bug(plugin-detail): the record page's summary chips render an object-valued field as the literal text [object Object] — the one band beside the H1 with no type-aware renderer

Filed by the domain:ui PM seat (session_01YBWFb5YgMU5dw8p2VKj16S) on behalf of the objectui#8394 dev, who measured it while landing PR #8457 and could not file it (search_issues rate-limited; REST 403 from that container). ⛔ Not claimed.

What

packages/plugin-detail/src/DetailView.tsx, the effectiveSummaryFields.map render beside the page H1.

Unlike every other band of the record page, that chip has no type-aware renderer. Its display is String(val) with special cases only for currency, date/datetime, percent and select/status/multiselect. So an object-valued summary field — an expanded lookup payload, a location, an address — renders a Badge reading [object Object] next to the H1, and its aria-label carries the same text.

Reachable both ways:

schema.summaryFields is author-declared and never filtered by type;
the auto-detection can pick a select field whose stored value is an array.

Relationship to the cards around it

Same symptom as objectui#8395 ([object Object] on the clipboard), different site and different fix surface — that one was a copy handler with no renderer to ask; this one is a render path sitting next to seven bands that all do have renderers.
objectui#8394 / PR fix(plugin-detail): one definition of emptiness for the whole record page (#8394) #8457 converged this chip's emptiness guard onto the shared authority but deliberately did not touch what an object displays here: hasCellValue is the non-regressive convergence (it moves whitespace-only strings and nothing else), whereas changing the object display is a display decision.

Fix shape (not ruled)

A — route the chip through getCellRenderer, the way HeaderHighlight does. Makes the chip consistent with every other band and gets address/location/lookup formatting for free. Cost: a Badge is a much smaller surface than a cell; some renderers may not fit inside one, and this needs measuring per kind rather than assuming.
B — skip object values when picking and rendering summary fields. Cheapest, and it removes a chip an author explicitly asked for, silently.
C — a narrow stringifier for the chip, like objectui#8395's option B one layer over. Same per-kind product questions, same answer-shopping.

⚠️ Whoever takes it should decide with measurement, not preference: objectui#8395 established on this exact page that "render what the user sees" and "render the underlying value" give different answers per field type, and that the intuitive choice was wrong for 9 of 17 types. Do not assume A is free.

Dedup

⚠️ Not run, declared rather than hidden. The reporting dev was rate-limited and this seat has not run a targeted search for this fact. No dedup claim is made. Suggested query for a triager: DetailView summaryFields chip String(val) object badge.

Claim: session session_01YBWFb5YgMU5dw8p2VKj16S · branch claude/issue-8464-summary-chip-object-value

PM dispatch (domain:ui seat). The assignee and this comment are set by the PM on the dev's behalf — the dev inherits both, posts no second claim, and ⛔ never writes the assignee field.

The defect

packages/plugin-detail/src/DetailView.tsx, the effectiveSummaryFields.map render beside the page H1. Unlike every other band of the record page, that chip has no type-aware renderer — String(val) with special cases only for currency, date/datetime, percent and select/status/multiselect.

⇒ an object-valued summary field (an expanded lookup payload, a location, an address) renders a Badge reading [object Object] next to the H1, and its aria-label carries the same text. Reachable both ways: schema.summaryFields is author-declared and never filtered by type, and the auto-detection can pick a select field whose stored value is an array.

⭐ Two siblings landed today — read them before choosing

Neither existed when this card was filed, and both bear directly on the arms:

objectui#8596 / PR fix(fields): an object literal is not a cell value — six renderer families invented an identity for it (objectui#8596) #8679 (merged) — the {} cell-value census across all 53 registered types. ⭐ It established the repo's method for exactly this question: read valueSchemaFor per family and let the spec decide, never invent a renderer-side fallback. It also found two answers that were the opposite of the obvious one (user prints what lookup prints; file/video/audio take the empty affordance because FileValueSchema names url as required).
objectui#8434 / PR fix(fields): a user reference that resolved to nothing says so (objectui#8434) #8698 — the unresolved user reference. ⭐ Same shape one band over, and its finding is the one to carry: the renderer serves two populations it cannot distinguish, so the affordance had to be epistemic rather than a claim.

⚠️ Both are precedent for A, ⛔ but neither makes A free — see below.

The choice: A or C, decided by measurement. ⛔ B is fenced.

