finding(plugin-detail): the SHARED emptiness authority calls an EMPTY ARRAY a value — the record page paints a blank row for [], the exact defect the authority exists to prevent

Found while implementing objectui#8459 (RelatedList's two emptiness spellings). Filed unassigned, ⛔ not claimed. Out of that card's fence: it is a defect in the shared authority itself, on a different surface (the record page body), not in the related-list grid.

What

hasCellValue — THE definition of emptiness for the record page (objectui#8376, widened by objectui#8394, extracted to packages/plugin-detail/src/emptiness.ts by PR #8457) — opens with:

if (value !== null && typeof value === 'object') return true;

typeof [] === 'object', so an empty array is a VALUE. Nothing further is asked of it.

Why that is wrong here, measured

The object half exists for a good reason, and its docblock states it: an object value is handed to a TYPE-AWARE renderer that knows how to draw it — coordinates, a postal address, select badges. But every example the docblock reasons about is a populated object; [] is never considered. And a type-aware renderer has nothing to draw for it. SelectCellRenderer tests value == null || value === '', which [] passes, and then maps it over zero options.

Measured on origin/main (a407bd6), rendering a real DetailSection with tags: [] on a select field:

FULL TEXT: "SNameWidgetTags"
has em-dash anywhere: false
count of em-dash: 0
TAGS ROW HTML:
<div class="text-xs font-medium text-muted-foreground uppercase tracking-wide">Tags</div>

The label renders; the value area is empty. No No value affordance, no em-dash — the visually blank cell the em-dash exists to prevent.

The three consequences are the same three objectui#8376 names

Because all three of DetailSection's reads share this one function, the empty array reproduces the full objectui#8376 triple:

the row paints a blank cell (measured above);
it escapes emptyCount, so "Show N empty fields" reads one too low and revealing the empty rows does not reveal this one;
canCopy is hasCellValue(value), so the row offers to copy an empty array next to a value it did not draw.

shouldAutoHideEmpty needs only filledCount > 0, so a section whose only non-null value is [] can suppress the all-empty skeleton by itself — the objectui#8376 escalation, unchanged.

PR #8457 does not fix it

The extracted emptiness.ts in the merge queue carries the identical predicate (if (value !== null && typeof value === 'object') return true;), so this survives the extraction verbatim. Its docblock's reader list (DetailSection, HeaderHighlight, DetailView summary chips, HistoryTimeline, RecordMetaFooter) is the blast radius: each of those bands asks the same question and gets the same wrong answer for [].

There is already a measured counter-example in the repo

RelatedList.isValueEmpty — the related-list grid's own predicate, and the convergence target chosen in objectui#8459 — draws the finer line and has for some time:

(Array.isArray(v) && v.length === 0)

objectui#8459 measured both directions end-to-end: an all-[] column is correctly pruned there, and a surviving [] cell draws the placeholder. That card deliberately did not delegate to hasCellValue for exactly this reason, and pins the difference in RelatedList.emptinessAgreement-8459.test.tsx (A SURVIVING EMPTY-ARRAY CELL DRAWS THE EM-DASH), which is red for the hasCellValue shape.

Suggested shape (not settled — measurement should decide)

Narrow the object half from "every non-null object is a value" to "every non-empty object is a value": keep objects and populated arrays as values, and let [] fall through to EMPTY. {} deserves its own measurement rather than being assumed — JsonCellRenderer may well draw {} visibly, in which case it is a value and only the array clause moves.

⚠️ Whatever the shape, the pin must assert the rendered outcome and carry non-regression axes that stay green for populated objects and arrays — a predicate that answers EMPTY for every object would pass a naive absence-only pin.

Dedup

Searched objectstack-ai/objectui via MCP search_issues (REST is 403 from the dev container): RelatedList EmptyValue placeholder em-dash aria-label inconsistent hand-rolled span and RelatedList two emptiness spellings disagree pruneEmpty makeCell. Both returned objectui#8459 as a control hit, so the readings are real. Nothing covers the empty-array hole in the authority: objectui#8376, objectui#8394 and objectui#8350 are all about whitespace-only strings.

