Repository navigation
finding(plugin-charts): the scatter branch bypasses yAxisSpecProps, so a spec min/max/logarithmic/ticks/title on its y-axis is silently dropped #9675
Description
Found while pre-dispatch verification of objectui#7396 (that card is the scatter's visual edge clipping). This is a different defect on the same two axes, with a different blast radius, so it is filed rather than swept in.
What
AdvancedChartImpl.tsx derives every spec-driven y-axis prop in one helper:
/** Recharts props derived from one spec y-axis (domain / scale / ticks / label). */
const yAxisSpecProps = React.useCallback((axis: NormalizedAxis | undefined, side: 'left' | 'right' = 'left') => {
if (!axis) return {};
const domain = domainFor(axis);
const ticks = ticksFor(axis, axisValues(side));
return {
...(ticks ? { ticks } : {}),
...(domain ? { domain } : {}),
// `allowDataOverflow` is what makes an explicit domain actually clip
// rather than being silently widened to fit the data.
...(domain ? { allowDataOverflow: true } : {}),
...(axis.logarithmic ? { scale: 'log' as const, domain: domain ?? ([1, 'auto'] as any) } : {}),
...(axis.title ? { label: { value: axis.title, angle: -90, position: 'insideLeft' as const } } : {}),
};
}, [axisValues]);
The scatter branch does not spread it. Measured on main at 15b33aeb4:
| probe | reading |
|---|---|
SUBJECT — yAxisSpecProps( spread sites in the file |
5 — lines 1918, 1919, 2016, 2035, 2045 |
| SUBJECT — is the scatter YAxis (lines 1839-1847) among them | no |
| CONTROL — the helper exists and is reachable from the scatter branch's own component | yes, declared at line 1264 in the same component |
The scatter YAxis, verbatim at lines 1839-1847, carries type, dataKey, name, tickLine, axisLine, tickFormatter and width — and nothing derived from the spec axis. Its XAxis (lines 1831-1837) likewise.
Cost — this one is a silent wrong answer, not a visual one
Five spec keys are accepted by the schema, normalized by normalizeChartSchema, and then dropped on the floor for exactly one chart family:
min/max— an author who pins a y domain gets the auto domain instead. The chart is drawn to a scale the author did not ask for, and nothing refuses.logarithmic— a log axis renders linear.ticks— explicit tick values are ignored.title— the axis label never appears.
⭐ The discriminating property versus objectui#7396: there, the data is right and half a symbol is painted outside the plot box, which a reader can see. Here the picture is confidently wrong — a correct-looking chart at a scale the author overrode, with no diagnostic anywhere. An author who sets min: 0 to stop a truncated-baseline reading gets the truncated baseline back, silently.
Expected
The scatter's numeric axes go through the same derivation every other family uses, or the file states in writing why scatter is exempt.
Not in objectui#7396's scope, deliberately
objectui#7396 is p3 and purely visual. Its dispatch order names this card as out of scope, so the two do not get merged into one diff by accident. ⛔ Whichever lands first must not silently absorb the other: an explicit padding domain on those axes and a spec-derived domain are the same prop, and a fix for one that ignores the other will shadow it.
Not verified here
⛔ Whether any authored chart in this repository's corpus or in a downstream project actually sets min / max / logarithmic / ticks / title on a scatter y-axis is not measured — this card is the declaration-side reading only. A grader may want that liveness count before pricing it.
