-value set is stored as []. record-validator.ts (the [Decision] Multi-value lookup after set_null member removal: does the emptied array read back as [] or null? #9447 ruling): 「the empty set is representable (it reads back as [], never null)」.
• So ruling B's text and null arms are expressible in a lowered FilterCondition, and its multi-value arm is not. The shape the spec half takes depends on a choice between two public-contract changes: a new $empty operator, or reopening [finding] the comparand-SHAPE face declares it closes the door "for every driver at once", but an array in the IMPLICIT-EQUALITY slot passes it — and driver-mongodb alone answers it, as an exact-array match #19757 for []. That choice is the maintainer's.

The order's fallback was not taken, and the seat agrees. The order said: if a site outside packages/spec must change, implement the spec half and write Part of. The dev read that fallback as unexecutable, because what the spec half IS depends on the open choice. That is correct: either spec half would be speculative code.

Premise A2 was falsified as well (report premise_evidence.A2_sites). protocol.ts:10862, engine.ts:1056 and service-analytics filter-normalizer.ts:2184 lower is_empty without the field types they hold or can reach. protocol.ts receives is_empty from a named production producer: objectui viewFilterRuleToNode, re-read by the seat on objectui 9f0c84a. Under option A of #20399 that stops mattering; under B it becomes three sibling cards.

Disposition. Ruling B stands and is not re-presented. The new question is on #20399: needs-user-decision, with options A ($empty), B (reopen #19757 for []) and C (two arms now). The seat recommends A, with the full four-axis analysis there. This card goes to pm:blocked behind it, and domain:spec · p2 · area:records stay. The next claimant re-claims under the full protocol once #20399 is ruled. objectui#10813 stays blocked on this card.

Leftovers: the branch claude/issue-20311-is-empty-per-type-lowering sits at ab6fb027 (no commits, the dev's route probe), and the worktree is removed. The branch is not a claim; the next dispatch may reuse or replace it.

Report hygiene: mcp_calls 0; api_writes 1 (this report). No write tool is named.