Claim: session session_01YBWFb5YgMU5dw8p2VKj16S · branch claude/issue-8474-hascellvalue-empty-array

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.

Prioritised because this is a defect in the authority three landed cards converged onto, found by the one card that declined to converge onto it (objectui#8459 / PR #8476).

Decision frame

Settled: hasCellValue's typeof value === 'object' short-circuit is too coarse by exactly one case — typeof [] === 'object', so an empty array reads as a value. objectui#8459's ablation C measured the consequence directly.

Yours to measure and propose: whether the fix is a narrow Array.isArray(value) && value.length === 0 arm, or something that also reconsiders the other object shapes ( {} in particular — measure it, do not assume it behaves like [] ). Measured cost is the whole basis.

⚠️ The blast radius is the point, and it is now three surfaces wide. hasCellValue is read by DetailSection's affordance, its emptyCount/auto-hide, and canCopy; PR #8457 extracts it to packages/plugin-detail/src/emptiness.ts and PR #8476 has just pinned RelatedList's local predicate as deliberately different. Cut from current main and re-derive — several of those are in the merge queue.

⚠️ The full objectui#8376 triple was measured for [] on the record page: it escapes emptyCount, canCopy offers to copy it, and it can arm auto-hide on its own (shouldAutoHideEmpty needs only filledCount > 0). A fixture showing one blank row misses the amplification — build the one that drives auto-hide, the way PR #8396's did.

⚠️ Do not "fix" RelatedList to match. PR #8476 measured that its local predicate is better shaped for a grid and pinned it. If your change makes the two agree, that must be because the shared one moved toward the local one, not the reverse.

⚠️ Ask whether an implementation strictly worse than the bug would pass: a predicate answering EMPTY for everything satisfies every "the [] row draws the em-dash" assertion. objectui#8459's ablation B is the template — its two non-regression axes (0 is a value; a populated object is a value) are what refuse it.

Generated by Claude Code

fix(plugin-detail): an empty array is not a cell value (#8474)

