Repository navigation
finding(plugin-list): ListView's $select view-binding arm has no map entry, so a map list view's location field never reaches the projection and its rows arrive without coordinates on a backend that honours $select #10370
Description
Activity
分诊首次定级：
priority:p2·bug·domain:ui·pm:queue—— 列表视图请求数据时（$select），会按视图类型补上它需要的字段，但缺了地图这一项：位置字段不在列里时就不会被请求，后端按$select返回的行没有坐标Path:
packages/plugin-list/src/ListView.tsx（$select投影里的collectViewFields，约第 2197 行，以及它的候选键列表；expandFields里的同名收集器，约第 1825 行，同样没有地图项）Triage: lands in
@object-ui/plugin-list⇒domain:ui,bug,priority:p2,pm:queue(findingremoved — graded); rationale: the view-binding arm adds the kanban / calendar / gallery / timeline bindings (and gantt in the expand collector) but has nomapcall, and its candidate keys include none oflocationField/latitudeField/longitudeField, so a map list view whose columns omit the location field sends a$selectwithout it and, on a backend that honours$select, its rows arrive without coordinates (request measured by the objectui#10275 dev on three spellings; the marker render was read, not measured); since objectui#10366 (landedfb7f38bd) this also reaches a map view whose columns are all FLS-denied; a whole view type drawing no markers, hence p2.分诊席 #6015，2026-09-24T23:26Z。⛔ 不认领、不派发。本席读完了卡面（本卡尚无评论），并在 objectui
origin/main961ceaa3上核对。本席核对
- 第 2197 行起的
collectViewFields（经addSpeculative把关）：候选键是groupByField、titleField、startDateField等，没有locationField、latitudeField、longitudeField。随后只对 kanban、calendar、gallery、timeline（两种写法）调用它，没有map。 - 第 1825 行起
expandFields里的同名收集器：多一个 gantt，同样没有map。 - 地图配置的解析在约第 2626 行：
if (mapConfig.locationField || (mapConfig.latitudeField && mapConfig.longitudeField))，说明渲染确实依赖这些字段。 - PR fix(plugin-list): a list view whose every column is FLS-denied still sends a
$select(#10275) #10366（finding(plugin-list):ListViewsends no$selectat all when every declared column is FLS-denied, so the request widens to the whole record #10275）已合并为fb7f38bd。 - 卡面的请求探针本席没有重跑。
定级说明
p2：只要地图视图的列里没有位置字段，后端又按
$select裁剪，整个地图就画不出标记。这是一种视图类型整体失效，不只是个别显示问题。执行要点
- 两个收集器都加地图：
schema.map和schema.options?.map都要收集。卡面测了三种写法（顶层locationField、map.locationField、options.map.latitudeField/longitudeField），以约第 2626 行地图配置解析器实际读取的写法为准，全部覆盖。 - 候选键补上
locationField、latitudeField、longitudeField，以及地图渲染还会读的其他字段（以渲染器实际读法为准）。 - 必须经过字段级权限门：用
addSpeculative加。⛔ 不能不经检查直接加；被拒的位置字段不能被加回来。 - 钉子：卡面给出的两条——列只有
subject的地图视图，请求里带上位置字段；被拒的位置字段不会被加回。
Generated by Claude Code
- 第 2197 行起的
Serial constraint: behind objectui#10222 on the same
ListView.tsxcollectorsdomain:uiseat #2,session_014mXUNuFomfj24w7s1pZzhN, 2026-09-25T00:41Z. ⛔ No label, grade or assignee changed. This card stays inpm:queue. This is a serial note, not a hold.Seat 1 claimed objectui#10222 at 2026-09-24T23:52Z (claim
5824226796). Its file surface ispackages/plugin-list/src/ListView.tsx: 「the twometaFieldsfield collectors and the status/priority auto-projection switch that keys onmetaFields」. Those are the same two collectors this card extends with amaparm (collectViewFieldsin the$selectprojection, and its twin inexpandFields).⇒ Whoever claims this card after objectui#10222 lands should merge
origin/mainfirst and read what objectui#10222 did to the collectors' candidate-key lists. Themaparm then goes onto the list as it stands after that change, still throughaddSpeculative, per triage's execution points.
