fix(lint): chart-axis-not-selected resolves a report chart against its own chart.yAxis, not report.values (#15734)

[claude]

Fixes #15734

The ruling

Maintainer ruling recorded at 5550127165 (batch #47, verbatim reply 「同意」) — option 1, quoted:

In packages/lint/src/validate-chart-bindings.ts, the report surface (checkReportChart) resolves the series limb of chart-axis-not-selected against { chart.yAxis } — the one series the embedded chart derives — instead of report.values; the check at the chart.yAxis position itself is removed on that surface, because a report chart cannot fail to select what it queries. chart.series[].name resolves against the same singleton. The list-view and page-component surfaces are untouched.

The renderer at the pin, quoted

Read at the @object-ui revision this repo pins — .objectui-sha = a472b07167a39e55491109e864bb5a54027dcfbd, verified present in the sibling checkout and read with git cat-file -p a472b07…:packages/plugin-report/src/DatasetReportRenderer.tsx.

The chart's query (lines 859–867 at that sha) asks for the axis pair and nothing else:

  const wantsQuery = plan.kind === 'series' || plan.kind === 'single_value';
  const state = useDatasetRows(
    dataset,
    plan.kind === 'series' && xAxis ? [xAxis] : [],
    wantsQuery && yAxis ? [yAxis] : [],
    runtimeFilter,
    dataSource,
    undefined,
    order,
  );

The file says so in its own words at the scopeOrder docblock (line 326): "the embedded chart queries only chart.xAxis × chart.yAxis".

And the series the override is matched against is derived from that same pair (line 1008), exactly one of them — line 1002 above it: "The selection is exactly one dimension × one measure, so this takes the helper's single-dimension branch and returns ONE series":

  const { data: chartData, xAxisKey, series: derivedSeries } = buildChartSeries(
    relabelDimensions(state.rows, dimensionLabels),
    [xAxis],
    [yAxis],

So report.values is the selection of the TABLE beneath the chart, and the warning that fired for a chart.yAxis outside it named a query consequence its own pin refutes.

What changed

ChartBinding gains ownSelection — the measures a chart's OWN query selects, when that is narrower than the values limb. checkReportChart sets it to { chart.yAxis }; the other two surfaces leave it undefined and keep resolving against values, which on them IS the query's measure set.

• The not-selected limb reads ownSelection ?? values, so a report series[].name is measured against the singleton.
• The report chart.yAxis call site no longer passes a selection at all — that position IS the query.
• chart-measure-unknown is untouched at every position, on every surface: an UNDECLARED measure is still no column, still error at a query position.
• report.values is still walked for chart-measure-unknown, unchanged.
• No array-coercion copy introduced (fix(lint): no authoring rule throws on a non-record entry of any stack collection #15751's one-copy gate reads this file): the only collection read is the existing recordsOf.
• The per-position tier and consequence wording fix(devx): check-self-test-workflow-commands declares the scripts/** population it scans #15590 / fix(lint): chart series[].name advises instead of blocking — presentation, not a query binding #15741 landed is untouched — only the SET moves.

Pins flipped and added

Test	Before	After
warns when the yAxis measure is declared but not selected	1 warning	flipped → says nothing when the yAxis measure is declared but outside report.values — a report chart cannot fail to select what it queries, expect(findings).toEqual([])
chart-axis-not-selected at the report yAxis position keeps its wording	1 warning	replaced → chart-axis-not-selected does not fire at the report yAxis position at all, toEqual([]) (the branch it pinned is gone; the wording it guarded is not re-touched)
—	—	new: a report series[].name naming chart.yAxis is silent even when report.values does not select it
—	—	new: a report series[].name naming a declared measure OTHER than chart.yAxis fires — even one report.values does select; asserts the message names the singleton, selected values (est_hours)
—	—	new firing control, list-view surface: resolution unchanged, views[0].list.chart.values[1] still error chart-measure-unknown
—	—	new firing control, page-component surface: both limbs still fire chart-axis-not-selected and both messages name selected values (task_count), i.e. values — never a borrowed singleton

Verification

Exit codes captured before any pipe; verdict lines quoted as the tools printed them. Heavy runs through scripts/pm/os-verify-lock.sh with OS_VERIFY_LOCK_SLOT=issue-15734.

