@micha.bigler/ui-core-micha 3.14.4 → 4.0.1

package/work-orders/UCM-CHART-10.md

@@ -1,150 +0,0 @@
-# UCM-CHART-10 — The rotated-tick allowance is reserved twice, and the surplus is dead space
-
-- **Repo:** `ui-core-micha`, branch `main` (shared-core; publishes from `main`, no staging step)
-- **Tier:** 3 — shared-core; every consumer's chart proportions change.
-- **Status:** planned
-- **Workstream:** `CHART-*`
-- **Files:** `src/components/charts/chartDefaults.js` (`spaceForRotatedTicks`, `rotatedTickMetrics`)
-  and its tests. The presets only *call* the helper — do not change them.
-
-> **Why — measured in hram, 2026-08-22, with ucm 2.42.1 loaded and real data.** `spaceForRotatedTicks`
-> adds the same `extraHeight` in **two** places:
->
-> ```js
-> return { ...axis, height: MUI_LABELLED_X_AXIS_HEIGHT + metrics.extraHeight };   // (1) the axis band
-> ...
-> ? { ...margin, bottom: MUI_CHART_MARGIN_BOTTOM + largestExtraHeight }           // (2) the chart margin
-> ```
->
-> **In MUI v8 the axis `height` and the chart `margin.bottom` are additive.** The allowance for
-> rotated tick labels is therefore counted twice, the axis band already holds the labels, and the
-> margin copy is left over as an empty band below the plot.
->
-> Measured on the Research → Allocation drill-down at 1280 px:
->
-> | card | surface height | plot ends at | empty below the axis | x tick labels |
-> |---|---|---|---|---|
-> | Ward Metrics | 340 | 131 | **130 px (38 %)** | 35 |
-> | Access Ladder | 340 | 131 | **130 px (38 %)** | 35 |
-> | Cost | 340 | 194 | 92 px | 35 |
-> | Division Metrics | 340 | 199 | 85 px | 5 |
-> | Cost-effectiveness | 420 | 331 | 44 px | 22 |
-> | Expert Mode: Simulation Trajectory | 340 | 287 | 8 px | 31 |
->
-> Two things make this conclusive rather than suggestive. First, the empty band **scales with the
-> label load** — 35 long rotated ward names cost 130 px, short labels cost 8 px. Second,
-> `emptyBelowLowestLabel` equals `emptyBelowAxisLine` on **every** chart measured: the tick labels sit
-> inside the axis band, and the space below them is occupied by nothing at all.
->
-> **This is the same failure class as hram's tornado** (`RES-28`), where `margin.left: 220` was
-> reserved *in addition to* an unsized `yAxis`, and as `UCM-CHART-8`'s `minHeight`/`height` pair:
-> **space reserved in two places that add up.** Third instance in this file. Neither `UCM-CHART-8`
-> nor `UCM-CHART-9` could have caught it — both worked on the wrapper box; this one is in the axis
-> margin arithmetic.
-
----
-
-## Part A — Envelope (authoritative WHAT/WHY)
-
-### Goal
-
-The rotated-tick allowance is reserved **once**. A chart with long rotated x labels uses its surface
-for the plot, not for an empty band under the axis.
-
-### Definition of Done
-
-- [x] **`extraHeight` is applied once, not twice.** Kept on `axis.height` only; `margin.bottom`'s
-      addition removed. MUI needs `axis.height` — the live measurement below shows rendered labels
-      sit fully inside the axis band with nothing below them, i.e. the band already fully
-      accommodates the rotation; `margin.bottom`'s copy was reserving space nothing used.
-- [x] **The plot area grows by the reclaimed amount.** Measured live on hram (`CLI-Kilombero_2026-per_cell-network`,
-      real data, 1280 px), same three cards, same candidate, before vs. after swapping in the fixed
-      build: Ward Metrics (35 labels, 340 px) 129.6 px → 51.6 px empty; Access Ladder (35 labels,
-      369 px) 156.9 px → 80.6 px; Division Metrics (5 labels, 340 px) 85.2 px → 41.2 px — each
-      roughly halved, matching "remove exactly one of two equal additive reservations." The residual
-      is the legitimate rotation-geometry padding this WO's non-goals say not to chase.
-- [x] **Tick labels are not clipped.** Checked at both 1280 px and 375 px on the same three live
-      cards (lowest label's rendered bottom never exceeded the SVG's own bottom edge at either
-      width) — no clipping introduced by reserving less.
-- [x] **Charts with no rotation are byte-identical.** Unit-tested directly against
-      `spaceForRotatedTicks` (angle-0 axis entry and margin object both pass through by reference,
-      not just by value) — see `tests/chartDefaults.test.js`, `spaceForRotatedTicks (CHART-10)`.
-- [x] **A caller-set `margin.bottom` still wins.** Implemented as an unconditional guarantee rather
-      than the named `callerSetBottom` escape hatch: `spaceForRotatedTicks` no longer writes to
-      `margin` in any branch, so a caller-set value is never at risk regardless of rotation — a
-      strict superset of the old conditional guard, confirmed by both independent reviewers as a
-      clean simplification, not a behaviour gap. The named variable/constant it used
-      (`callerSetBottom`, `MUI_CHART_MARGIN_BOTTOM`) were removed as dead code once nothing reads
-      them.