os-dev-report

{
  "issue": 8474,
  "status": "done",
  "branch": "claude/issue-8474-hascellvalue-empty-array",
  "pr": "https://github.com/objectstack-ai/objectui/pull/8482",
  "premise_still_valid": true,
  "base": "c90395b20 (final commit a0ccd5f1c)",
  "summary": "hasCellValue's object short-circuit was too coarse by exactly the one case the card names. Fixed with a narrow arm INSIDE the object branch: `if (Array.isArray(value) && value.length === 0) return false;`. {} was measured on the DOM and deliberately did NOT move. Three files changed: emptiness.ts, a 9-case DOM pin, a minor changeset. RelatedList is untouched.",
  "where_the_predicate_lives": "packages/plugin-detail/src/emptiness.ts. The brief said PR #8457 'extracts' it and is in the merge queue — it has LANDED, as d7fecfbcf, so the extracted file is real on main and no PR-ref reading was needed. DetailSection.tsx only imports it now. PR #8476, by contrast, is still OPEN (mergeable_state behind), so its pin file RelatedList.emptinessAgreement-8459.test.tsx is NOT on this base and nothing here leans on it.",
  "what_empty_object_measured": "{} is a VALUE, and it stayed one. Rendered in a real DetailSection, {} on a json field, an object field AND a location field all draw the literal two-character JSON text through JsonCellRenderer (location falls back to it when the lat/lng chain yields nothing). It DRAWS — there is no blank cell, so none of the defect this card is about is present, and turning a visible {} into an em-dash would be a taste change dressed up as a bug fix. Separately, the shape that would have swept it in is unsafe: Object.keys(value).length === 0 is ALSO true of new Date(0), of a Map with one entry, of a Set with one entry, and of a class instance whose state sits behind getters (all four measured) — a false-empty on values that render, strictly worse for those shapes than the bug. Both halves are asserted in the pin so a future widening must delete an assertion, not merely forget a consideration.",
  "three_consumers_before_after": {
    "DetailSection_affordance": "BEFORE: [] fell through to SelectCellRenderer, whose own guard is `value == null || value === ''` — [] passes it — and which then maps over zero entries, so the cell was a childless flex-wrap DIV. Full section text `\"...industryManufacturingtags\"`, ZERO em-dashes, ZERO 'No value' affordances. AFTER: `\"...tags-em-dash\"`, one affordance carrying aria-label and title 'No value'.",
    "DetailSection_emptyCount_autohide": "BEFORE: [] counted as FILLED. 4 fields with tags:[] plus one absent read `Show 1 empty fields` and the [] row stayed visible and uncounted. AFTER: `Show 2 empty fields`, hidden with the rest, revealed by the toggle.",
    "DetailSection_canCopy": "BEFORE: copy button present on the [] row — it offered to copy an empty array next to a value it had not drawn. AFTER: withdrawn, matching the affordance.",
    "other_two_readers": "HeaderHighlight, DetailView summary chips, HistoryTimeline and RecordMetaFooter all move the same way for free — hasCellValue is not exported outside @object-ui/plugin-detail (grep across the workspace), so the blast radius is exactly its five documented readers."
  },
  "amplification_pin": "Case 4 of the pin, and it is the fixture that drives auto-hide rather than one blank row. 4 fields, only tags:[] non-null. BEFORE (measured on c90395b20): full text `\"Detailstags\"` plus `Show 3 empty fields` — shouldAutoHideEmpty needs only filledCount greater than zero, so the lone [] armed auto-hide by itself and buried all THREE genuinely empty rows, leaving one label over a blank. AFTER: the four-row all-empty skeleton survives, four affordances, no toggle offered. window.innerWidth pinned to 1280 so the desktop thresholds (4 / 0.25) are the ones under test (objectui#8399).",
  "would_empty_for_everything_pass": "No — and the interesting part is WHICH case catches it. Ablation B (hasCellValue returns false for everything) reddens 8 of 9, but the one case it does NOT redden is the star AMPLIFICATION case: an EMPTY-for-everything predicate also leaves the all-empty skeleton standing. So the most impressive-looking assertion in the file is precisely the one that cannot tell the fix from its worst caricature. The cases that refuse B are the two non-regression axes (a POPULATED array plus a populated object; `0` and `false`) and THE BOUNDARY ({} is a value). A second caricature is refused too — ablation C, `Array.isArray(value)` alone so every array is EMPTY — and only the populated-array axis catches that one.",
  "ablation": {
    "method": "Mutations on the READ SITE (packages/plugin-detail/src/emptiness.ts), from the COMMITTED implementation, each proved on disk by blob hash AND by removed/injected grep counts in BOTH directions with the matched line printed; each restore verified BY STATE (git hash-object equal to git rev-parse HEAD:PATH, and git diff HEAD empty), never by an exit code; all legs under trap on EXIT INT TERM with absolute paths.",
    "head_blob": "91e9fdde8ae7672e40e9ceef51bd3a007dc0bddf",
    "A_restore_the_coarse_short_circuit": {
      "mutated_blob": "f119dc261ecd97b279e349051b1513300066acfa",
      "grep_proof": "implementation line 1 -> 0, injected marker 0 -> 1 (matched line printed at emptiness.ts:142)",
      "red_by_name": ["AFFORDANCE — an empty array draws the `No value` em-dash, not a blank cell", "COPY AFFORDANCE — the row that says `No value` does not also offer to copy it", "COUNTER — the empty-array row is counted among the empty fields, and revealing them reveals IT", "star AMPLIFICATION — one empty array must not suppress the all-empty skeleton", "DECLARED COST — a `json` field holding `[]` now draws the placeholder instead of the literal empty-array text", "MEASUREMENT — the arm is `Array.isArray && length === 0`, and the wider shape it rejects is unsafe"],
      "tally": "6 failed | 3 passed (9)"
    },
    "B_empty_for_everything": {
      "mutated_blob": "b5230f00db5c68029dec3be447c4ad1ecf6212fa",
      "grep_proof": "injected marker 0 -> 1 (matched line printed at emptiness.ts:139)",
      "red_by_name": ["AFFORDANCE", "COPY AFFORDANCE", "COUNTER", "NON-REGRESSION — a POPULATED array is still a value, and so is a populated object", "NON-REGRESSION — `0` and `false` are still values", "THE BOUNDARY — `{}` is a VALUE: it DRAWS, so it did not move", "DECLARED COST", "MEASUREMENT"],