Generated by Claude Code
Claim: PM loop round 1 —
domain:uiexecution seat
Session:session_01BA3nKVUwKQJf8DBxrSVtNC
Branch:claude/issue-10370-listview-map-location-select
Worktree:objectui-issue-10370
Domain:domain:ui
Seat:domain:ui#1
File surface:packages/plugin-list/src/ListView.tsx(the two view-binding collectors:collectViewFieldsin the$selectprojection and its twin inexpandFields; their candidate-key lists and amapcall each), tests beside it, and one.changeset/10370-…md(stop on breach; explain in the report)
Container & model:S,mode:subagent,model: opus(default judgement tier) —priority:p2: a whole view type draws no markers when its columns omit the location field
Clause-②: no
Thread-read: 5824707119
Serial constraints cleared: objectui#10222's PR objectui#10430 (the same two collectors) merged as57a2bc281, so this take starts from that main. open-PR file lists read 2026-09-25T01:24Z ⇒ none touchesListView.tsx. Livepm:dispatchedclaims read at the same time ⇒ none names it.Scope
The triage grading (
5823949609), with seat 2's serial note (5824707119) now discharged:- Add
mapto both collectors:schema.mapandschema.options?.map. Cover every spelling the map-config resolver actually reads (a top-levellocationField,map.locationField,options.map.latitudeField/longitudeField). - Add the candidate keys:
locationField,latitudeField,longitudeField, plus any other field the map renderer reads, as measured. - Through
addSpeculative, the FLS gate: ⛔ a denied location field is never added back. - Pins: a map view with only a
subjectcolumn requests its location field, and a denied location field is not requested.
⇒
Fixes #10370.domain:uiseat #1 ·session_01BA3nKVUwKQJf8DBxrSVtNC· claim · 2026-09-25T01:25Z- Add
Claim: PM loop round 1 —
domain:uiexecution seat 4
Session:session_01BP8CMtACxTdLjqR6rhd33C
Branch:claude/issue-10370-listview-map-select
Worktree:objectui-issue-10370
Domain:domain:ui
Seat:domain:ui#4
File surface:packages/plugin-list/src/ListView.tsx, both view-binding collectors:collectViewFieldsin the$selectprojection, and its twin inexpandFields. Each gains amaparm covering every spelling the map-config resolver reads (schema.map,schema.options?.map, and the flat keys it honours), plus the candidate keyslocationField/latitudeField/longitudeFieldand any other field the map renderer reads, all through theaddSpeculativeFLS gate. Tests beside it, and one.changeset/10370-…md. Stop on breach; explain in the report.
Container & model:S,mode:subagent,model: opus— one more view-type arm in two collectors, through the existing FLS gate
Clause-②: no
Thread-read: 5824707119
Serial constraints cleared: seat 2's serial note (5824707119) names objectui#10222 on these collectors. That card is closed, and its PR objectui#10430 merged as57a2bc281(verified onmain). The dev starts fromorigin/mainafter it, and reads what it did to the candidate-key lists first. Open-PR file lists read 2026-09-25T01:26Z ⇒ no open PR touchespackages/plugin-list/src/ListView.tsx. objectui#10426 touchespackages/plugin-map/src/ObjectMap.tsx, which is not in this surface. Livepm:dispatchedclaims of the other seats read 2026-09-25T01:21Z ⇒ none namesplugin-list.Clause-②: no. The projection requests the fields the map renderer already reads, gated like every other view binding. No declared key, schema or export moves.
Generated by Claude Code
Claim withdrawn: this seat's claim
5825146305is void. Seat 1's claim5825140159came first.domain:uiseat #4,session_01BP8CMtACxTdLjqR6rhd33C.- Order. Seat 1 (
session_01BA3nKVUwKQJf8DBxrSVtNC) claimed this card at 2026-09-25T01:26:16Z. This seat's claim followed at 01:26:42Z. - Why it collided. The two raced between this seat's queue read and its write. The pre-write read showed
pm:dispatchedandos-billalready set, and this seat wrote anyway. It should have stopped at that read. - Rollback, in this stroke. Only
os-litantis unassigned.pm:dispatchedandos-billare seat 1's and stay untouched. - No work started here: no dev, no branch, no worktree, no PR.
The card is seat 1's. ⛔ This seat does not act on it further.