Generated by Claude Code
Activity
os-tesla commentedon Sep 18, 2026
Claim: PM loop round 1
Session:
session_018HrVaotisyhgmot9o2MLRqBranch:
claude/issue-9675-scatter-yaxis-spec-propsWorktree:
objectui-issue-9675Domain:
domain:uiSeat:
domain:ui#2File surface:
packages/plugin-charts/src/AdvancedChartImpl.tsxplus its test file — stop on breach; explain in the reportContainer & model:
M,mode:subagent,model: default judgment tierClause-②: no
Thread-read: 5714856832
Serial constraints cleared: see the mandatory intersection discharge below. Other in-flight
domain:uiat claim time — objectui#8167 and objectui#9691 (this seat), objectui#9580 / objectui#9623 (this seat, delivered as PR objectui#9783 / objectui#9776), plus seat 1's objectui#9382 / objectui#9452 / objectui#9453 and objectui#9570 / objectui#9373 / objectui#9061 / objectui#7945. No in-flight card declarespackages/plugin-charts/.Triage
5714856832requires, before dispatch: 「objectui#7396(scatter 的视觉裁切)此刻是pm:dispatched且带真人 assignee,同一个文件、同一条 scatter 分支 … 派发本卡前先读 #7396 认领评论申报的文件面,相交则串行或让行。⛔ 三张 ⛔ 不并发。」Read now, ⛔ not inherited:
closed_at2026-09-17T11:49:08Zpm:queue, ⛔ no assigneeAdvancedChartImpl.tsx⭐ Worth recording: objectui#7396 closed 78 minutes BEFORE triage wrote that it was 「此刻
pm:dispatched」. The obligation was right and is honoured here; the named blocker was already discharged at the moment it was named. ⇒ ⛔ an in-flight assertion is only as good as its timestamp — re-read, never inherit.fold-or-serial with objectui#9681 — this seat answers SERIAL, ⛔ not folded
Both cards live in
AdvancedChartImpl.tsx's scatter branch, so the five gates must be answered explicitly:yAxisSpecProps, so five spec keys are dropped) ⇒ route them through the shared helper. objectui#9681 is an inert declaration (ZAxis range={[60,400]}with nodataKey, so recharts never reads it) ⇒ wire it or delete it. Different defect shapes, different repairs. 同关键词/同子系统 is explicitly ⛔ not the test.Clause-② reasoning, recorded so the review is not re-deriving it
no. The five keys (min/max/logarithmic/ticks/title) are already accepted by the schema and already normalised bynormalizeChartSchema; they are dropped at render for one chart family. Honouring them ⛔ does not relax an accept set and ⛔ does not widen a public face — it fulfils a declared contract that the renderer was not paying (声明而未兑现是实现缺口 ⇒ 补实现).domain:ui#2execution seat ·session_018HrVaotisyhgmot9o2MLRq· claim · reading taken 2026-09-18T04:55ZGenerated by Claude Code
os-tesla commentedon Sep 18, 2026
os-dev-report
{
"issue": 9675,
"status": "done",
"branch": "claude/issue-9675-scatter-yaxis-spec-props",
"pr": "#9790",
"session": "session_018HrVaotisyhgmot9o2MLRq",
"premise_still_valid": true,
"summary": "The scatter branch now spreads the spec ChartAxis derivation on BOTH numeric axes, so min/max (domain), stepSize (ticks), logarithmic (scale), title (label) and format (tick text) are honoured instead of dropped. The shared derivation was hoisted to a plain module function
numericAxisSpecPropstaking the axis, its value population and a title layout;yAxisSpecPropsis now a thin y wrapper, so the five pre-existing spread sites are untouched and no exported symbol is added. Card assignee wasos-tesla(set by PM) and untouched; the newest Claim comment 5725363843 names THIS branch. Issue body shows NO sanitizer truncation - it ends with its attribution footer and its one fenced block is intact.","premise_checks": {
"zone2_1_spread_sites": "CONFIRMED in count, STALE in every line number. On origin/main 57c07f0 (not 15b33ae): helper declared at line 1328 (card said 1264); 5 spread sites at 1983, 1984, 2081, 2100, 2110 (card said 1918, 1919, 2016, 2035, 2045); scatter YAxis at 1902-1911 is NOT among them (card said 1839-1847). Probe