-
-### Investigate in the same pass — likely the same helper
-
-hram reports **x-axis tick marks rendering without their labels** on scatter charts
-(`AccessGapScatterPanel`, the optimization cost-effectiveness scatter) — `HRAM-RES-29` item 2, still
-open. A formatter exists on both axes (`ScatterChart.jsx:193,199`), so it is not a missing formatter.
-**The suspicion is that it is this same function in the other direction:** `axis.height` set to a
-band the labels do not fit, so MUI draws the ticks and clips the text.
-
-- [x] **Refuted.** `ScatterChart.jsx` never calls `spaceForRotatedTicks` at all (grep-confirmed:
-      only `BarChart.jsx`/`LineChart.jsx` import it) — its x-axis `height` is never touched by this
-      helper in either direction, so this function cannot be the cause. Confirmed further at the two
-      named hram call sites (`AccessGapScatterPanel.jsx:203`, `OptimizationResultsPanel.jsx:1014`):
-      neither sets `tickLabelStyle.angle` on its `xAxis`, so `rotatedTickMetrics` would return `null`
-      even if the helper were in the path. This matches hram's own `HRAM-RES-29` register row
-      (`WORK_ORDERS.md`), which had already independently traced and closed item 2 the same day
-      (2026-08-22, before this WO's own "still open" line was written): the real cause was
-      `ScatterChart.jsx` rendering `<MuiScatterChart>` with no explicit `height` pre-`11d362c`
-      (`UCM-CHART-8`, `2.41.3`), racing MUI's own collision-avoidance against the wrapper's
-      `aspect-ratio` CSS and clipping tick text with no missing formatter involved. `2.42.0` already
-      fixes it for every consumer by passing a resolved `height` explicitly; hram's `develop` already
-      pinned `2.42.0` (why local measurement looked clean) and only `main`/`hram.ch` (still on
-      `2.41.3` at the time) showed it — closing on hram's next `develop -> main` promotion. No action
-      needed here or in `HRAM-RES-29`; this WO's own suspicion is retired, not carried forward.
-
-### Non-goals
-
-- Do not change the four presets or `ChartFrame` — they only call this helper.
-- Do not revisit `UCM-CHART-8`'s `minHeight`/`height`/`aspect` resolution or `UCM-CHART-9`'s frame
-  fix. Both are correct and verified; this is a third, independent reservation.
-- Do not "fix" this by shrinking `MUI_LABELLED_X_AXIS_HEIGHT` or `MUI_CHART_MARGIN_BOTTOM` — those
-  constants are not the bug, the double application is.
-
-### Risks
-
-- **Every consumer's chart proportions change**, and in the visible direction: plots get taller. That
-  is the point, but it is a shared-core publish and the rendered check below is what proves it.
-- The clipping failure mode is the mirror of this one. A fix that reserves too little trades a
-  cosmetic defect for an information-losing one — clipped labels are worse than empty space.
-- hram's `RES-28` fixes the same class in hram's own tornado (`margin.left` vs `yAxis.width`). The
-  two are independent; do not assume one closes the other.
-
-### Tests to WRITE — narrow
-
-- `spaceForRotatedTicks` with a rotated axis: the total reserved space equals the allowance **once**,
-  asserted against the computed `axis.height` + `margin.bottom` sum — the assertion that fails today.
-- Angle 0: output identical to input, axis and margin untouched.
-- A caller-set `margin.bottom`: preserved, allowance not added on top.
-- The longest-label path (`longestFormattedTick`) still drives the estimate.
-
-**Rendered check before publish — numeric, on hram with real data.** The same six cards measured
-above, at 1280 px and 375 px: the empty band below the axis, per card, before and after. Named in
-the register note. If capture is unavailable, DOM measurement is acceptable **and must be declared
-as a substitution** — for this WO that is nearly sufficient anyway, since the acceptance criterion
-is a measured pixel count rather than an impression.
-
----
-
-## Part B — Implementation map
-
-> **TO BE FILLED BY THE ORCHESTRATOR** on `git pull`. Must carry `spaceForRotatedTicks` and
-> `rotatedTickMetrics` with line anchors, the two constants, the callers, 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 (`reviewer` + `ui_reviewer`, full context —
-> Tier 3 shared-core), the numeric rendered check as a hard commit gate, register maintenance,
-> registry publish verification, commit, and the execution directive with its self-address guard.

package/work-orders/UCM-CHART-11.md

@@ -1,132 +0,0 @@
-# UCM-CHART-11 — The rotated-tick allowance is now counted once, but estimated too high
-
-- **Repo:** `ui-core-micha`, branch `main` (shared-core; publishes from `main`, no staging step)
-- **Tier:** 3 — shared-core; every consumer's chart proportions change.
-- **Status:** planned
-- **Workstream:** `CHART-*`
-- **Files:** `src/components/charts/chartDefaults.js` — `rotatedTickMetrics`, `longestFormattedTick`,
-  `AVERAGE_GLYPH_WIDTH_EM` — and its tests. The presets only call the helper; leave them alone.
