epiphany/crates/epiphany-render-svg
Levi Neuwirth c818992835 Push 3 Standard-tier: optimal break search (replace greedy first-fit)
Casting-off's greedy first-fit + tail-only widow rebalance is replaced by a
deterministic badness-minimizing break search (optimal_breaks, a Knuth-Plass-
style dynamic program over the measure boundaries). ENGRAVER_VERSION 9 -> 10.

Objective: minimize the sum over ALL systems of the squared normalized underfill
((width_limit - w)/width_limit)^2. Squaring evens the systems; including the
FINAL system in the sum subsumes the old widow rebalance (the optimizer won't
leave a narrow final stub if a balanced partition is cheaper). It is the
additive, DP-tractable analog of the retired distribution_cost (max of the
catalog's break penalty and width-CV imbalance). On the ten-measure fixture the
search settles on 5/4 measures where greedy left a fuller-then-shorter split,
filling the final system more and pulling casting_off_quality down (~0.80 ->
~0.61) — the payoff, visible now that horizontal justification drives
system_break to ~0 so casting_off mostly sees the last system's fullness.

Break requirements (hard/soft/page) bound the DP's segments — a system may not
span a forced break — and walk_region still honours them and records skipped
content-less soft breaks as IrOverride, unchanged; optimal_breaks reports only
the automatic breaks. A system may exceed the width only as a single
unsplittable measure. Minimal still makes no optimality claim.

Deterministic: minimizes lexicographic (cost, system_count). Tests:
optimal_breaks_{balances_systems_and_avoids_a_final_widow, never_spans_a_forced_
break, is_deterministic_and_empty_when_unbounded}; the widow test now checks the
balanced measure distribution; wrapping-fixture metrics updated (casting_off
improved). Removed rebalance_widows/rebalance_region/distribution_cost + their
tests. Goldens regenerated (balanced systems; view_box stable — justification
still fills to width). 944 tests, clippy 0, docs -D warnings, conformance 8/8.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-08 21:29:08 -04:00
..
examples Phase 3 tranche 1: casting-off, K1 schema-fill, value-restoring undo 2026-07-02 21:55:26 -04:00
src Push 3 (slur quality) 1/3: render dashed/dotted slurs faithfully 2026-07-08 17:13:32 -04:00
tests Push 3 Standard-tier: optimal break search (replace greedy first-fit) 2026-07-08 21:29:08 -04:00
tools Schema major 2 Phase E1: repeat barlines + volta brackets render 2026-07-07 21:51:11 -04:00
Cargo.toml Agent I-4c: embedded @font-face glyph mode (a second self-contained renderer) 2026-06-27 11:09:34 -04:00
DECISIONS.md Schema major 2 Phase E2: slur curves + cubic-bézier primitive 2026-07-08 15:33:00 -04:00
README.md Agent I-4c: embedded @font-face glyph mode (a second self-contained renderer) 2026-06-27 11:09:34 -04:00

README.md

epiphany-render-svg

Agent I's SVG renderer behind the Epiphany RenderIR interface (spec Chapter 7): turns a ResolvedLayoutIR into well-formed SVG 1.1, drawing each glyph from genuine Bravura SMuFL data — inline outline <path>s by default (GlyphMode::PathOutline), or <text> set in an @font-face-embedded Bravura subset (GlyphMode::EmbeddedFont). It is the visible end of the v0 Score → layout IR pipeline.

Status

The Score → layout IR → SVG pipeline renders recognizable notation — clefs, noteheads at clef-relative staff positions, accidentals, key/time signatures, rests, barlines, and the staff lines and stems that connect them. Output is golden-locked against both the interface-only stub solver and Agent I's real epiphany-engrave solver (whose horizontal spacing pass re-spaces the glyphs), and the layout round-trip (criterion 6) runs through both. What the renderer itself guarantees, independent of engraving quality: real Bravura glyphs, provenance preserved to the score graph, output XML-valid and deterministic. The renderer consumes any solver's ResolvedLayoutIR.

Demo

# Render a fixture to SVG (stub solver by default):
cargo run -p epiphany-render-svg --example render_fixture -- \
    ten_measure_single_staff > out.svg

# Drive Agent I's engrave solver instead, to bisect renderer-vs-solver:
cargo run -p epiphany-render-svg --example render_fixture -- \
    ten_measure_single_staff --solver=real > out.svg

# Use the embedded-font glyph mode (<text> + @font-face) instead of inline paths:
cargo run -p epiphany-render-svg --example render_fixture -- \
    ten_measure_single_staff --glyph-mode=embedded > out.svg

Fixtures: ten_measure_single_staff, valid_score_rich, valid_score. Stats and diagnostics go to stderr; the SVG goes to stdout.

Library

use epiphany_render_svg::{render, RenderOptions};

let out = render(&resolved_layout_ir, &RenderOptions::default());
assert!(out.is_well_formed());
println!("{}", out.svg);

render is pure and deterministic. RenderOptions controls SVG-encoding choices only (display scale, margin, provenance attributes, and glyph_mode — inline PathOutline vs EmbeddedFont) — nothing that changes engraving.

Bundled Bravura data

Two generated artifacts come from the official OFL Bravura.otf via tools/extract_bravura_outlines.py — the font is not vendored, only the generated Rust is committed:

  • src/outlines_generated.rs — the inline glyph outlines (geometry-only, so byte-stable across fontTools versions);
  • src/font_subset_generated.rs — a base64 OTF subset (just the pipeline's glyphs) for GlyphMode::EmbeddedFont. As a Modified Version, its primary font name is renamed off the Reserved Font Name "Bravura" per the OFL; a content BLAKE3 + decoded length are committed alongside as an integrity lock.

Bravura is © Steinberg Media Technologies GmbH under the SIL Open Font License 1.1 (tools/OFL.txt); both artifacts are redistributed under the same license. To regenerate both (the subset step also needs the blake3 package):

cd crates/epiphany-render-svg/tools
python3 -m venv .venv && . .venv/bin/activate && pip install fonttools blake3
python3 extract_bravura_outlines.py --font-out ../src/font_subset_generated.rs \
    > ../src/outlines_generated.rs