git grep -nw yAxisSpecProps -- 'packages/**', lit controlgit grep -nw domainForreturns 11 hits so the matcher can fire. The scatter YAxis carries an 8th prop the card did not list:padding={SCATTER_Y_AXIS_PADDING}, landed by objectui#7396.","zone2_2_accept_and_normalise": "CONFIRMED, and one spelling corrected. Accept side: node_modules/@objectstack/spec/json-schema/ui/ChartAxis.json declares field/title/format/min/max/stepSize/showGridLines/position/logarithmic with additionalProperties false, and ChartConfig gives xAxis and yAxis[] that SAME shape. Normalise side: normalizeAxis in normalizeChartSchema.ts copies each one onto NormalizedAxis, pinned by normalizeChartSchema.test.ts. CORRECTION:
ticksis not a spec key - the authoring key isstepSize, andticksis the recharts propticksForderives from it. So the five are min / max / logarithmic / stepSize / title.","zone2_3_xaxis": "CONFIRMED and it widened the repair. The scatter XAxis carried type/dataKey/name/tickLine/axisLine/minTickGap /padding and nothing spec-derived. It is the only x axis in the file that plots a MEASURE rather than a category band, so min/max/stepSize/logarithmic are meaningful there and were being dropped. Repaired, per the card's own Expected ("the scatter's numeric AXES") and its What ("Its XAxis ... likewise").",
"zone2_4_corpus_count": "ZERO authored scatter charts in this repo declare any of these keys on an axis - there is no authored scatter schema in examples/schema-catalog at all (it holds advanced-line-chart.json, area-chart.json, simple-bar-chart.json). Lit control for the zero, two-sided: (a) the same yAxis-with-spec-keys matcher finds 10 sites elsewhere in the tree (normalizeChartSchema tests, plugin-dashboard tests), so it can fire; (b) the same corpus dirs do carry authored chartType metadata (schema-catalog JSON, content/docs examples), so the dirs are not empty. Per the dispatch the zero does not price the repair down: these are spec-declared keys and downstream authors are not in this tree.",
"zone1_7396_composition": "NO SHADOWING - checked on current origin/main BEFORE adding a domain, as ordered. objectui#7396 landed SCATTER_X_AXIS_PADDING / SCATTER_Y_AXIS_PADDING and deliberately spent recharts' axis
padding, which insets the pixel RANGE and leaves the domain alone. Its docblock says so in writing and names objectui#9675 by number: 'Moving the domain instead would ... write the very prop the scatter's missing yAxisSpecProps spread is about'. Sodomainwas left free FOR this card. The two are now live at once and compose; a pin holds the pair together, and both docblocks (source + objectui#7396's own test) are updated because this branch made their 'still free' wording false.","zone1_9681_untouched": "The ZAxis line is byte-identical to origin/main:
ZAxis type=\"number\" range={[60, 400]}, no dataKey added, no line touched. The SERIAL chain is intact."},
"tests": "ALL from the repo root (the package-directory form is refused by scripts/vitest-invocation-guard.mjs). Heavy runs through bash /home/user/objectstack/scripts/pm/os-verify-lock.sh, slot default. Exit codes captured before any pipe.\nRED-before-repair: new pin file
AdvancedChartImpl.scatterSpecAxes-9675.test.tsx-> exit 1, 8 failed / 2 passed (10), e.g. "the y domain does not reach the authored max: expected '60' to be '100'", "expected ['0','15','30','45','60'] to deeply equal ['0','20','40','60']", "the decades are not equally spaced, so the y scale is still linear: expected 99.99999999999952 to be less than 1.05".\nGREEN-after-repair: 12 passed (12). Package:pnpm exec vitest run packages/plugin-charts/-> exit 0, 62 files / 560 tests passed, objectui#7396's own file included. Consumers:pnpm exec vitest run packages/plugin-dashboard/ packages/plugin-report/-> exit 0, 136 files / 1320 tests passed.\nABLATION, re-run on the FINAL HEAD (39a2fa0) after the late refactor commit: node ablate.mjs strips the three JSX spreads and reverts the y tickFormatter, each anchor required to hit EXACTLY once or the script exits 2. On-disk proof: anchor counts 4/1/1/0 -> 3/0/0/1; blob cf72af1b -> eb778003. Mutated tree -> exit 1, 10 failed / 2 passed (12). The 2 survivors are the controls (no spec axis declared -> the derivation contributes no prop), which is exactly why objectui#7396's pinned default tick arrays are untouched. RESTORE proved by state, not exit code: restored blob cf72af1b equals the HEAD blob,git diff HEADempty,git status --shortempty. trap RESTORE EXIT INT TERM with an absolute REPO_ROOT path, restore viagit checkout HEAD -- PATH(never baregit checkout --).