Render semantic chord parts, Unicode symbols, and muted rhythm markers

[isaiahdahl]

Stack position

3 of 4

• Base PR: Normalize directives and preserve semantic chord-line tokens #2229
• Merge after: Normalize directives and preserve semantic chord-line tokens #2229
• Next: Refine contextual enharmonic spelling #2237

The effective diff contains chord shaping, Unicode/font work, configurable chord-part rendering, and muted-x styling. Directive and token-model changes remain in the base PR.

Review first

1. src/rendering/chord_shaper.ts
2. src/layout/engine/item_processor.ts
3. src/rendering/pdf/js_pdf_renderer.ts
4. src/rendering/html/chord_run_renderer.ts
5. src/chord_sheet/chord_line_token.ts for muted x
6. script/build_symbol_fonts.py and script/check-symbol-fonts.sh

Problem

PDF and measured HTML need the same understanding of a displayed chord. Roots, qualities, extensions, and basses can use different sizes, baselines, and fallback fonts, but measurement and rendering must still agree or line wrapping changes between passes.

Unicode accidentals also need glyphs and metrics that are available in generated PDFs. Relying on an environment font can change output between machines.

Changes

Semantic chord shaping

Displayed chords are split into semantic runs:

marker | root | quality | extensions | bass | marker

The same run model drives:

• line measurement and wrapping;
• PDF chord rendering;
• measured-HTML spans and classes;
• chord-diagram titles.

Public configuration

new PdfFormatter({
  useUnicodeModifiers: true,
  chordRendering: {
    quality: {
      fontSizeRatio: 0.9,
    },
    extensions: {
      fontSizeRatio: 0.85,
      baselineShiftRatio: 0.3,
    },
  },
});

Omitted values keep neutral rendering. The older chordSuperscript input remains readable for compatibility but is not part of the new configuration surface.

Unicode fonts

• Parse Unicode accidentals in roots, extensions, keys, and slash basses.
• Add a generated ChordSheet Symbols font with source attribution and drift checks.
• Install pinned uv in CI for deterministic font generation.
• Skip binary font sources in the trailing-whitespace text check.
• Use the same glyph metrics during measurement and PDF output.
• Preserve Unicode text extraction.

Muted rhythm symbols

• Keep x semantically classified as rhythm-symbol with variant mute.
• Resolve only its visual role to noChord.
• Keep / and - on the ordinary rhythm-symbol role.
• Apply the role consistently to measurement and positioned HTML.

Integration guarantees

• Current master numeric and lowercase Roman behavior remains intact.
• Responsive pagination from Normalize directives and preserve semantic chord-line tokens #2229 remains intact.
• The formatter and layout engine use the same effective column count.

Boundaries

This layer does not choose contextual enharmonic spellings. It renders the finalized Chord model supplied by the parser and song operations. That provenance work is in #2237.

Validation

• Full suite at this layer: 184 suites passed; 45,525 tests passed; 3 skipped.
• Type checks, ESLint, package exports, and symbol-font drift checks passed.
• Tests cover PDF, measured HTML, positioned HTML, chord-diagram titles, Unicode extraction, responsive pagination, and muted x semantics.