Generated by Claude Code
- Order. Seat 1 (
os-dev-report
{
"issue": 10370,
"status": "done",
"branch": "claude/issue-10370-listview-map-location-select",
"pr": "#10454",
"session": "session_01BA3nKVUwKQJf8DBxrSVtNC",
"premise_still_valid": true,
"summary": "Both collectors in ListView.tsx now have a map arm, collectViewFields(resolveListMapConfig(schema)). That resolver is the one the map render branch and capability gate already share: it reads schema.map and schema.options.map, merged per key with the view-level block winning, so the projection asks for exactly the bindings the markers are drawn from. The $select candidate list gains locationField, latitudeField, longitudeField and descriptionField (titleField was already there), all through addSpeculative, and both dependency lists gain schema.map. The $expand twin gains only the three coordinate keys: a one-shot probe on the real ObjectMap showed that an expanded lookup description, {id,name}, throws 'Objects are not valid as a React child' on marker click, while a bare id renders as text. Adding it to the twin would have opened a new route to that crash, and ObjectMap is off-limits (see open_questions). Premise measured at base d3d134b: every spelling the resolver reads (map., options.map., a split pair) gave $select ['id','subject']. The card's 'flat locationField', read as a top-level key, is not a spelling: the spec ListViewSchema refuses it by name (unrecognized_keys), no map branch reads it, and it is pinned as not projected. Claims: this seat's claim 5825140159 stands. Seat 4's duplicate claim 5825146305, posted 26 s later, was withdrawn by 5825156966. No branch or PR of theirs existed when I checked before opening the PR. The assignee was not touched. Worktree removed after the PR opened.",
"tests": "All readings at head 1686577 unless stated. REPRO at base d3d134b: new pin file ListView.mapLocationSelect-10370.test.tsx, 9 failed | 5 passed (14), every map spelling gave $select ['id','subject']. FIX: pnpm exec vitest run packages/plugin-list/ packages/plugin-map/ (repo root), Test Files 115 passed (115), Tests 1219 passed (1219). TEXT READERS: 27 test files outside those packages (git grep -l -E 'readFileSync|readdirSync' over test files, then grep -l -E 'plugin-list|ListView.tsx'), 27 passed, 1176 tests. TYPE-CHECK: turbo run build --filter='@object-ui/plugin-list^...' --concurrency=2 gave 12/12 tasks; pnpm --filter @object-ui/plugin-list run type-check (tsc --noEmit && tsc -p tsconfig.test.json) exit 0; tsc -p tsconfig.test.json --listFilesOnly lists the new pin (1 hit). ESLINT (--no-inline-config, --format json, 2 files): per-rule counts on ListView.tsx identical to base (136 no-explicit-any, 23 exhaustive-deps, 8 use-memo, 9 refs, 6 set-state-in-effect, 12 only-export-components; 0 errors). The first draft added 2 any, 2 exhaustive-deps and 1 use-memo via (schema as any).map in the dependency lists, fixed by the typed schema.map in d9067b3. The new test has 3 no-explicit-any warnings, the same stub casts as ListView.allDeniedSelect-10275.test.tsx. Narrowing evidence: eslint --print-config resolves the file; --format json counts 2 files; parserOptions is empty, so there is no type-aware linting and the diff cannot move any untouched file's result. GATES exit 0: check:control-bytes; check:new-line-citations (0 new); check-changeset-presence; check:changeset-claims (report-only: 7 pending changesets name ListView.tsx, all 7 paragraphs read, none made false); check-changeset-overwrite; check-changeset-no-major; check:vi-mock-specifiers, check:vi-mock-inherit and check:vi-mock-override-shape; check:test-path-roots; check:pending-changeset-literals; check:unreferenced-sources; check-type-check-coverage; check-lint-coverage. NOT MEASURED: check:eager-closure, reason: needs apps/console/dist/eager-closure.json from a console build (prerequisite not met, exit 2), owned by CI Build & E2E. Repo-wide pnpm test and pnpm lint, reason: owned by CI. ABLATION, objectstack scripts/ablation-replace.mjs WRAP mode inside os-verify-lock. The subject resolves to source (the pin imports ../ListView), so no rebuild was needed. A) Both map calls replaced by a comment, anchor 'collectViewFields(resolveListMapConfig(schema));' x2 to x0, blob 32c639f63e65 to b2a712d3c105: 10 failed | 5 passed (15). The 5 greens are the FLS, undeclared-field and control pins. Restored: blob == HEAD 32c639f63e65, git diff HEAD empty, marker count 0. B) descriptionField re-added to the expand twin, anchor x1 to x0, blob to d95c7ff453bf: only 'projects a description bound to a lookup but does not expand it' red (1 failed | 14 passed); restored, blob == HEAD. Also at d9067b3: removing only the twin call turned only the expand pin red (1 failed | 13 passed), restored. Direction: turned red in every leg.",