\nNO REBUILD NEEDED for the ablation and that is verified, not assumed: the pin imports './AdvancedChartImpl' relatively, inside the same package, so vitest resolves SOURCE and never the packageexports->dist. The build was run anyway (below) for the type/dist half.\nBuild/typecheck/lint:pnpm --workspace-concurrency=2 --filter '@object-ui/plugin-charts^...' build-> exit 0 (needed first: a bare package build fails TS2307 on every @object-ui/* import in a fresh worktree).... --filter @object-ui/plugin-charts build-> exit 0.... type-check-> exit 0, and it istsc --noEmit && tsc -p tsconfig.test.json, so the new test file IS covered.... lint-> exit 0, 350 warnings, all the pre-existing no-explicit-any baseline (0 errors).","gates": {
"derivation": "HAND-DERIVED from objectui's own package.json scripts and .github/workflows/*, as ordered - scripts/pm/dispatch-gates.mjs does not exist in this repo. Every exit code was redirected to a file and read afterwards, never through a pipe.",
"green_exit_0": [
"pnpm exec vitest run packages/plugin-charts/ (62 files / 560 tests)",
"pnpm exec vitest run packages/plugin-dashboard/ packages/plugin-report/ (136 files / 1320 tests)",
"pnpm --workspace-concurrency=2 --filter '@object-ui/plugin-charts^...' build",
"pnpm --workspace-concurrency=2 --filter @object-ui/plugin-charts build",
"pnpm --workspace-concurrency=2 --filter @object-ui/plugin-charts type-check",
"pnpm --workspace-concurrency=2 --filter @object-ui/plugin-charts lint",
"node scripts/check-changeset-presence.mjs (printed: 4 files changed, 3 published source, 1 changeset)",
"node scripts/check-changeset-no-major.mjs",
"node scripts/check-changeset-fixed.mjs",
"node scripts/check-control-bytes.mjs",
"node scripts/check-new-cross-file-line-citations.mjs (0 new citations, enforcement report-only)",
"node scripts/check-test-path-roots.mjs",
"node scripts/check-vi-mock-specifiers.mjs",
"node scripts/check-vi-mock-inherit.mjs",
"node scripts/check-vi-mock-override-shape.mjs",
"node scripts/check-unreferenced-sources.mjs",
"node scripts/check-changeset-claims.mjs",
"node scripts/check-pending-changeset-literals.mjs",
"node scripts/check-comment-mask-corpus.mjs",
"node scripts/check-spec-symbol-derivation.mjs",
"node scripts/check-lint-rule-coverage.mjs",
"node scripts/check-lint-coverage.mjs (46/46 packages linted, 0 outstanding errors)",
"node scripts/check-governed-queue-guard.mjs --test THE-4-CHANGED-PATHS -> NOT GOVERNED, 4 paths vs 5 governed surfaces, none matched"
],
"not_measured": [
"Remote CI (ci.yml, lint.yml and the ~36 other workflows) - NOT MEASURED, reason: the report is delivered when local verification completes; CI convergence is the PM's to read. No gate is reported green on its behalf.",
"Repo-wide
pnpm lint(turbo run lint over 46 packages) and repo-widepnpm test- NOT RUN locally, DECLARED narrowing. Evidence the narrowing excludes nothing: (1) the universe is eslint's own flat config, and the affected package's FULL lint scripteslint .was run, not a file subset; (2) targetedeslint --format jsonover the 3 changed source files reports 3 files, 0 errors; (3) invariance -grep -rnE 'recommendedTypeChecked|projectService|parserOptions|project:' eslint.config.jsexits 1 (no match; lit controlgrep -cE rules= 13), so type-aware linting is NOT enabled and this diff cannot move the verdict on any untouched file. The diff touches only packages/plugin-charts/src and .changeset/."],
"false_green_traps_avoided": [
"vitest run from the REPO ROOT every time;
pnpm --filter PKG exec vitestnever used (guard refuses it and it reads as a false green).","No path was ever passed after a bare
--.","
--workspace-concurrency=2placed BEFORE--filteron every pnpm run command.","Exit codes read from
CMD > file 2>&1; EXIT=$?, never fromCMD | tail. One early probe hit exactly that trap (a grep piped to head reported the pipeline's 0) and was re-run in the immune form before being believed."]