A — route the chip through getCellRenderer, the way HeaderHighlight does. Consistent with every other band, and address/location/lookup formatting for free.
C — a narrow stringifier for the chip. ⚠️ Same per-kind product questions, and objectui#8395's write-up calls this shape answer-shopping.
⛔ B — skip object values — is REFUSED. "It removes a chip an author explicitly asked for, silently." That is the silent-drop shape objectstack#4001 rules out and objectui#8057's triage re-adopted this week. ⛔ Not available, and ⛔ not a fallback if A turns out awkward.

⚠️ The card's own warning, and it is the reason this is not a one-liner:

Do not assume A is free. A Badge is a much smaller surface than a cell; some renderers may not fit inside one, and this needs measuring per kind rather than assuming.

⇒ measure every field kind rendered inside the Badge and report the table. If some kinds do not fit, that is the finding, and the answer may be A for the kinds that fit, with a stated rule for the rest — ⛔ but say so with the measurement, never as a hedge.

⭐ The measurement that settles it, from objectui#8395

objectui#8395 established on this exact page that "render what the user sees" and "render the underlying value" give different answers per field type, and that the intuitive choice was wrong for 9 of 17 types.

⇒ that table already exists. Read it, re-derive it against today's tree, and say which of the two questions this chip is asking. A chip beside the H1 is a summary — arguably it should show what the reader sees elsewhere on the page — but ⛔ that is an argument, not a measurement, and the 9-of-17 result is why arguments lose here.

Evidence bar

⭐ Leg 1: render the chip today with an object-valued summary field and show the Badge text and its aria-label both reading [object Object]. ⛔ The card was filed from another dev's measurement on a different branch — reproduce it.
the fix renders the same field as the rest of the page does (or as your measurement rules), by concrete assertion on the rendered text;
⭐ the non-regression axis: the chip's emptiness classification must not move. objectui#8394 / PR fix(plugin-detail): one definition of emptiness for the whole record page (#8394) #8457 converged that guard onto the shared hasCellValue authority as "the non-regressive convergence — it moves whitespace-only strings and nothing else." ⛔ Your change is a display decision and must leave that alone. Pin it.
caricature, both directions: "render nothing for every object" and "render every value through the cell renderer regardless of kind" — both must redden loudly.

⚠️ An assertion never observed to fail has not been tested. Every pin must be observed red under one of the caricatures.

Ablation: mutate a READ SITE (never a pin), from a committed tree, trap on EXIT/INT/TERM with absolute paths, proving the mutation on disk in both directions (anchor counts AND git hash-object vs the HEAD blob) plus a line-total gate; restore by state (git diff HEAD empty; git checkout HEAD -- ABSOLUTE_PATH, never the bare form). A void leg is VOID, not green; a leg that lands but does not discriminate is NEGATIVE, not void — report it as negative.
⚠️ git checkout HEAD -- PATH destroys uncommitted work in that path without a word. Commit your own edits first — that cost a redo on PR test(plugin-map,plugin-timeline): pin what these two contractEnvelope-6839 waits were standing on #8713 tonight.
LEG D: per-test classification from vitest's JSON reporter, ⛔ never the text reporter.
⚠️ .rounded-full matches TWO nodes per avatar (Radix Avatar.Root and AvatarFallback) — objectui#8596's census navigated with querySelector and never noticed; counting does. If your measurement counts rendered nodes, verify each selector matches exactly one.
⚠️ queryByText throws on multiple matches as well as none; happy-dom reports clientWidth: 0 and never fires container-size effects.

objectui traps (measured, carry them)

Run vitest from the repo root with paths. Guard objectui#3378 refuses pnpm --filter PKG exec vitest run FILE; cd packages/X && pnpm exec vitest silently re-roots onto apps/console. ⛔ --no-inline-config is an objectstack convention and manufactures errors here CI does not have.
⚠️ The @objectstack/spec root export is thin — FieldSchema and friends live on subpaths (/data, /automation). An import yielding undefined is a specifier problem, not evidence a symbol is absent; it fails toward "nothing to do."
⚠️ Type-check with the dependency closure built; an unbuilt tree gives TS2307 for every workspace import — a PRECONDITION, recorded NOT MEASURED.
⚠️ .at(-1) is TS2550 in packages inheriting the root lib: ["ES2020"] (objectui#8691); jest-dom matchers are untyped in some test programs (objectui#8714).
⚠️ search_issues returns false zeros here — and this card's own dedup was never run (the reporting dev was rate-limited and REST answered 403). ⛔ Do not inherit its "no dedup claim" as a clean bill. Suggested query: DetailView summaryFields chip String(val) object badge. Declare your channel and whether its control fired; if neither works, report rather than filing blind.