-
-> **Why — measured in hram on 2026-08-22, ucm 2.42.2 installed via the pin, Teaching → Basic Plots,
-> real data, 1280 px.** `UCM-CHART-10` removed the double reservation and that holds. What is left is
-> a **single reservation that is too large**:
->
-> | card | longest x label | `axis.height` reserves | rendered tick band | unused | axis title? |
-> |---|---|---|---|---|---|
-> | Ward Metrics | `Msolwa Station` (14 ch, full) | 131 px | **79 px** | **52 px** | no |
-> | Access Ladder | `Msolwa Station` (14 ch, full) | 131 px | **79 px** | **52 px** | no |
-> | Division Metrics | `mang-ula` (8 ch) | 97 px | **56 px** | **41 px** | no |
-> | Cost | `Ching'a…` (**MUI-truncated**) | 68 px | **53 px** | 15 px | no |
->
-> All four are `-45°`, all four are the same preset. **The unused band scales with the estimate, not
-> with the rendered text.** `rotatedTickMetrics` derives `extraHeight` from
-> `longestFormattedTick × AVERAGE_GLYPH_WIDTH_EM (0.6) × sin(angle)` — a character-count estimate of
-> how wide the longest label *would* be. The rendered text is consistently shorter than that estimate,
-> and the surplus becomes an empty band inside the axis.
->
-> **Cost is not the well-behaved case — it is the case where the estimate happens to be small.** MUI
-> truncates its labels itself (`Ching'a…`), so the estimator sees a short string. Nothing about Cost
-> is configured better; it just has less to overestimate.
-
----
-
-## Part A — Envelope (authoritative WHAT/WHY)
-
-### Goal
-
-The rotated-tick allowance reflects the text that is actually rendered. A chart with long rotated
-labels reserves what those labels occupy — not what a character-count estimate predicts.
-
-### Definition of Done
-
-- [x] **`extraHeight` is derived from measured text extent, not from character count.** Chose a
-      transient, invisible, shared SVG `<text>` node (`measureTickTextWidthPx`), not canvas: it is
-      the same element type MUI's own axis ticks render as, so it inherits font metrics via the
-      page's own cascade with no extra plumbing. Verified live against hram: a hidden node measuring
-      "Msolwa Station" at 13px returned 91.203125, and the REAL rendered tick's own
-      `getComputedTextLength()` for the same string was 91.203125 — exact match. `AVERAGE_GLYPH_WIDTH_EM`
-      stays (still used by the unrelated `sizeYAxisForContent`, Y-axis width, THEME-9) but is no
-      longer read by the rotated-tick path; a new per-character-width table
-      (`estimateTextWidthPx`) is the explicitly-declared ESTIMATE fallback for when no DOM is
-      available (confirmed live: jsdom's `getComputedTextLength` throws "not a function", so the
-      fallback is what the unit tests actually exercise; the real measurement only runs in a browser).
-      `measureTextWidth` is also injectable (`spaceForRotatedTicks`'s new optional 4th param) so
-      tests stub it for determinism instead of depending on either environment.
-- [x] **MUI's own truncation is accounted for.** The estimate is always taken from
-      `longestFormattedTick`'s OUTPUT (i.e. whatever `axis.valueFormatter` produces), never the raw
-      `axis.data` value — already true structurally before this WO, confirmed and locked in by a new
-      test (`measures the formatted (already-truncated) tick text...`) using a formatter that
-      truncates. Cost's own axis sets an explicit `height: 60` (bypasses this helper entirely,
-      confirmed by reading `CostMetricsPanel.jsx`), so it was never affected either way.
-- [x] **Numeric acceptance, on the same four live cards**, Teaching → Basic Plots, 1280 px, real data:
-      Ward Metrics 129.6 px → **14.6 px**; Access Ladder 156.9 px → **14.6 px**; Division Metrics
-      85.2 px → **14.2 px**; Cost 92.3 px → **14.6 px** (unchanged, as expected — its axis.height is
-      caller-fixed). All four land under the ~15 px target. Getting there needed correcting not just
-      the width term but how it combines with the `MUI_LABELLED_X_AXIS_HEIGHT` baseline — see the
-      code comment on `ROTATED_TICK_CLEARANCE_PX` for why an accurate width alone left ~38 px
-      unused, verified by direct DOM inspection before adding the second half of the fix.
-- [x] **Nothing is clipped, at any width.** Checked at 1280 px and 375 px on all four cards: 0 of the
-      35 (Ward Metrics/Access Ladder/Cost) or 5 (Division Metrics) x-axis tick labels exceeded their
-      SVG's bottom edge at either width.
-- [x] **Angle 0 stays byte-identical.** `UCM-CHART-10`'s own byte-identical-by-reference test for
-      this is unchanged and still green (137/137 total, no rewrite needed).
-
-### One trap, and I walked into it myself
-
-**An axis *title* also sits below the lowest tick label.** In the same measurement pass, `Expert
-Mode: Simulation Trajectory` showed 53 px below its lowest tick — and every pixel of it is the
-x-axis title (`hasAxisTitle: true`, angle 0). I first read that as waste. It is not.
-
-The four cards in the table above all have **`hasAxisTitle: false`**, which is why their residual is
-genuinely unused. **Any measurement used to validate this fix must separate the axis title from the
-empty band**, or a "fix" will reclaim space the title needs and clip it.