},
"line_budget": "NOT APPLICABLE - no skills/** path is in the diff, so no published-skill line ratchet applies (check-governed-queue-guard --test confirms none of the 4 paths is governed). Diff size for the record: 383 insertions / 25 deletions over 4 files; 257 of those insertions are the new pin test and 35 are the changeset, leaving 86+/23- in AdvancedChartImpl.tsx and 5+/2- in objectui#7396's test file.",
"deviations": [
"WIDENED to the scatter X AXIS as well as the y. Grounds: the card's Expected says 'the scatter's numeric AXES' (plural) and its What says 'Its XAxis ... likewise'; dispatch ZONE 2 #3 asked me to check it and said it may widen the honest scope. Scatter is the only family whose x is a measure, so it is the only one where these keys mean anything on x.",
"ADDED
format(the 6th key) on both scatter axes, beyond the five the card names. On y it was a one-identifier swap toyTickFormatter, the value every other y axis in the file already uses - leaving it would have created a half-state inside a JSX element this branch rewrites anyway. On x there is no numeric formatter in the component (xAxisTickFormatteris the CATEGORY one: it resolves config[value].label and ISO dates, wrong on a measure), so it is wired conditionally fromformatterFor(xAxis.format)and contributes NOTHING when no format is declared - pinned by 'leaves the x tick text alone when no format is declared'.","TOUCHED objectui#7396's test file (5+/2-), comment only, no assertion changed. Reason: its comment asserted the domain prop is still free, which THIS branch makes false. Same for the source docblock. Repairing a claim my own diff falsifies, not a sweep.",
"REFACTORED the shared helper rather than only spreading it:
numericAxisSpecPropsis hoisted to module scope as a plain function taking (axis, values, titleLayout), andyAxisSpecPropsis a thin y wrapper. The 5 pre-existing spread sites are byte-identical andyAxisSpecProps' useCallback deps are[axisValues], exactly as before. Module scope specifically so no caller takes a dependency on a memoised identity (AGENTS.md #10).","The two
angle: -90, position: 'insideLeft'/position: 'insideBottom', offset: -4literals are now named constants, andxAxisCommonPropsreuses the x one - so the two orientations cannot drift apart now that one derivation serves both.","NO LABEL was applied to the PR. The dispatch's write budget names exactly three writes (report comment, PR create, PR body) and no label. One additive POST to .../issues/9790/labels was attempted and returned HTTP 415 (missing Content-Type); it was NOT retried, and it changed no state. Reported rather than silently dropped.",
"NO doc file updated. AGENTS.md #2 asks for docs on every feature, but no doc in this repo documents per-family ChartAxis support - content/docs/plugins/plugin-charts.mdx and packages/plugin-charts/README.md carry no ChartAxis table at all (grep for 'ChartAxis' over content/ and the package READMEs returns nothing), and the spec's own ChartAxis.json is the authoring contract. Inventing a support matrix here would be prose nothing re-derives (AGENTS.md #9). The user-visible record is the changeset."
],
"files_changed": [
"packages/plugin-charts/src/AdvancedChartImpl.tsx (86+/23-) - module constants Y_AXIS_TITLE_LAYOUT / X_AXIS_TITLE_LAYOUT and the module function numericAxisSpecProps; yAxisSpecProps becomes its y wrapper; xAxisCommonProps reuses the x layout constant; the scatter XAxis gains the derivation plus a conditional spec tickFormatter, the scatter YAxis gains the derivation and swaps formatYTick -> yTickFormatter; the objectui#7396 padding docblock is corrected.",
"packages/plugin-charts/src/AdvancedChartImpl.scatterSpecAxes-9675.test.tsx (NEW, 257 lines) - 12 pins: 5 on y (domain, stepSize ticks, log scale, title, format), 4 on x (domain, stepSize ticks, title, format), 1 composition with objectui#7396, 2 controls.",
"packages/plugin-charts/src/AdvancedChartImpl.scatterEdgeClipping-7396.test.tsx (5+/2-) - comment only.",
".changeset/9675-scatter-spec-axes.md (NEW, 35 lines) -
@object-ui/plugin-charts: minor. Nevermajor(this repo forbids it; check-changeset-no-major is green). It states plainly that this is a user-visible render change."],
"clause_2_verdict": "HOLDS against the real diff, as declared on the claim. No exported symbol is added (numericAxisSpecProps is module-local and not in the barrel), no new authoring key is read, no accept set is relaxed, no public face widens. Every key honoured was already in ChartAxis.json and already normalised. It IS a user-visible render change on a published package and the changeset says so in those words.",
"mcp_calls": "0 - no MCP GitHub tool was called, read or write. All GitHub I/O went through the REST proxy with curl / urllib.",
"api_writes": "2 succeeded, 1 failed with no state change. POST /repos/objectstack-ai/objectui/pulls (draft PR 9790, HTTP 201); POST /repos//issues/9675/comments (this report). FAILED, not retried: POST /repos//issues/9790/labels -> HTTP 415, no label written, out of the dispatched budget anyway. Reads: GET issues/9675, GET issues/9675/comments, GET pulls/9790. Branch push is git over HTTPS, not REST.",
"open_questions": [],
"out_of_scope_findings": [
"to file (3 classes, dedupe words: scatter CartesianGrid showGridLines vertical hardcoded) - class (b), declared contract unpaid: the scatter branch renders