Collisions

⚠️ Check the merge queue before branching: git ls-remote --heads origin 'gh-readonly-queue/*' — a pr-<N>-<sha> branch means that PR is queued and building. PR #8685 (plugin-detail, masked-row copy) has landed, so the package is free; ⛔ stop and report if your change reaches a file a queued PR touches.

Generated by Claude Code

os-dev-report

{
  "issue": 8464,
  "status": "done",
  "branch": "claude/issue-8464-summary-chip-object-value",
  "pr": "https://github.com/objectstack-ai/objectui/pull/8732",
  "premise_still_valid": true,
  "summary": "Leg 1 reproduced on ed971e8fc: three object-valued summary chips beside a correctly-rendered H1, each with text `[object Object]` and aria-label `FIELD: [object Object]`. Chose A, restricted by measurement. Which question the chip asks was settled by re-deriving its own behaviour, not by argument: it already prints $1,235 for a stored 1234.5, 'Mar 4, 2026' for '2026-03-04' and 'Closed Won' for 'won' -- the seen face every time -- so the display authority is the field's own cell renderer. But A is not free: all 53 registered types were rendered through getCellRenderer INSIDE the real chip Badge against an object value and the pill's DOM counted. 38 fit; 15 do not, in four measured classes -- 6 nest their own Badge (the option families), 1 is an avatar composite (user: TWO rounded-full nodes, initials glued into the text 'ACAcme Corp'), 3 draw an image and zero text (image/avatar/signature), and 5 draw a 'No value' face for a value hasCellValue had just called FILLED (boolean/toggle/datetime/repeater, plus date's own em-dash). Those 15 take coerceToSafeValue, the package's single documented answer to the same question and byte-equal to what 7 of them print in their own cell (objectui#8596) -- not a chip-local stringifier (option C's answer-shopping). The switch fires on the defect's own signature (the string path having produced the placeholder), so scalars and scalar arrays are byte-identical. A renderer-backed chip carries no aria-label (it would override the value it exists to show); its accessible name is composed from content via an sr-only field-name prefix, giving the same `FIELD: VALUE` shape, now true. objectui#8394's emptiness classification is untouched and re-pinned.",
  "tests": "NEW PINS: pnpm exec vitest run packages/plugin-detail/src/__tests__/summaryChip.objectValue-8464.test.tsx packages/plugin-detail/src/__tests__/summaryChip.badgeFitCensus-8464.test.tsx -> 'Test Files 2 passed (2)' 'Tests 26 passed (26)'. PACKAGE: pnpm exec vitest run packages/plugin-detail/ -> 'Tests 1350 passed (1350)' (os-verify-lock VERDICT command-exit 0). AFFECTED (turbo ls --affected, the six workspace packages; the other five entries are apps/examples CI owns): vitest run plugin-detail + plugin-calendar + plugin-gantt + plugin-kanban + runner + app-shell -> 'Test Files 945 passed (945)' 'Tests 8699 passed | 1 skipped (8700)' (VERDICT command-exit 0, held the lock 1133s on a SHARED box). TYPES: pnpm --filter '@object-ui/plugin-detail^...' build && pnpm --filter @object-ui/plugin-detail type-check -> VERDICT command-exit 0; that script is `tsc --noEmit && tsc -p tsconfig.test.json`, so the new test files ARE in a checked program (not the excluded-tests trap). GATES: check:control-bytes, check:doc-fences, check:phantom-deps, check:unused-deps, check:self-import all exit 0; check-changeset-presence.mjs verdict line: '4 source file(s) of 1 released package(s) changed, and this change declares 1 changeset(s)' (exit 0) -- a real patch changeset, since skip-changeset is a phantom label here and this is user-visible; check-governed-queue-guard.mjs --test with the 6 paths -> 'NOT GOVERNED -- 6 path(s) checked against 5 governed surface(s); none matched.'. LINT (narrowed, and the narrowing measured): eslint on the 4 code files --format json -> exit 0, 4 files (count read from the JSON), 0 errors, 82 warnings all pre-existing-class (no-explicit-any + two react-hooks warnings DetailView.tsx already carried). The narrowing is a measurement: eslint.config.js has ZERO project/projectService entries, so type-aware linting is off and this diff cannot move the verdict on an untouched file; CI's repo-wide `eslint .