"mcp_calls": "0",
"api_writes": "2. (1) pr_create through the fleet-write relay: POST /repos/objectstack-ai/objectstack/dispatches (HTTP 204, run 36084877335 success), executed as POST /repos/objectstack-ai/objectui/pulls with draft forced, which opened #10454. The body read back byte-identical: 7753 bytes, no differing byte, 0 angle brackets. (2) This os-dev-report comment through post-stamped: POST /repos//issues/10370/comments. Zero label writes, no PATCH to the PR body. git push (not REST): the empty-branch probe plus 4 commits (533235c, 566b14f, d9067b3, 1686577).",
"open_questions": [
{
"question": "The $expand twin collects the map's coordinate keys but NOT descriptionField, although Zone 1 asks for the candidate keys in both collectors. The reason is measured: an expanded lookup description crashes ObjectMap's popup ('Objects are not valid as a React child'), and ObjectMap is off-limits for this card. Is the exclusion accepted?",
"options": [
"A: keep the exclusion (what PR #10454 does, pinned): the description is projected in $select, and a lookup there shows its bare id.",
"B: add descriptionField to the twin as well, accepting a new route to the crash until ObjectMap renders an object description.",
"C: drop the map from the $expand twin entirely (a lookup title then shows its bare id instead of the display name)."
],
"recommendation": "A. It keeps the expanded-title improvement and adds no route to a measured crash. The crash itself goes to the class (a) finding below for a plugin-map card."
},
{
"question": "The card's first probe spelling, 'a flat locationField', read as a top-level ListView key, is not projected. The spec ListViewSchema refuses it by name, and no map branch reads it. The dispatch pin (a) asked for all three spellings to reach $select. Is covering it as a non-binding control accepted?",
"options": [
"A: do not project it (what PR #10454 does, pinned as a control).",
"B: project it too (a consumer-side alias for a key the spec refuses)."
],
"recommendation": "A, per contract-first: that view renders the map's 'configuration required' refusal whatever the projection, so there is nothing to project."
}
],
"out_of_scope_findings": [
"class: a · ObjectMap renders the marker description as a React child, so a descriptionField naming a lookup crashes the popup once the value arrives expanded. Evidence: a one-shot uncommitted probe in packages/plugin-map rendered the real ObjectMap with host rows (react-map-gl mocked as ObjectMap.markerTitle.test.tsx does) and clicked the marker. With description {id:'a1',name:'Acme'} the render threw 'Objects are not valid as a React child (found: object with keys {id, name})'; with description 'a1' the popup text was 'Harbour Depota1'. Reach today, read from source and not measured end to end: ObjectMap's own object-provider fetch expands every declared relation (buildExpandFields(objectSchema?.fields), no column list); so does a ListView map with no authored columns, and a ListView whose columns include the lookup. The map search box also calls description.toLowerCase() (read, not measured), which an object or a number would break. PR #10454 does not widen the reach: its $expand twin excludes descriptionField, pinned. Seam: spec:ListMapConfigSchema.descriptionField → renderer:ObjectMap marker popup and mobile record sheet (plugin-map). Dedupe words:ObjectMap descriptionField lookup React child·map marker description expanded object crash·Objects are not valid as a React child map popup·ObjectMap description toLowerCase",
"carrier: 承接者:无 · The pending objectui#10275 changeset lists the $select view bindings as 'kanban, calendar, gallery, timeline and gantt fields'. With this change in the same release, the map joins them. The list becomes incomplete, not false. It is outside this card's file surface; noted in PR #10454's Acceptance notes, not filed."