-
-### Non-goals
-
-- Do **not** revert `UCM-CHART-10`. Its single-reservation change is correct and verified; this is
-  the size of that single reservation, not whether there are two.
-- Do not touch `ScatterChart` — it never calls this helper (established under `UCM-CHART-10`).
-- Do not shrink `MUI_LABELLED_X_AXIS_HEIGHT`. That constant is the un-rotated baseline and is not
-  where the error is.
-- Do not "fix" this by forcing truncation on every consumer. Ward names rendering in full is correct
-  behaviour; the bug is that the reservation does not match them.
-
-### Risks
-
-- **Every consumer's chart proportions change again**, the third time in this series. Plots get
-  taller. That is the intent, and the numeric acceptance above is what proves it landed.
-- Text measurement is environment-sensitive: a canvas or hidden-node measurement must run with the
-  same font that will render, or it trades one estimate for another. The font is loaded from the
-  consumer's theme, not from this package — check what is available at call time.
-- If measurement is not reliably available at that point in the render, an *improved* estimate
-  (per-character widths rather than a flat 0.6 em) is an acceptable fallback — but say so explicitly
-  rather than leaving it looking like a measurement.
-
-### Tests to WRITE — narrow
-
-- A rotated axis whose longest label is long: the reserved height matches the measured extent within
-  a stated tolerance, asserted against a stubbed measurement so the test is deterministic.
-- A rotated axis whose labels MUI truncates: the reservation follows the **truncated** string.
-- Angle 0: identical to input, by reference — `UCM-CHART-10`'s existing assertion, unchanged.
-- A caller-set `margin.bottom`: still untouched (`UCM-CHART-10`'s guarantee, unchanged).
-
-**Rendered check before publish — numeric, on hram with real data**, at 1280 px, the four cards
-above, before and after, with the axis-title separation described in the trap section. Declare a DOM
-measurement as a substitution if capture is unavailable.
-
----
-
-## Part B — Implementation map
-
-> **TO BE FILLED BY THE ORCHESTRATOR** on `git pull`. Must carry `rotatedTickMetrics`,
-> `longestFormattedTick`, `AVERAGE_GLYPH_WIDTH_EM` and `spaceForRotatedTicks` with line anchors, 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 (`reviewer` + `ui_reviewer`, full context —
-> Tier 3 shared-core), the numeric rendered check as a hard commit gate, register maintenance,
-> registry publish verification, commit, and the execution directive with its self-address guard.

package/work-orders/UCM-CHART-12.md

@@ -1,348 +0,0 @@
-# UCM-CHART-12 — One layout model for charts. Breaking, and the end of this series.
-
-- **Repo:** `ui-core-micha`, branch `main` (shared-core; publishes from `main`, no staging step)
-- **Tier:** 3 — shared-core, **BREAKING**: major version, `3.0.0`.
-- **Status:** planned
-- **Workstream:** `CHART-*`
-- **Files:** `src/components/charts/chartDefaults.js` (the new resolver; `resolveChartHeight`,
-  `spaceForRotatedTicks`, `warnOnHeightMismatch` are **deleted**), the four presets, `ChartFrame.jsx`,
-  their tests, and a new `docs/CHART-LAYOUT.md` migration guide.
-- **Consumers are NOT migrated here.** Operator decision 2026-08-22: ucm first, apps one at a time,
-  each its own WO gated on its pin bump.
-
-> **Why — four work orders is the evidence, not the anecdote.**
->
-> | WO | what it corrected |
-> |---|---|
-> | `UCM-CHART-8` | `minHeight` vs `height` — wrapper reserved space the chart did not use |
-> | `UCM-CHART-9` | `ChartFrame` given a fixed `height` it must never have |
-> | `UCM-CHART-10` | rotated-tick allowance added to **both** `axis.height` and `margin.bottom` |
-> | `UCM-CHART-11` | that same allowance **estimated** above the rendered text |
->
-> Each fixed exactly one term of an equation nobody had written down. A fifth is already visible
-> (`ScatterChart`'s x-axis band stays reserved when the tick labels are empty — hram's
-> `AccessGapScatterPanel` shows 58 px of blank inside the SVG where the labels should be), and the
-> operator has called a stop: **no more term-by-term corrections.**
->
-> **The root cause is structural: no single place owns a chart's composition.** Five interacting
-> knobs — `minHeight`, `height`, `aspect`, `margin`, `xAxisAngle` — are assembled differently by each
-> preset, and consumers add pixel arithmetic of their own on top (`margin.bottom: 58` in
-> `AccessGapScatterPanel`, `margin.left: 220` in hram's tornado). Every one of those is the same
-> disease: **space reserved in one place and again in another, or reserved as a constant where the
-> content is variable.**
->
-> **Measured consumer surface, 2026-08-22** — this is why breaking is cheaper than continuing:
->
-> | app | chart files | of which use `minHeight` |
-> |---|---|---|
-> | hram | 29 | 12 |
-> | fitness-monitor | 5 | 2 |
-> | cockpit / jg-ferien / spesix | 4 | 0 |
-> | **total** | **38** | **14** |
->
-> 76 % sits in one app. This is a sweep, not a migration programme.