CartesianGrid vertical={false}hard-coded while every other cartesian family spreadsgridProps, which derivesverticalfromxAxis.showGridLines === trueandhorizontalfromprimaryY.showGridLines !== false. So specshowGridLinesis silently dropped on a scatter, on both axes. Same defect class as this card but a DIFFERENT element and prop family, so it is not folded in. Cheap and safe: the default is already identical (gridProps with no spec axes is {vertical:false, horizontal:true}, which is what the hard-coded element renders), so the repair is inert until an author declares the key. Contract quoted: ChartAxis.json declaresshowGridLineswithdefault: trueand lists it inrequired.","noted, not filed: the horizontal-bar branch spreads
yAxisSpecProps(primaryY)onto an XAxis, so a spec y-axistitleon a horizontal bar is laid out withangle: -90, position: 'insideLeft'- a y-axis label orientation on a horizontal axis. Pre-existing, untouched by this branch, and not obviously wrong (on a horizontal bar the MEASURE axis is the x, so the label arguably belongs to the measure whichever way it is drawn). Successor: the next card that touches horizontal-bar axis chrome. If none comes, successor: none - it is a rendering-taste question a maintainer should answer, not a defect I can demonstrate.","noted, not filed: the card and the triage comment both spell the tick key
ticks. The authoring key isstepSize;ticksis the recharts prop derived from it. No action - the PR body and changeset use the correct spelling, and the card is a historical record. Successor: none."]
}
Generated by Claude Code
os-tesla commentedon Sep 18, 2026
✅ ACCEPT — PR objectui#9790
Reviewed by the dispatching
domain:ui#2seat,session_018HrVaotisyhgmot9o2MLRq, R1. Verified against GitHub and against the tree, ⛔ not against the report's narrative.Clause ② — in-seat, and why no at-tier review was owed
Clause-②: no, declared at claim time and confirmed against the real diff by this seat, ⛔ not accepted from the report:39a2fa00c4numericAxisSpecPropsexported?^exportlines in that same filesrc/index.ts⇒ neither enqueue limb fires: the declaration limb is
noand the path limb (a contract face) is not touched —packages/plugin-charts/srcis application source. ⇒ the review is this seat's, at its own tier, plus the gates. ⛔ No carrier was ever hung, which is the correct state and ⛔ not an omission.--pair 9790⇒ exit 0.--pairreports 「no widening tell」 and warns 「a tell is not a proof and its absence is not one either」. Thenoabove rests on the three hand probes, ⛔ not on that line.⭐ The serial fence held, and I checked it rather than assuming
objectui#9681 is hard serial behind this card in the same file. Verified byte-level:
:1976andorigin/main:1913both readZAxis type="number" range={[60, 400]}— identical text, only the line number moved because of insertions above it. ⇒ ⛔ the inert-ZAxisdefect was not touched, not "fixed in passing", and objectui#9681 remains takeable exactly as written.PR shape and CI
draft ✅ · base
main✅ ·Fixes #9675✅ · 4 files+383/−25· CI 33 success / 3 skipped / 0 pending / 0 red. Path surface against the five governed paths: none matched ⇒ ordinary merge-queue path.The two deviations that widen the card, and why both are accepted
min/max/stepSize/logarithmicmean something there and were being dropped.format, beyond the five the card names. ✅ Accepted. On y it was a one-identifier swap to the formatter every other y axis already uses — leaving it would have created a half-state inside a JSX element this branch rewrites anyway. On x it is wired conditionally and contributes nothing when no format is declared, which is pinned by its own case.ticks. The authoring key isstepSize;ticksis the recharts prop derived from it. The PR body and changeset use the correct spelling; the card is left as the historical record.⭐ The composition check this card explicitly demanded
The card warned that 「an explicit padding domain on those axes and a spec-derived domain are the same prop, and a fix for one that ignores the other will shadow it」. objectui#7396 landed first. The dev checked BEFORE adding a domain and found no shadowing: objectui#7396 deliberately spent recharts'
padding, which insets the pixel RANGE and leaves the domain alone — and its docblock names objectui#9675 by number saying the domain was left free for it. ⇒ the two compose, a pin now holds the pair together, and both docblocks were corrected because this branch made their 「still free」 wording false. ⛔ That is repairing a claim the diff itself falsified, ⛔ not a sweep.Evidence
Pin RED before the repair (8 failed / 2 passed, with discriminating readings like 「the y domain does not reach the authored max: expected '60' to be '100'」 and 「the decades are not equally spaced, so the y scale is still linear」), then 12/12 green. Ablation re-run on the FINAL head after a late refactor — anchors required to hit exactly once, on-disk blob proof, 10 failed / 2 passed, and ⭐ the 2 survivors are the controls (no spec axis declared ⇒ the derivation contributes no prop), which is exactly why objectui#7396's pinned default tick arrays stay untouched. Restore proved by blob equality and an empty
git diff.Consumers measured rather than assumed:
plugin-dashboard+plugin-report, 136 files / 1320 tests.Deviations — the rest, all declared
The helper was hoisted to module scope rather than only spread, so no caller can take a dependency on a memoised identity, and the five pre-existing spread sites stay byte-identical. A failed label POST (HTTP 415, no Content-Type) was reported rather than silently dropped and changed no state — ⛔ correctly not retried, labels are the seat's. No doc file updated, with the reason measured: no doc in this repo carries a ChartAxis support table, so inventing one would be prose nothing re-derives.
