# UCM-CHART-14 — The removal was right. The reason on the record is false, and the error message says the opposite of the docblock. - **Repo:** `ui-core-micha`, branch `main` - **Tier:** 3 — shared-core. Version **`3.1.1`** (patch — see "Why 3.1.1 and not 4.0.0"). - **Status:** planned - **Workstream:** `CHART-*` - **Files:** `src/components/charts/ChartFrame.jsx`, `docs/CHART-LAYOUT.md`, `tests/ChartFrame.test.jsx`, `package.json`, and one new dev script + its tests. > **`UCM-CHART-13` shipped the right change for a stated reason that is false.** The mechanics are > correct and are not being reopened: `height` and `aspect` are gone, `minHeight` stays, > `warnOnChartFrameHeightMismatch` is deleted, and parameterising `assertRemovedChartProp` with > `removedIn` was a genuine improvement over what the WO asked for — the error now attributes the > right WO instead of always blaming `UCM-CHART-12`. > > What is wrong is the **justification** recorded alongside it, and one user-visible consequence. ## Part A — Envelope (authoritative WHAT/WHY) ### The three findings **F1 — the shipped consumer measurement is false.** Three places state that nothing passed `aspect`: | where | text | |---|---| | `ChartFrame.jsx:29-30` (docblock) | "no consumer across the five apps passes it (measured against 3.0.1)" | | `docs/CHART-LAYOUT.md` | "it is gone because no consumer across the five apps passed it (measured against 3.0.1)" | | commit `676af5c` message | "aspect was applied … but unused by all five consuming apps" | **Four `ChartFrame`s pass `aspect`**, at `origin/main` of fitness-monitor and in its working tree right now: ``` fitness-monitor/frontend/src/pages/BodyHistoryPage.jsx:295 fitness-monitor/frontend/src/pages/BodyHistoryPage.jsx:444 fitness-monitor/frontend/src/pages/EnvironmentPage.jsx:245 fitness-monitor/frontend/src/pages/EnvironmentPage.jsx:279 ``` Measured 2026-08-22 by parsing JSX opening tags with brace- and string-awareness, across all 13 app repos plus ucm. Same scan confirms **hram is clean**: 16 `ChartFrame`s, 11 carrying `minHeight` only, zero carrying `aspect` or `height` — so `HRAM-CHT-4` (landed, `4e55d1ae`) is unaffected by this WO. **F2 — the runtime error message states the opposite of the docblock 50 lines above it.** ```js // ChartFrame.jsx:79 'Removed -- it was never applied to the frame. Use size on the chart inside the frame …' ``` ```js // ChartFrame.jsx:28, the same file // it WAS wired to `aspectRatio` on the box at 3.0.1, and worked. ``` **The four consumers who hit this error are the four whose cards actually change shape**, and the message tells them the prop never did anything. Deleting it on that advice is a silent layout change. This WO's predecessor argued that "a prop accepted and silently ignored is worse than one that errors" — an error that misdescribes what it is guarding is worse than both. `tests/ChartFrame.test.jsx:254` repeats the same false claim in a comment. **F3 — `3.1.0` is a minor for removing an applied prop with four live call sites.** The removal is breaking, and the minor was chosen *because* of F1. `3.1.0` is published and cannot be unpublished. **The remedy is a corrected record and a truthful error message, NOT a compensating major** — see below; this was argued the other way first and the operator was right. ### Goal The record matches what was measured, the error message tells the truth to the four callers who will see it — and **the consumer count stops being a claim and becomes a command.** ### The actual deliverable: a census, not three corrected sentences `UCM-CHART-12` established the pattern — *the deliverable is the invariant, not the fix*. The analogue here is blunt: **this is the sixth wrong consumer count in this series**, and every one of them came from a grep. | # | claim | reality | |---|---|---| | 1 | `UCM-CHART-8`: 3 `aspect` sites in fitness-monitor | 8 hits; `EnvironmentPage` never listed | | 2 | "8 call sites" | 4 charts (4 frame+chart pairs) | | 3 | `HRAM-CHT-4`: "18 preset files" | 14 files, 19 old-API sites | | 4 | "0 consumers pass `aspect`/`height` to `ChartFrame`" | 4 pass `aspect` | | 5 | "`aspect` is never applied in `ChartFrame`" | applied at `:139`; a `head`-truncated grep hid it | | 6 | `UCM-CHART-13`: "no consumer across the five apps passed it" | the same 4 | Greps cannot see JSX structure: a prop three lines below its component name, an arrow function whose `=>` ends the match early, an apostrophe inside a comment, a `{a > b}` in a sibling prop. Every entry above is one of those. **Ship the parser, and "measured against X" becomes reproducible output instead of a sentence somebody has to trust.** ### Definition of Done - [ ] **`scripts/chart-api-census.mjs`** — takes a workspace root, walks every sibling repo's `frontend/src` (falling back to `src/`), and reports per app: every `` and every ucm chart preset element, with which of `aspect` / `height` / `minHeight` / `margin` / `xAxisAngle` each one carries, and file:line. Only counts components actually imported from `@micha.bigler/ui-core-micha` — a local `BarChart` of the same name is not a consumer. - [ ] **It parses, it does not grep.** Opening-tag scanning must track string literals, template literals, `{}` depth, and `//` + `/* */` comments. **A tag it cannot close is reported as `UNPARSED`, never skipped and never counted as clean** — a census that silently drops what it cannot read is the failure it exists to prevent. - [ ] **Its own tests, over fixtures carrying the four traps that actually broke the ad-hoc scans:** an arrow function in a prop, an apostrophe inside a JSX comment, nested braces, and a prop on a line far below the component name. Each fixture asserts the exact props found. - [ ] **`aspect`'s error message rewritten to be true and actionable.