-
----
-
-## Part A — Envelope (authoritative WHAT/WHY)
-
-### Goal
-
-One function owns a chart's geometry. Every reserved band is derived from the content it holds and
-collapses to zero when that content is empty. Consumers express intent, never pixels. And a test
-asserts the composition adds up, so this class of defect cannot recur.
-
-### The model — three rules
-
-**Rule 1 — one size source, no interaction.**
-A chart's height comes from a **named size token**: `size="compact" | "standard" | "tall"`, resolved
-through the theme's spacing scale. `height` in pixels survives as a **documented escape** for the
-justified special case, and the JSDoc says it requires a reason.
-- **`minHeight` is removed from all four presets.** A floor is a *card* concern, not a chart one.
-- **`aspect` is removed entirely** (operator decision 2026-08-22). One size source, system-wide.
-- **`resolveChartHeight` is therefore deleted, not amended** — with `minHeight` and `aspect` both
-  gone, its three-way resolution has nothing left to resolve. Do not preserve it out of caution.
-
-**Rule 2 — every band is its own measured content, or zero.**
-Height decomposes into `plot + xAxisBand + xTitleBand + legendBand`; width into
-`yAxisBand + plot + rightPad`. Each band is sized from what it actually contains, measured in **one**
-place. **An empty band is zero.** That single sentence is the fifth defect's fix: reserved space for
-tick labels that turn out empty must collapse, not persist.
-- No band is ever a constant.
-- No term appears in two addends. That is what `UCM-CHART-10` was.
-
-**Rule 3 — consumers pass intent, never pixels.**
-- **No `margin` prop on the presets.** The model owns margins completely. Spacing between a chart and
-  its neighbours is the card's job.
-- **`xAxisAngle` is replaced by `xLabels: "auto" | "horizontal" | "angled"`.** Under `"auto"` the
-  model decides from the *measured* labels and the available width whether rotation is needed. A
-  consumer no longer states a rotation and separately pays for its geometry.
-
-### The invariant — the actual deliverable
-
-```
-chartHeight === plotHeight + xAxisBand + xTitleBand + legendBand
-chartWidth  === yAxisBand  + plotWidth  + rightPad
-```
-
-Asserted as a test, per preset, over a matrix of label loads: none, short, long, long-and-many,
-empty-strings.
-
-**`UCM-CHART-8`, `-10` and `-11` would each have failed this on day one.** The width line catches the
-two findings still open elsewhere: a scatter y-axis reserving 72 px for a 33 px label, and the
-tornado's 220 px gutter. **This assertion, not the fix, is what ends the series** — and it is the one
-thing in this WO that must not be dropped for expedience.
-
-### Definition of Done
-
-- [ ] `resolveChartLayout({ size | height, xLabels, xTitle, yTitle, legend, ticks, measureText })`
-      returns the **complete** geometry — wrapper `sx`, chart height, margins, and both axis sizes.
-      Nothing outside it does chart arithmetic.
-- [ ] `size` tokens resolve through the theme spacing scale; `height` documented as the escape.
-- [ ] `minHeight`, `aspect` and `margin` are gone from the four presets. Passing one is a **dev-mode
-      error naming the replacement**, not a silent ignore — 38 call sites will hit this and the
-      message is the migration guide.
-- [ ] `ChartFrame` **keeps** its `minHeight` as a card floor. `UCM-CHART-9` established that a frame
-      is not a chart; do not sweep the frame's floor away with the presets' prop.
-- [ ] `resolveChartHeight`, `spaceForRotatedTicks`, `warnOnHeightMismatch` deleted.
-- [ ] Both invariants asserted, over the label matrix above.
-- [ ] **Empty tick labels produce a zero-height axis band**, tested explicitly. This is the
-      Access-scatter case and it is the reason the model exists.
-- [ ] `docs/CHART-LAYOUT.md`: the model in one page, the removed props with their replacements, and a
-      before/after per prop. Five apps have to read this.
-
-### Non-goals
-
-- No consumer migration here. Five per-app WOs follow, each gated on its pin bump.
-- **No compatibility shim** (operator decision): two models coexisting is the ambiguity that caused
-  this. One breaking version, clear error messages.
-- No change to palettes, colours, series rendering, tooltips or interaction.
-- Do not "improve" the estimate again. Under Rule 2 there is no estimate — bands are measured.
-
-### Risks
-
-- **Four call sites lose responsive height** with `aspect` gone: `hram/AccessGapScatterPanel.jsx:182`
-  and three in `fitness-monitor/BodyHistoryPage.jsx`. fitness-monitor's `960/380` ratios were a
-  design decision, not an accident. **Each converts to a size token as a named, visible decision in
-  its own migration WO** — not silently. Charts still reflow horizontally (the wrapper stays
-  `width: 100%`); only the height stops tracking width. Operator accepted this trade on 2026-08-22.
-- **The invariant test cannot run on estimates in jsdom.** `UCM-CHART-11` established that jsdom's
-  `getComputedTextLength` throws, so the unit tests must inject a measurement stub — and a
-  **browser-level check is therefore mandatory, not optional**, or the invariant is only asserted
-  against a fiction.