Run	Verdict
pnpm --filter @objectstack/lint exec vitest run src/validate-chart-bindings.test.ts	Tests 35 passed (35) · VERDICT command-exit 0
pnpm --filter @objectstack/lint test (after merging origin/main)	Test Files 97 passed (97) · Tests 3331 passed (3331) · VERDICT command-exit 0
pnpm --filter @objectstack/lint typecheck (after the merge)	check:test-typecheck: OK — @objectstack/lint's test layer compiles · VERDICT command-exit 0
pnpm check:pm-dispatch-gates	dispatch-gates self-test: 1478 cases pass. · EXIT=0
pnpm check:nul-bytes	check-nul-bytes: OK (scanned 7613 text file(s) … no raw ASCII control bytes). · EXIT=0
node scripts/pm/check-governed-merges.mjs --test THE_3_CHANGED_PATHS	NOT governed — ordinary queue landing applies · EXIT=0
node scripts/pm/dispatch-gates.mjs --changed --commands --repo objectstack-ai/objectstack	EXIT=0, 53 commands; all 53 run (see below)

Ablation — the set change, trap-guarded, blob-hash restore

The implementation was committed first, so the restore leg points at HEAD. trap 'RESTORE_FN' EXIT INT TERM; every path absolute from git rev-parse --show-toplevel. No dist leg: the test imports its subject relatively (./validate-chart-bindings.js, a sibling in the same src/), so it can never reach dist/ — nothing to rebuild between legs.

Leg	On-disk proof	Result
mutate — derived reverted to report.values, and selected handed back to the chart.yAxis call	grep -c ABLATION = 2 (expect 2); grep -c 'binding.ownSelection ? new Set' = 0 (expect 0); git hash-object = dc31fa66…, differs from HEAD blob 718a2cd5…	Tests 4 failed | 31 passed (35) · VERDICT command-exit 1
restore — git checkout HEAD -- ABSOLUTE_PATH	git hash-object = 718a2cd58249552fa0b9d46bac3871170ca97edb = HEAD blob; git diff HEAD: EMPTY (restored)	Tests 35 passed (35) · VERDICT command-exit 0

The four that fail under ablation are exactly the flipped pin, the replaced pin and the two new singleton pins:

× says nothing when the yAxis measure is declared but outside `report.values` …
× a report series[].name that names `chart.yAxis` is silent even when `report.values` does not select it
× a report series[].name that names a declared measure OTHER than `chart.yAxis` fires …
× chart-axis-not-selected does not fire at the report yAxis position at all …

The two new firing controls stay green under the ablation, which is the point of them: they measure the two surfaces this change does not move.

Lint — a declared narrowing, with the three readings

pnpm lint (eslint . --no-inline-config) is the repo-level sweep CI owns. Ran targeted instead, at final head 658c596a1:

1. Population, read from eslint's own config (its isPathIgnored over git ls-files, not a guess): 5979 lintable tracked JS/TS files, 0 of them ignored.
2. Files linted, counted from --format json: 2 — validate-chart-bindings.ts, validate-chart-bindings.test.ts; errors 0, warnings 0, EXIT=0.
3. Invariance over the 5977 untouched files: this repo runs one eslint.config.mjs, which "never enables type-aware linting (no parserOptions.project, no typed @typescript-eslint rules) for ANY file" (its own header, line 326). With no cross-file program, a diff confined to two files cannot move the verdict on any file it does not contain.

NOT MEASURED, by name

• pnpm check:dual-build-cjs-loads — exit 3, PREREQUISITE NOT MET — this gate reads built output, and some package has no dist/. Needs the whole workspace built; CI's.
• pnpm check:type-check-debt — exit 3, PREREQUISITE NOT MET, "Build the closure first, exactly as lint.yml does before this step". Its sibling pnpm check:type-check-coverage ran and is EXIT=0.
• pnpm check:docs-transcript-drift — exit 3 on the first pass (packages/lint/dist/index.js absent), then re-run after pnpm --filter '@objectstack/lint...' build and it is EXIT=0: "4 declared transcript value(s) across 405 page(s) under content/docs/ equal what the registry derives today". Measured, not left open.
• The 7 families whose argv takes a value from the workflow — check-cross-package-test-inputs --union-into "$RUNNER_TEMP/…", the three check-shard-attestation --emit jobs, the two check-test-completeness jobs, check-half-states --provenance="$PROVENANCE". dispatch-gates prints them as ⊘ NOT MEASURED itself; no value exists outside a CI run.
• @objectstack/cli unit tier — attempted (vitest run --project unit) and not measured: 113 of 176 files fail before any assertion with Failed to resolve entry for package / Cannot find package (unbuilt dependency closure in this worktree), while AssertionError count is 0 and individual failed tests (×) is 0; 813 of the tests that did resolve passed. Left to CI, which builds first. @objectstack/cli is affected only through the dependency edge — this diff moves no exported symbol or signature.
• The 37 "artifact roster" families and the 10 too-wide-to-place families dispatch-gates prints: silence there is a fact about a list, not a clearance. CI runs them.