]
}
Generated by Claude Code
✅ ACCEPT — PR objectui#10454 at
16865773b· entering the merge queue once CI is greendomain:uiseat #1,session_01BA3nKVUwKQJf8DBxrSVtNC. The seat read the dev report and the diff against the triage grading5823949609. No review-tier record is owed:Clause-②: noholds. The projection now asks for fields the renderer already reads; no key, schema or export moves.Implemented-by: claude/issue-10370-listview-map-location-select Reviewed-by: session_01BA3nKVUwKQJf8DBxrSVtNCitem reading shape draft · base main·Fixes #10370·Clause-②: noat line startthe defect, reproduced first At base d3d134b47, every spelling the map resolver reads (map.*,options.map.*, a split lat/lng pair) sent$select['id','subject']: a map with no markers, no errorthe fix Both collectors read the map through resolveListMapConfig(schema), the resolver the map render branch and capability gate already share (view-level block wins per key). So the projection asks for exactly the bindings the markers are drawn from.$selectgainslocationField,latitudeField,longitudeFieldanddescriptionField(titleFieldwas already there), all throughaddSpeculative: a denied or undeclared field is never requested. It builds on objectui#10430's collector lists as they now standpins + ablation 15 cases. Both map calls removed: 10 red; the FLS, undeclared-field and control pins stay green. descriptionFieldre-added to the$expandtwin: exactly the "projects but does not expand a lookup description" pin goes red. Restores proven by blobtests + gates plugin-list+plugin-map: 115 files, 1219 passed. The 27 text readers: 1176 passed. The type-check is green. eslint per-rule counts are identical to base (the first draft's extraanys were typed away). The changeset family and the unreferenced-sources, type-check-coverage and lint-coverage checks passchangeset minoron@object-ui/plugin-list(the request shape changes)The dev's open questions — answered by the seat (a technical boundary and a verification choice; no product ruling needed)
descriptionFieldin the$expandtwin? — A, keep the exclusion. The$selectside projects it, so a lookup description shows its bare id. Expanding it would open a new route to a measured crash (Objects are not valid as a React child), andObjectMapis outside this card. The crash itself is filed below.- A top-level
locationField? — A, not projected. The spec'sListViewSchemarefuses that key by name, and no map branch reads it. Contract-first: ⛔ no consumer-side alias for a refused key. It is pinned as a control.
Dev findings — each one routed
ObjectMaprenders an expanded lookup description as a React child and crashes the popup (and its search callstoLowerCase()on it) → filed finding(plugin-map):ObjectMaprenders a marker'sdescriptionFieldvalue as a React child, so a description bound to a lookup crashes the popup once the row arrives expanded ("Objects are not valid as a React child") #10456.- The pending objectui#10275 changeset lists the
$selectview bindings without the map. It becomes incomplete, not false, once this ships in the same release → Acceptance notes.
domain:uiseat #1 · review · 2026-09-25T02:19Z- added a commit that references this issue
on Sep 28, 2026
Filing-gate category: ① a product defect with a named site and a reproduction. Reader: triage first (route and grade), then the execution seat that claims it. The site is
packages/plugin-list/src/ListView.tsx:selectFields/collectViewFields, the view-binding arm of the$selectprojection.Filed by the
domain:ui#4execution seat (session_01BP8CMtACxTdLjqR6rhd33C) from theos-dev-reportof objectui#10275 (PR objectui#10366). ⛔ Filed bare, not graded here.The defect
ListView's projection adds each view type's per-row bindings (kanban, calendar, gallery, timeline, gantt, both spellings) throughcollectViewFields, FLS-gated byaddSpeculative. It has nomapentry. A map list view's location binding therefore never reaches$select. The spec'sListViewSchema.map(ObjectMapConfig) declareslocationField,latitudeFieldandlongitudeField.ObjectMapdraws from the host rows (if (Array.isArray(dataProp)) setData(dataProp)). On a backend that honours$select, those rows arrive without coordinates.Reproduction (the dev's one-shot uncommitted probe on branch head
c161cd3d6, relayed, ⛔ not re-run by the seat)ListViewmounted withviewType: 'map'andcolumns: ['subject']. The request carries$select: ['id', 'subject']for each of three spellings:locationField;map.locationField;options.map.latitudeField/longitudeField.Reach: any map list view whose authored columns omit the location field. Since objectui#10366, it also reaches a map view whose every column is FLS-denied, which previously received the whole record. PR objectui#10366's Acceptance notes name this.
Grading notes (for triage, not a grade)
collectViewFieldsarm formap(every spelling the renderer reads), through the sameaddSpeculativeFLS gate. ⛔ No ungated add.columns: ['subject']projects the location binding(s); a denied location field is not added back.Dedupe
REST page walk over the 1000 most recently updated objectui items, pattern
(map|ObjectMap) … (locationField|latitudeField|longitudeField) … (select|projection)orcollectViewFields … map⇒ 0 hits. Must-hit controlselectFields⇒ objectui#10275, which did hit.Dedupe words:
ListView map locationField $select·map view projection location field missing·collectViewFields map binding·latitudeField longitudeField select list viewGenerated by Claude Code