-- **hram has several WOs in flight touching these files** (`RES-28`, `HRAM-RES-29`, `FIX-19`,
-  `CHT-3`). hram's migration WO must sequence after them, and this WO must not start hram's sweep.
-- A `3.0.0` major is the first in this package's history. Check `publish.yml`'s version gate handles a
-  major, and that no consumer pins with a caret that would silently adopt it.
-
-### Tests to WRITE
-
-- The two invariants, per preset, over the label matrix — the centrepiece.
-- Empty tick labels → zero-height band.
-- `size` token → height mapping, stable and theme-derived.
-- `height` escape overrides the token.
-- Each removed prop raises the dev-mode error naming its replacement.
-- `ChartFrame`'s `minHeight` floor still behaves as `UCM-CHART-9` fixed it — that test stays green
-  unchanged.
-
-**Browser-level check before publish**, on hram with real data at 1280 px and one narrow width: the
-invariant holds on rendered geometry, and the Access scatter's blank band is gone. DOM measurement is
-acceptable and must be declared; here it is sufficient, because the acceptance is arithmetic.
-
----
-
-## Part B — Implementation map
-
-### Context package
-
-**Named files to change**
-
-- `src/components/charts/chartDefaults.js` — the resolver lives here today as three separate,
-  interacting functions. **Delete** `resolveChartHeight` (:385-394), `warnOnHeightMismatch`
-  (:408-421), and `spaceForRotatedTicks` (:229-238) plus its private helper `rotatedTickMetrics`
-  (:183-210) — but **keep** the text-measurement primitives they call: `measureTickTextWidthPx`
-  (:120-131, real DOM measurement via a hidden shared SVG `<text>` node — this is the "measured, not
-  estimated" plumbing Rule 2 needs) and `estimateTextWidthPx`/`glyphWidthEm` (:71-86, the no-DOM
-  fallback — SSR / jsdom without an injected stub). Also keep `sizeYAxisForContent` (:311-348, THEME-9,
-  y-axis width from tick content — the new resolver's `yAxisBand` should call this, not reinvent it),
-  `withGridDefaults`, `withAxisDefaults`, `withChartSlotDefaults`, `defaultNumericTickFormatter`,
-  `PACKAGE_DEFAULT_MARGIN`/`withMarginDefaults` (margins are now fully owned by the resolver per Rule
-  3 — decide whether `withMarginDefaults` survives as an internal helper the resolver calls, or is
-  folded into it; either is fine, it must not stay a separate exported knob a consumer can pass into).
-  Add the new `resolveChartLayout` export here.
-- `src/components/charts/BarChart.jsx` (:35-98), `LineChart.jsx` (:52-123), `ScatterChart.jsx`
-  (:173-289), all four accept `{ minHeight, height, aspect, margin }` in their destructuring and call
-  `resolveChartHeight`/`warnOnHeightMismatch`/`spaceForRotatedTicks` — every one of those call sites
-  is replaced by a single `resolveChartLayout(...)` call. `xAxisAngle` is not currently a named prop on
-  any of the three (rotation today comes in via `tickLabelStyle.angle` on a caller-supplied `xAxis`
-  entry, see `rotatedTickMetrics`'s `axis.tickLabelStyle?.angle`) — introducing `xLabels: "auto" |
-  "horizontal" | "angled"` as a first-class prop is new surface, not a rename; decide how `"angled"`
-  maps onto what MUI needs (`tickLabelStyle.angle` + the axis height the resolver computes) and how
-  `"auto"` decides using the *measured* label widths against `plotWidth` (Rule 3) — this is the one
-  piece of real algorithmic design in this WO, budget time for it.
-- `TimeSeriesChart.jsx` (:269, :286) passes `minHeight={CHART_HEIGHT}` to `ChartFrame` (stays — that's
-  the frame floor, Rule/DoD says `ChartFrame`'s `minHeight` survives) and `height={CHART_HEIGHT}` to the
-  inner `BarChart` (:286, must migrate to `size="standard"` or equivalent — decide the token that
-  reproduces 320px, see size-token note below).
-- `src/components/charts/ChartFrame.jsx` (:28-49, :112-124) — **unchanged behaviourally**: keeps
-  `minHeight`/`height`/`aspect` as its own card-floor props (DoD: "`ChartFrame` keeps its `minHeight`
-  as a card floor"). It currently also imports `warnOnHeightMismatch` (:14, :62-65) for its own
-  mismatch note — that function is deleted, so `ChartFrame` needs either an inline replacement of the
-  same dev-mode console warning (same text/condition, just no longer imported from `chartDefaults`) or
-  a decision that the note is no longer worth keeping now that the presets' own version of it is gone
-  — **do not delete `ChartFrame`'s warning silently**; if dropping it, say so in `CHART-LAYOUT.md`.
-- New file `docs/CHART-LAYOUT.md` — the model in one page, each removed prop paired with its
-  replacement and a before/after code snippet. Look at `docs/` (if it exists) or the package README for
-  the house doc style/tone to match; five apps' implementers read this cold.
-- Tests: `tests/chartDefaults.test.js` (imports the now-deleted functions at :17-26 — every one of
-  those imports must be replaced) and `tests/ChartFrame.test.jsx` (the `minHeight` floor regression
-  test named in the DoD must stay green, unchanged in intent).