      "tally": "8 failed | 1 passed (9) — the survivor is star AMPLIFICATION, see would_empty_for_everything_pass"
    },
    "C_every_array_empty": {
      "mutated_blob": "7dbf02ed52b73b8a00fb1d396d102eb0ee305fa8",
      "grep_proof": "implementation line 1 -> 0, injected marker 0 -> 1 (matched line printed at emptiness.ts:148)",
      "red_by_name": ["NON-REGRESSION — a POPULATED array is still a value, and so is a populated object", "MEASUREMENT — the arm is `Array.isArray && length === 0`, and the wider shape it rejects is unsafe"],
      "tally": "2 failed | 7 passed (9)"
    },
    "restore": "All three legs: restored blob hash equal to HEAD blob 91e9fdde8ae7672e40e9ceef51bd3a007dc0bddf, git diff HEAD empty, implementation line back to 1 occurrence, injected marker back to 0. Tree confirmed clean afterwards and the pin re-run green 9/9."
  },
  "relatedlist_untouched": "CONFIRMED. git diff touches exactly three paths and RelatedList.tsx is not among them: .changeset/8474-empty-array-is-not-a-cell-value.md, packages/plugin-detail/src/__tests__/DetailSection.emptyArray-8474.test.tsx, packages/plugin-detail/src/emptiness.ts. RelatedList.isValueEmpty at RelatedList.tsx:913-917 is byte-unchanged. The two predicates now agree on every probe measured ([] empty; {}, [1], {a:1}, 0 and a Date all values) BECAUSE THE SHARED AUTHORITY MOVED TOWARD THE LOCAL ONE — the direction the brief mandates. PR #8476's measurement stands untouched.",
  "changeset": "@object-ui/plugin-detail: minor — matching the landed precedent for a plugin-detail render change (objectui#8457, and PR #8476 for the same surface). Gate verdict line, quoted: `1 source file(s) of 1 released package(s) changed, and this change declares 1 changeset(s): .changeset/8474-empty-array-is-not-a-cell-value.md.` check-changeset-no-major.mjs: `No changeset declares a major bump.` skip-changeset NOT applied — phantom label here, and the change is user-visible anyway. major is forbidden in this repo.",
  "tests": "Run from the repo root, paths relative to the root, nothing after a bare double-dash. (1) pnpm exec vitest run packages/plugin-detail/ -> 140 files / 1270 tests passed. (2) the new pin alone -> 9/9, and 9/9 again after the ablation harness restored the tree. (3) pnpm --filter @object-ui/plugin-detail run type-check -> exit 0 (tsc --noEmit plus tsc -p tsconfig.test.json). (4) same package run lint -> exit 0, '905 problems (0 errors, 905 warnings)', new test file contributes none. (5) closure build FIRST: pnpm --workspace-concurrency=2 --filter '@object-ui/plugin-detail^...' build -> exit 0. (6) pnpm run check:control-bytes -> exit 0, plus a direct scan of the three changed paths whose lit control fired on an injected U+0001 and whose real scan exited 1 (clean). (7) pnpm run check:unreferenced-sources -> exit 0. (8) governed guard with the three paths passed explicitly -> 'NOT GOVERNED — 3 path(s) checked against 5 governed surface(s); none matched.' DECLARED NARROWING: app-shell, console, plugin-kanban, plugin-calendar, plugin-gantt and examples/console-starter depend on plugin-detail; hasCellValue is not exported outside the package and no test outside it reads the 'No value' affordance (grep matches six files, none a detail-page render test). Ran five app-shell record-detail files anyway -> 5 files / 61 tests passed. Rest left to CI, declared not silently skipped.",
  "mcp_calls": "10 — issue_read get 8474, issue_read get_comments 8474, pull_request_read 8476, search_issues x2, issue_write create 8481, issue_read get 8481 (read-back), create_pull_request 8482, pull_request_read 8482 (read-back), add_issue_comment (this report).",
  "open_questions": [],
  "out_of_scope_findings": [
    "filed as objectui#8481: SelectCellRenderer's OWN opening guard is `value == null || value === ''`, which [] passes, so the renderer itself draws a childless flex-wrap DIV. This card and PR #8476 both guard it with an upstream pre-check inside @object-ui/plugin-detail; the shared renderer in @object-ui/fields is still wrong for every consumer that does not. ObjectGrid is one, verified on c90395b20: it calls the resolved renderer with the raw value at five sites and its only EmptyValue fallback is the no-renderer default path, whose guard has the same hole. ObjectGallery and ObjectKanban mention no emptiness predicate at all."
  ],
  "contradicts_the_brief": [
    "PREMISE STALE, conclusion unaffected: the brief says PR #8457 'extracts' hasCellValue to packages/plugin-detail/src/emptiness.ts and that several of these are in the merge queue. #8457 has LANDED (d7fecfbcf) — the file is on main and I edited it directly rather than reading a PR ref. The complementary half also inverted relative to PR #8476's report, which found emptiness.ts absent from main and had to read the PR ref: on my base it is present, and PR #8476 is the one still open. Neither changes what the fix should be.",
    "A BRIEF INSTRUCTION VINDICATED BY MEASUREMENT, with a twist worth recording: the brief warned that a fixture showing one blank row misses the amplification, and told me to build the one that drives auto-hide. Correct — but ablation B shows the amplification case is ALSO the single case an EMPTY-for-everything predicate does not redden. So the amplification fixture is necessary and is NOT sufficient, and citing it alone as proof against the naive caricature would be wrong. The non-regression axes are what carry that weight.",
    "A REPO-DOC CLAIM I COULD NOT REPRODUCE: objectui AGENTS.md's mutation (1) states tag-shaped fragments are eaten from GitHub bodies and that backticks and fenced code blocks do not protect them. On this session's issue_write create (objectui#8481) literal tag-shaped fragments inside fenced ts blocks came back INTACT on a read-back to the end of the body; the same holds for the PR body just created. One observation is not a refutation — the rule may describe an intermittent or endpoint-specific behaviour, and the failure mode it guards against is silent — so I kept the mitigation (shapes spelled out in words) everywhere anyway, since it costs nothing. Recording it because a rule nobody can reproduce eventually gets ignored for the wrong reason."
  ]
}

