Current coverage

Essenfont covers 137,117 of the 159,866 codepoints assigned by Unicode 17.0.

  • 85.8% of assigned codepoints have real glyph outlines.
  • 92 of 346 blocks are fully covered (every assigned codepoint renders).
  • Private Use Areas are not assigned by Unicode and are excluded from coverage counts.

Known gaps

CFF outline extraction (Tangut and similar)

Status: Fixed in build repo (CffOutlineFiller, commit 7b0166f). Website WOFF2 subsets for Tangut and Tangut Components have been regenerated directly from the Noto Serif Tangut donor font.

Root cause: fontisan’s UFO converter (FromBinData.extract_cff_glyphs) is a stub in v0.4.23 — it creates UFO glyphs with names and advance widths but zero contours for CFF-based OTF donors. The CffOutlineFiller module bridges fontisan’s OutlineExtractor (which can parse CFF charstrings) into the UFO after the stub conversion.

Color emoji in per-block WOFF2

Status: Blocked on fontisan subsetter.

The OTC carries CBDT/CBLC color-bitmap tables for emoji blocks (Emoticons, Transport and Map Symbols, etc.). When the per-block WOFF2 subsetter processes these blocks, it drops the CBDT/CBLC tables because it doesn’t recognize them. The website serves these blocks in monochrome (Noto Emoji glyf outlines) instead of color.

Fix path: The fontisan subsetter needs a table-passthrough mode that preserves CBDT/CBLC tables alongside glyf outlines.

Chrome OTS requirements

Chrome’s OpenType Sanitizer (OTS) enforces several requirements that the build must meet:

  1. loca indexFormat must be 0 for short loca tables. Fixed in fontisan via LocaFormat integration.
  2. head.modified timestamp must be outside Chrome’s rejection window. Fixed by stamping the timestamp explicitly during build.
  3. WOFF2 must be spec-compliant — fontisan’s WOFF2 encoder had a bug in origLength fields for reconstructed short loca. Fixed via LocaFormat integration.

All three fixes are verified end-to-end via Puppeteer tests in the build repo.

Roadmap

Item Status Blocker
Tangut glyph outlines ✅ Fixed —
CFF2 charstring index corruption ⚠ Partial fontisan CFF2 compilation bug
Color emoji WOFF2 subsets ❌ Blocked fontisan subsetter CBDT/CBLC passthrough
UFO disk cache ❌ Blocked fontisan Ufo::Reader Phase 2 (contour decoding)
Per-codepoint SVG export cache ✅ Done BuildCache keyed by OTC sha256
WOFF2 encoding cache ✅ Done BuildCache keyed by TTF sha256

The “missing characters” taxonomy

When a codepoint doesn’t render on this site, there are three possible causes:

  1. The font has no glyph for this codepoint. The donor font’s cmap doesn’t include it. Tangut and similar CFF-based donors had this issue (now fixed).

  2. The face lookup picks an empty face. The TTC has 5 faces; if the code-to-face lookup resolves to the wrong face, the glyph is empty. Fixed via the face_lookup correct-face-first algorithm.

  3. The WOFF2 subset stripped a table the glyph needs. CBDT/CBLC tables for color emoji are stripped during per-block WOFF2 subsetting. Fixed for monochrome; blocked for color.

Each is diagnosed and tracked in the coverage data (public/site-stats.json → coverage).