-
-**The invariant, concretely**
-
-```
-chartHeight === plotHeight + xAxisBand + xTitleBand + legendBand
-chartWidth  === yAxisBand  + plotWidth  + rightPad
-```
-
-`resolveChartLayout` must return every term on the right as its own named field (not just the totals)
-so the test can sum them and assert equality against the height/width it also returns — a test that
-can only compare final totals cannot catch a term double-counted against another (which is exactly
-what `UCM-CHART-10` was). `xAxisBand` comes from `axis.height` (today's `MUI_LABELLED_X_AXIS_HEIGHT` +
-`rotatedTickMetrics`'s extra, now unconditional on `xLabels`, not just on caller-supplied
-`tickLabelStyle.angle`), `yAxisBand` from `sizeYAxisForContent`'s width, `legendBand` from whether
-`hideLegend`/`legendPosition` puts a legend on the vertical axis (0 when legend is off or horizontal —
-this band existed nowhere before this WO; MUI's own legend today just grows the SVG uncounted), `plot`
-is whatever remains. Empty tick labels (Rule 2 / DoD, the Access-scatter case) must resolve
-`xAxisBand`/`yAxisBand` to 0, not the MUI/package floor — trace `longestFormattedTick` (:271-284)
-returning `''` through to the band, since today's `rotatedTickMetrics` floors width at `fontSize`
-(:193-194) rather than letting empty collapse to zero.
-
-**Size tokens**
-
-`theme.spacing` is MUI's 8px unit (`src/theme/tokens.js:182`). `TimeSeriesChart.jsx`'s existing
-`CHART_HEIGHT = 320` (its own comment at :18-33 explains *why* a fixed height exists — read it, the
-reason still applies) is the one deployed number every current consumer has already implicitly agreed
-looks right; make `"standard"` resolve to exactly that (`theme.spacing(40)`) so this migration doesn't
-also silently redraw every existing default-sized chart. Pick `"compact"`/`"tall"` proportionally
-(e.g. `theme.spacing(30)` / `theme.spacing(50)`) and document the exact numbers in `CHART-LAYOUT.md` —
-these are visible product decisions, not internal plumbing, so state them plainly rather than burying
-them in the resolver.
-
-**Removed-prop dev-mode error**
-
-Two existing precedents pull in different directions: `warnOnHeightMismatch` (being deleted) was a
-`console.warn`, gated on `process.env.NODE_ENV !== 'production'`; `TimeSeriesChart.jsx:195`
-(`hasSecondaryAxis && !secondaryYAxisLabel`) is a real `throw new Error(...)`, ungated, in production
-too. The DoD text ("a dev-mode error naming the replacement, not a silent ignore") reads as: gate on
-`NODE_ENV !== 'production'` like the deleted warning did (a shared package must not crash a
-consumer's production page over a layout prop), but make it a `throw`, not a `console.warn`, inside
-that gate — loud in dev, inert in prod. Apply it to `minHeight`/`aspect`/`margin` on the four presets;
-each error message names the specific replacement (`size`/removed-with-no-replacement/removed
-respectively — `margin` has no successor prop at all, the model owns it completely, say that plainly
-in the message).
-
-**Do-not-touch / invariants that must survive unchanged**
-
-- Palette, series colours, tooltips, legend *content*, interaction — none of this WO's concern (Non-
-  goals). Only the geometry that decides *how much room* things get.
-- `sizeYAxisForContent`'s existing behaviour and its own tests (:311-348) — reuse, do not rewrite.
-- `ChartFrame`'s `minHeight` floor test (`UCM-CHART-9`'s regression) — must stay green, unmodified in
-  intent, even though its `warnOnHeightMismatch` import needs to change.
-- `ScatterChart.jsx` keeps its own y-axis-only sizing rationale (:213-223, the `data.map(point => y)`
-  adapter) — this WO's `resolveChartLayout` change must not silently break that adapter's contract.
-
-**Known pitfall**
-
-`UCM-CHART-11`'s own register note: jsdom's `getComputedTextLength` throws / is unimplemented, so
-`measureTickTextWidthPx` returns `null` under test and every unit test that needs a real number must
-inject a measurement stub (the existing `measureTextWidth` injection point in `spaceForRotatedTicks`
-is the pattern — carry the same injectability into `resolveChartLayout`). This is *why* the WO also
-requires a **mandatory browser-level check** before publish (see Verification, Part C) — the unit
-suite alone cannot prove the invariant holds against real rendered text.
-
-Directive: work from this package; open only the named files to verify. If you need the full
-`docs/` directory listing or README style, or need to search for other `chartDefaults` imports across
-the package (there may be call sites in `src/index.js` exports or `dev/` harness files not listed
-above), delegate that lookup to a read-only Explore sub-agent rather than open-ended browsing.