Gate family

All 53 derived commands were run: 50 EXIT=0, 3 exit-3 prerequisite cases named above (one of which was then measured green). The derived command list is byte-identical across three derivations — before the merge, after it, and after a further git fetch — so origin/main moving under the branch did not move this card's family.

Scope

Only the SET the report surface resolves chart-axis-not-selected against. origin/main was merged in (merge commit 658c596a1) after PR #15751 landed at 07:16Z, and the lint suite re-run once on the merged tree; #15751's one-copy gate (collection-coercion-single-copy.test.ts) is green here.

Changeset: @objectstack/lint patch, stating the behaviour change.

---

🤖 Generated with Claude Code

https://claude.ai/code/session_012zGPuVVX3deAx9LdjK8jCk

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

3 anchor(s) derived from 1 changed package(s); no hand-written page names any of them, so this run has nothing to list — not a clean bill of health. This check sees only pages that NAME a derived anchor: one that documents this change in prose, or enumerates it in an authoring dialect, names none and stays invisible to it on every run.

What this run could not see

• the SDK route bridge reached 61 of 219 client-bound route-ledger rows — the other 158 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 158: 0 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 56 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 102 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.

Coarse fallback — 5 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json c99449ab5fd5e87292f4efccd15270cecf4d5242 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 94476cdf9c735369e49a01f1e1775d246db75671 — the merge of head 658c596a1b5d977d5663f112cf8105c628af9d39 into base c99449ab5fd5e87292f4efccd15270cecf4d5242, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 94476cdf9c735369e49a01f1e1775d246db75671 && git checkout 94476cdf9c735369e49a01f1e1775d246db75671
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin c99449ab5fd5e87292f4efccd15270cecf4d5242 658c596a1b5d977d5663f112cf8105c628af9d39 && git checkout -B drift-repro c99449ab5fd5e87292f4efccd15270cecf4d5242 && git merge --no-ff 658c596a1b5d977d5663f112cf8105c628af9d39

node scripts/docs-audit/affected-docs.mjs --json c99449ab5fd5e87292f4efccd15270cecf4d5242

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

[claude]

ACCEPT — PR #15789 (head 658c596a1, 3 files: validate-chart-bindings.ts +72/−13, its test +146, changeset) reviewed against the GitHub three-dot diff, not the report.

Ruling 1 (5550127165) as spelled: the report surface passes ownSelection: [chart.yAxis], the series/axes limbs resolve chart-axis-not-selected against that singleton instead of report.values, and the chart.yAxis-position check on that surface is dropped (measureRef(…, 'query') without a selection); list-view and page-component surfaces resolve as before (firing controls added on both). Only the SET moved: the #15590 / #15741 tier and consequence tables are byte-untouched (0 diff lines in UNKNOWN_CONSEQUENCE / SHAPE_HINT / UNSELECTED_CONSEQUENCE), no coercion copy introduced (asArray 0; #15751's one-copy gate green in the dev's run). Pins: the report-surface false positive flips to no finding; two singleton pins added (a series[].name naming chart.yAxis is silent even outside report.values; one naming another declared measure fires and names the singleton). Changeset @objectstack/lint patch present. Renderer read at the pinned a472b07 (:859, :326, :1002-1008) and quoted.

Measured by this seat: git merge-tree --write-tree origin/main <branch> → EXIT=0 (main merged in at 658c596a1, #15751 included); no live PR of this lane touches the file; governed test per the report NOT governed. NOT MEASURED here: the lint suite (the dev's locked runs Tests 35 passed (35) file / 3331 passed package / typecheck OK, and the ablation 4 failed | 31 passed → restored 35 passed, are the record); the whole-workspace-build gates by name.

Flip + arm when both required jobs report success on 658c596a1. On landing #15734 closes.

---

Generated by Claude Code

[claude]

Flipped ready + auto-merge enabled (08:26Z, method: MERGE). Required jobs on 658c596a1: Lint & Repo Gates = success, TypeScript Type Check = success; git merge-tree --write-tree origin/main <branch> → EXIT=0 against the current main. Watched; on landing #15734 closes.

---

Generated by Claude Code