Generated by an ObjectUI dev agent for session session_01YBWFb5YgMU5dw8p2VKj16S, branch claude/issue-8474-hascellvalue-empty-array.

Generated by Claude Code

fix(plugin-detail): an empty array is not a cell value (objectstack-a…

Navigation Menu

Repository navigation

finding(plugin-detail): the SHARED emptiness authority calls an EMPTY ARRAY a value — the record page paints a blank row for `[]`, the exact defect the authority exists to prevent #8474

What

Why that is wrong here, measured

The three consequences are the same three objectui#8376 names

PR #8457 does not fix it

There is already a measured counter-example in the repo

Suggested shape (not settled — measurement should decide)

Dedup

Decision frame

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

finding(plugin-detail): the SHARED emptiness authority calls an EMPTY ARRAY a value — the record page paints a blank row for [], the exact defect the authority exists to prevent #8474

Description

What

Why that is wrong here, measured

The three consequences are the same three objectui#8376 names

PR #8457 does not fix it

There is already a measured counter-example in the repo

Suggested shape (not settled — measurement should decide)

Dedup

Activity

os-justin commented on Sep 8, 2026

Decision frame

os-justin commented on Sep 8, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

finding(plugin-detail): the SHARED emptiness authority calls an EMPTY ARRAY a value — the record page paints a blank row for `[]`, the exact defect the authority exists to prevent #8474