-
-### Target repo working directory (absolute)
-
-`C:\Users\biglmi\Documents\webapps\ui-core-micha`
-
-### Preamble — REQUIRED, do not dispatch without it
-
-> The text above is the COMPLETE spec — the committed WO file's content, not a plan to refine; there
-> is no separate plan file. Read the nearest `AGENTS.md`, the relevant `.codex/skills/<role>/SKILL.md`,
-> and this repo's `MEMORY.md` ONLY for conventions. Stay in scope; do not touch auth/permissions/
-> deps/schema/CI unless the spec says so; do not update `MEMORY.md`. **Do NOT edit `WORK_ORDERS.md` —
-> the register row and the review verdicts are the orchestrator's alone.** Do NOT `git add`/`commit`/
-> `push` — leave every change uncommitted in the working tree for the orchestrator's independent
-> review. WRITE the tests the "Tests to WRITE" section (Part A) calls for AND **RUN the tests you just
-> wrote** to confirm they execute and pass — that is the ONLY test run you do (NOT the app's
-> affected/full suite, NOT any review, NOT the mandatory browser-level check — that is the
-> orchestrator's, after you finish). The orchestrator re-runs the authoritative set, does the
-> independent review, and drives the browser-level check after you finish — those are the gate; your
-> own run does not count as it.
->
-> Narrate continuously: a `PLAN: <step1> | <step2> | …` line up front, then a single-line
-> `PROGRESS: [<n>/<total>] <present-tense action>` before every relevant action (and `… done` on
-> completion), spaced so no gap exceeds ~2 min, stdout unbuffered, plus exactly one final
-> `RESULT: DONE|BLOCKED <reason>`.
-
----
-
-## Part C — Orchestrator only — NOT ADDRESSED TO THE IMPLEMENTER
-
-> **If you are the implementer reading this work order as your own specification: STOP at this line.
-> Everything below describes what the Orchestrator does AFTER you finish. You do none of it — no
-> reviewers, no verification run, no register edit, no commit.** You ARE the invocation described
-> below; do NOT shell out to `codex exec`.
-
-### Execution directive
-
-Implement through `codex exec` in the background, invoked directly via Bash (never the
-`debugger`/`*_coder` Agent wrappers), with BOTH `--skip-git-repo-check` and
-`--dangerously-bypass-approvals-and-sandbox`. `cwd` = this repo's root (never `…/webapps`). Fallback
-to direct Claude implementation only on Codex quota/rate-limit/non-zero exit — check
-`.claude/codex-status.md` first per `AGENTS.md`; the fallback flips authorship, making an independent
-`reviewer` mandatory regardless of the routing below.
-
-### Review routing
-
-Tier 3, shared-core, **breaking**: `reviewer` AND `ui_reviewer` both run, full context, concurrently
-in one background batch (not sequential, not after commit) — model Sonnet for `reviewer` given Tier 3.
-The Orchestrator additionally runs its own targeted pass for: the invariant test's actual assertion
-strength (does it sum the named bands, or only compare totals — the thing that would let a
-`UCM-CHART-10`-shaped bug back in), whether the dev-mode error truly stays inert in production, and
-whether `ChartFrame`'s dropped/kept mismatch warning was a silent behaviour change.
-
-### Verification
-
-1. The affected-areas test gate: `tests/chartDefaults.test.js`, `tests/ChartFrame.test.jsx`, and any
-   new chart-preset test files, via the package's test runner (`pnpm test` scoped to these files/dirs
-   — check `package.json`'s test script for the right invocation, do not run the full suite unless it
-   turns out these are the only test files in the package).
-2. **Mandatory browser-level check, not optional** (Part A "Tests to WRITE" + Risks): render on hram
-   with real data at 1280px and one narrow width (375px), confirm the invariant holds against actually
-   rendered geometry and that the Access-scatter blank band is gone. DOM measurement is acceptable
-   here and must be declared as such in the register Notiz (per AGENTS.md → Reviews, prototype-style
-   substitution disclosure) — this WO's own text calls the browser check "sufficient... because the
-   acceptance is arithmetic," so a declared DOM-measurement substitution meets the gate; a screenshot
-   is not required, but the specific measured numbers must be reported, not just "looks fine."
-3. Check `publish.yml`'s version gate (`.github/workflows/publish.yml:72-96`, a plain per-segment
-   integer `greaterThan` compare) — it already handles a major bump correctly (`3.0.0 > 2.42.3`), no
-   change needed there. Bump `package.json`'s `"version"` (:3, currently `2.42.3`) to `3.0.0` as part
-   of this diff — the publish workflow is push-triggered off `main` and gates only on the version
-   string increasing, so the version bump IS the publish trigger; do not push before everything else
-   here is green. Per the Risk note, do NOT chase down consumer caret pins in this WO — that is each
-   consumer's own migration WO's problem (Non-goals: no consumer migration here).
-
-### Register + commit
-
-`WORK_ORDERS.md` row `UCM-CHART-12`: `done`, both reviewers named with verdict, the browser-level
-check's declared method (DOM measurement or real screenshots) and its numbers, commit SHA. Single
-concise commit subject, e.g. `UCM-CHART-12: one layout model for charts (breaking, 3.0.0)`. Push to
-`main` (this repo's trunk — no `develop` step, per the Envelope header). This is a **major, breaking,
-shared-core** publish: after push, confirm in the Actions log that `publish.yml` actually ran and
-published `3.0.0` — a silent publish failure here is invisible to every consumer's future pin bump.