[botandbang]

What and why

Riding the pitch fader from +0% to +4% during a track gave one speed everywhere a mix is shown. The track stays in one chain, because each 4 s slice drifts well within a chain's 0.75 s tolerance of the last, and the chain's one line averaged both speeds. In a test the play read +0.0%: the averaged slope was 2% off the speed searched, so the matcher fell back to that speed. Its presence curve was traced at the average, which neither half played at, and it dipped to 0.18.

This takes option a from triage, which is a provisional choice that hasn't been confirmed: split a chain's fit where its speed changes. The other options were (b) report the range of speeds searched, quantised, and (c) keep one averaged speed and document it. Undoing it means removing lines/split/Bend in the matcher and the speed_range field.

Added in loclib-mixid:

• split tries two lines meeting at a knot, searched on a 0.5 s grid at least 24 s from either end. It keeps them when their speeds differ by 0.3% (RIDE) or more, each is within 2% of the speeds its slices voted for, and the Bayesian information criterion prefers them (k = 2 against 4). It recurses on each side. Strays are dropped relative to the bend, so the end of a short second speed isn't cut off.
• Found::speed_range() gives the slowest and fastest measured lines.
• BIC alone isn't enough: whole-frame rounding runs in sawtooth stretches and passes it. Rounding can tip two lines of 24 s or more apart by up to 0.29%, so the 0.3% check rules it out.

Changed in loclib-mixid:

• Found.fits has a line per speed.
• tempo is weighted by votes and line length, so a play at one speed is unchanged.
• trace listens near each line's own speed.
• least_squares, drop_strays, edges and chain_pairs are pulled out of fit to share with the new code.

Added in locallibrary and the website:

• Play.speed_range and the export's speedRange. Both are optional, so the export keeps version 2, and the RPC contract is regenerated.
• The range ("+0.0% to +4.0%") is shown in the HTML report (track list and both tooltips), the terminal tracklist, the player's Mixes view (tracklist column and details) and the website's mix report (tracklist and tempo chart).

Not included: the player's playhead and the tempo charts still follow one speed per play, since Play carries the range and not the lines.

Fixes LOC-122

How it was tested

• A matcher end-to-end test (45 s at +0%, then +4%) that gave one fit before the fix and now gives two lines meeting at 45 s, with presence at 0.4 or more throughout. Hegel properties that a play at one speed is never split and that a 1 to 4% step is found where it happened (150 cases each).
• A locallibrary end-to-end test through Identifier with WAVs, and a test per surface: the report page, the export (hegel), the tracklist, the TUI and the website's helper and schema.
• Broken once each, and failed: the per-line trace (presence back to 0.18), both split checks removed and RIDE removed alone, the export, the report page field, the TUI's fastest speed, and the website schema field.
• cargo fmt -- --check, and cargo clippy -p loclib-mixid -p locallibrary -p locallibrary-cli --all-targets -- -D warnings.
• cargo nextest run: loclib-mixid 75/75, locallibrary's mix, sets, mixes and sessions tests 106/106, and locallibrary-cli's mix, TUI mix, rpc and mix command tests 138/138 (the_contract_in_docs_is_current included).
• apps/web pnpm check, pnpm lint, oxfmt --check and pnpm test (800 pass).
• Not run: the full Rust suites, Windows clippy, web test:storybook and test:build, and screenshots.

Checklist

• Tests cover the change
• For locallibrary: apps/locallibrary/CHANGELOG.md has a line under Unreleased, if people using it will notice
• The READMEs, docs/ and --help still say what the code does (apps/locallibrary/docs/rpc/ too, for a sidecar change)

[linear-code]

LOC-122

[github-actions]

Coverage Report for apps/web

Status	Category	Percentage	Covered / Total
🟢	Lines	95.46% (🎯 93%) 🟰 ±0 pp	7750 / 8118
🟢	Statements	94.11% (🎯 91%) 🟰 ±0 pp	11857 / 12598
🟢	Functions	94.26% (🎯 92%) 🟰 ±0 pp	3075 / 3262
🟢	Branches	84.75% (🎯 80%) ⬆️ +0.01 pp	5541 / 6538

File Coverage

File	Stmts	Branches	Functions	Lines	Uncovered Lines
Changed Files
apps/web/src/lib/components/mix/MixTracklist.svelte	92.74% 🟰 ±0 pp	65.34% 🟰 ±0 pp	93.47% 🟰 ±0 pp	94.82% 🟰 ±0 pp	8, 58, 58, 78, 78, 51-53
apps/web/src/lib/components/mix/TempoChart.svelte	88.7% 🟰 ±0 pp	57.14% 🟰 ±0 pp	95.91% 🟰 ±0 pp	79.24% 🟰 ±0 pp	45-59, 70-101, 63
apps/web/src/lib/mix/report.ts	93.25% ⬆️ +0.07 pp	82.75% ⬆️ +0.61 pp	97.36% ⬆️ +0.07 pp	97.14% ⬆️ +0.04 pp	132-134, 179, 190
apps/web/src/lib/mix/schema.ts	100% 🟰 ±0 pp	92.59% 🟰 ±0 pp	100% 🟰 ±0 pp	100% 🟰 ±0 pp

Generated in workflow #653 for commit 61a662d by the Vitest Coverage Report Action

Size

Compared with main. KiB are 1,024 bytes.

	This pull request	main	Change
Client JavaScript	679.4 KiB	676.3 KiB	+3.2 KiB (+0.5%)
Client JavaScript, brotli	225.6 KiB	224.9 KiB	+813 B (+0.4%)
Client CSS	89.6 KiB	89.6 KiB	0 B (0%)
Client CSS, brotli	12.9 KiB	12.9 KiB	0 B (0%)

[botandbang]

Closing and reopening so CI reruns with the destinations test fix (#344) now on main. Branch unchanged.

[github-actions]

Rust coverage

From run 641. Both are held to the thresholds in .cargo/config.toml.

Tests	Lines	Functions	Regions
All	97.45% (83016/85189)	96.53% (13243/13719)	95.71% (182618/190804)
Hegel property tests alone	95.56% (81406/85189)	94.25% (12930/13719)	93.65% (178689/190804)