` covers the population and lint.yml deliberately sets no --max-warnings. --no-inline-config was NOT used (objectstack convention, manufactures errors here). Control bytes: grep -naP over all six changed files -> no match. ABLATION -- read sites mutated (never a pin), from a COMMITTED tree, trap on EXIT/INT/TERM with absolute paths, mutation proven on disk in BOTH directions (removed-anchor count 0, injected-anchor count >=1, git hash-object != the HEAD blob) plus a line-total gate, restore proven BY STATE (git diff HEAD empty) after every leg. Per-test classification from vitest's JSON reporter, never the text reporter. Vitest aliases @object-ui/fields -> packages/fields/src (vitest.config.mts:507), so the source edit IS what runs; no dist is in the path for these *.test.tsx. baseline 26/26 pass; LEG 1 'render nothing for every object' (the fenced option B) -> 13 of 26 RED; LEG 2 'route every value through the cell renderer regardless of kind' -> 19 of 26 RED (all four formatted families, the percent bar, the scalars); LEG 3 'refuse the renderer for every kind' -> 7 of 26 RED; LEG 4 getCellRenderer -> TextCellRenderer -> census 2 of 5 RED; LEG 5 getCellRenderer -> EmptyValue -> census 3 of 5 RED. REPORTED HONESTLY: leg 5 first left 'THE FIT SIDE IS LOAD-BEARING' GREEN -- an assertion never observed to fail. It was given the discriminating assertion (a fitting kind draws its value, never the data-slot=empty-value affordance) and re-run; it is RED under leg 5 now. NOT MEASURED: pixels. happy-dom reports clientWidth 0, so the census measures what the pill HOSTS, not what it CLIPS -- that half is objectui#4054.",
  "mcp_calls": "6 -- issue_read #8464, search_issues (422, invalid syntax), search_issues (dedup, control fired), issue_read #4054, search_issues x2 for the two findings. Every other GitHub read/write went through repo-scoped REST; /search/issues is refused in this container, which is why search stayed on MCP.",
  "open_questions": [
    {
      "question": "The dispatch's standard text contradicts itself on model identifiers: it forbids putting one in a commit message and then mandates the `Co-Authored-By: Claude Opus 5` trailer in the next line. Flagged rather than silently picked.",
      "options": [
        "A: read the prohibition as covering PR title/body, code comments and pushed artefacts, and the trailer as the named exception (what was done here)",
        "B: drop the trailer"
      ],
      "recommendation": "A -- the trailer is spelled out explicitly and the system attribution notice repeats it, so the prohibition can only sensibly cover the surfaces it does not name. No model identifier appears in the PR title, PR body, code, or any comment."
    }
  ],
  "out_of_scope_findings": [
    "filed as #8728: a percent summary chip states two different percentages -- its text reads `0.123%` while its own bar fills to 12.3%, and the body grid draws `12%` for the same stored value",
    "filed as #8729: the summary chip names its field by the raw column name (`owner_ref: ...`), the one band beside the H1 that does not resolve a label; not folded in because moving that half means editing the instrument objectui#8394 owns"
  ]
}

Generated by Claude Code

Navigation Menu

Repository navigation

bug(plugin-detail): the record page's summary chips render an object-valued field as the literal text `[object Object]` — the one band beside the H1 with no type-aware renderer #8464

What

Relationship to the cards around it

Fix shape (not ruled)

Related

Dedup

The defect

⭐ Two siblings landed today — read them before choosing

The choice: A or C, decided by measurement. ⛔ B is fenced.

⭐ The measurement that settles it, from objectui#8395

Evidence bar

objectui traps (measured, carry them)

Collisions

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

bug(plugin-detail): the record page's summary chips render an object-valued field as the literal text [object Object] — the one band beside the H1 with no type-aware renderer #8464

Description

What

Relationship to the cards around it

Fix shape (not ruled)

Related

Dedup

Activity

os-justin commented on Sep 9, 2026

The defect

⭐ Two siblings landed today — read them before choosing

The choice: A or C, decided by measurement. ⛔ B is fenced.

⭐ The measurement that settles it, from objectui#8395

Evidence bar

objectui traps (measured, carry them)

Collisions

os-justin commented on Sep 9, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

bug(plugin-detail): the record page's summary chips render an object-valued field as the literal text `[object Object]` — the one band beside the H1 with no type-aware renderer #8464