** It must say the prop *was* applied (`aspectRatio` on the card box, up to 3.0.1), that the card's height now follows its content with `minHeight` as the floor, and that a caller who wants the chart taller or shorter sets `size` on the chart inside. Do not describe it as never-applied anywhere. - [ ] **`height`'s message left as it is** — it *was* genuinely dead, so that one is already true. - [ ] **F1's text corrected in all three places** — docblock, `docs/CHART-LAYOUT.md`, and the comment at `tests/ChartFrame.test.jsx:254`. Replace "no consumer passed it" with what the census reports: four call sites in fitness-monitor, removed by `FM-CHART-1`. **Do not quietly delete the sentence** — the point is that a measurement was wrong, and that is worth leaving legible. - [ ] **Version `3.1.1`**, with a CHANGELOG entry saying plainly that **`3.1.0` removed an applied prop under a minor**, that four call sites were affected, and that the dev-mode error is what makes the break loud. Do not publish a compensating `4.0.0`. - [ ] **Run the census and paste its output into this WO's register Notiz.** A WO about unverified counts that lands with an unverified count is self-refuting. ### Why `3.1.1` and not `4.0.0` **This WO changes no API.** Corrected prose, one error-message string, one dev script that never ships. On its own diff it is a patch, and labelling it `4.0.0` would announce a breaking change it does not contain. The counter-argument, made first in this WO and **withdrawn**: no consumer has crossed `3.1.0` — hram is on `3.0.1` (`HRAM-CHT-4`, landed), fitness-monitor on `2.37.0` committed / `3.0.1` in its working tree, jg-ferien on `2.41.1` — so the break is still ahead of all three, and a major would put the boundary where they will actually meet it. **What settles it: the break is not silent.** `assertRemovedChartProp` throws on first render in dev and names the replacement. The protection a major version buys — "read the notes before upgrading" — is already delivered by a hard error at the exact call site, which is strictly better than a version number nobody reads. A major would buy a second copy of a signal that already exists, at the cost of a version number that lies about its own contents. **This is conditional, and the condition is F2.** The guard is only a safety net if it tells the truth. An error saying `aspect` "was never applied" invites precisely the wrong action from the four callers whose cards change shape. **`3.1.1` is defensible because F2 is fixed in it — not independently of that.** If F2 were deferred, this argument collapses and the major comes back. `3.1.0` may additionally be deprecated on npm (`npm deprecate`, operator-only — a registry mutation). Low value here since nobody is on it and `3.1.1` supersedes it immediately; noted as available, not recommended. ### Non-goals - **Do not reinstate `aspect`, and do not touch `minHeight`.** The removal stands; only its stated reason and its error text are wrong. - Do not migrate fitness-monitor's four call sites — `FM-CHART-1` owns them. - Do not rewrite history, unpublish `3.1.0`, or publish a compensating `4.0.0`. - Do not extend the census beyond chart props. It answers "who passes what to the chart API", not "who uses ucm". ### Risks - **The census must stay out of the published package.** It is a dev script: keep it off the `exports` map and out of `files`, and confirm the existing `packageTreeShaking` test still passes — that test has been flaky under concurrent load before (`UCM-CHART-12`'s note), so a failure there needs re-running alone before it is believed. - **The census reads sibling repos and must degrade gracefully** when a repo is absent, has no frontend, or is a fresh clone — report "not present", never throw. It must never write anything. - **`FM-CHART-1` is mid-flight** with fitness-monitor's pin at `3.0.1` and all four `aspect` props still in place. Until it lands, the census correctly reports 4 — that is the expected output, not a failure. Do not "fix" fitness-monitor from here. - The corrected error message will be seen by that session. Getting F2 right is what makes its remaining work safe, so **do F2 before anything else in this WO.** ### Tests to WRITE — narrow - `ChartFrame` + `aspect`: the thrown message names `minHeight` and `size`, and **does not contain "never applied"**. Assert on content, not just that it throws — `13`'s test matched `/ChartFrame.*aspect.*size/s`, which the false message passes. - `ChartFrame` + `height`: unchanged, still green. - The census fixtures above. - Nothing else. `13`'s suite stays as it is apart from the corrected comment and the tightened message assertion. No full suite. --- ## Part B — Implementation map > **TO BE FILLED BY THE ORCHESTRATOR** on `git pull`. Must carry: `ChartFrame.jsx:22-35` (docblock) > and `:72-81` (the two `assertRemovedChartProp` calls); `assertRemovedChartProp` itself at > `chartDefaults.js:566`; the `docs/CHART-LAYOUT.md` "What stays unchanged" block; the test file's > `describe` at `:257` and comment at `:254`; the absolute working directory; the progress contract; > and the preamble. **Do not dispatch while this placeholder stands.** --- ## Part C — Orchestrator-only > **TO BE FILLED BY THE ORCHESTRATOR.** Review routing (Tier 3: `reviewer` + `ui_reviewer` > concurrent), register maintenance including the pasted census output, publish verification for > `3.1.1`, commit, and the execution directive with its self-address guard.