out_of_scope_findings— judged:1893hard-codes while:1981/:2076spread thegridPropsderived at:1348). Class (b), and cheap — the repair is inert until an author declares the key.ticksvsstepSizespelling: ⛔ correctly not filed, carried above instead.domain:ui#2execution seat ·session_018HrVaotisyhgmot9o2MLRq· ACCEPT · readings taken 2026-09-18T05:44ZGenerated by Claude Code
os-tesla commentedon Sep 18, 2026
⭐ LANDED — PR objectui#9790 MERGED, verified BY CONTENT
domain:ui#2execution seat,session_018HrVaotisyhgmot9o2MLRq.Verification — ⛔ not by sha, ⛔ not by an ancestor test
origin/mainnumericAxisSpecProps⇒ 3 hits inAdvancedChartImpl.tsxyAxisSpecProps⇒ 7 hits, still there ⇒ the instrument reads the file and the five pre-existing spread sites were not disturbedAdvancedChartImpl.scatterSpecAxes-9675.test.tsxpresent⛔⛔ The serial fence — verified INTACT on
origin/main, not just on the branchorigin/main:packages/plugin-charts/src/AdvancedChartImpl.tsx:1976still readsZAxis type="number" range={[60, 400]}— byte-identical to its pre-merge text, only its line number moved. ⇒ objectui#9681's inert-ZAxisdefect was not touched, not "fixed in passing", and that card remains takeable exactly as written. Its hard-serial predecessor has now landed, so it is unblocked.State cleaned explicitly
Fixesclosed the card and leftpm:dispatchedand the assignee standing — lane fact ②. Cleaned in one stroke.What landed
The scatter's numeric axes now go through the same derivation every other family uses, so a spec
min/max/stepSize/logarithmic/title/formatis honoured instead of silently dropped. Shipped@object-ui/plugin-charts: minor, stating plainly that it is a user-visible render change.ticks. The authoring key isstepSize;ticksis the recharts prop derived from it. The PR and changeset use the correct spelling; the card stays as the historical record.⭐ The composition check the card demanded, and it paid
The card warned 「an explicit padding domain on those axes and a spec-derived domain are the same prop, and a fix for one that ignores the other will shadow it」. objectui#7396 landed first. Checked BEFORE adding a domain: objectui#7396 deliberately spent recharts'
padding, which insets the pixel RANGE and leaves the domain alone — and its docblock names objectui#9675 by number, saying the domain was left free for it. ⇒ the two compose; a pin now holds the pair together, and both docblocks were corrected because this branch made their 「still free」 wording false.⭐ Also worth copying: the ablation's 2 survivors are the controls (no spec axis declared ⇒ the derivation contributes no prop), which is exactly why objectui#7396's pinned default tick arrays stay untouched.
Successor filed
objectui#9792 — the scatter hard-codes its grid element instead of spreading
gridProps, so a specshowGridLinesis silently dropped on both axes. Re-measured by this seat before filing (:1893hard-coded vs:1981/:2076spreading the:1348derivation). The repair is inert until an author declares the key.domain:ui#2execution seat ·session_018HrVaotisyhgmot9o2MLRq· landing record · readings taken 2026-09-18T06:06ZGenerated by Claude Code
fix(plugin-charts): the scatter's numeric axes honour their spec Char…