Module unicode_font 

Available on crate feature font-subset only.

Embedding a caller-supplied Unicode font, so a replacement can contain scripts the document itself never had. See text_edit::FontFallback::EmbedUnicode. Embedding a caller-supplied Unicode font, so a replacement can contain characters the document’s own fonts have no glyph for. Embedding a caller-supplied Unicode font as a Type0 composite font, so a replacement can contain characters the original font has no glyph for.

Why this exists

The Phase 1B replacement path re-encodes through the original font’s reverse map, and its fallback ([FontFallback::InjectStandard]) injects Helvetica/WinAnsi. Both are bounded by what the document already had: WinAnsi tops out at U+00FF. That is fine for replacing a word in an English document and useless for writing Polish, Greek, Cyrillic, CJK or any other script the source never contained.

A Type0 font with Identity-H encoding removes that ceiling: character codes are 2-byte glyph indices into an embedded font program, so the reachable set is whatever the supplied font covers.

Why the caller supplies the font

Embedding glyphs redistributes the font. Rather than bundle one and inherit its licence terms (and its size — a full CJK face is tens of megabytes), the caller passes the bytes of a font they are licensed to embed. That also lets them pick a face per language, which a bundled default could not do. A bundled default can be layered on later without changing this API.

What is written

Type0 (/Encoding /Identity-H)
 └── DescendantFonts[0] → CIDFontType2
       ├── /CIDToGIDMap /Identity     CID == GID, no extra indirection
       ├── /W                          per-glyph advances
       └── /FontDescriptor → /FontFile2 (subsetted font program)
 └── /ToUnicode                        CMap so the text stays extractable

The /ToUnicode map is not optional here. Without it a translated document renders correctly and yields nothing on copy, search or screen-reader — which for a translation workflow is a silent, total loss of the actual product.

Subsetting and glyph identity

Glyphs are remapped to a compact range via subsetter::GlyphRemapper, so the embedded program carries only what is used. Because Identity-H ties the character code to the glyph index, the codes written into the content stream are the new (post-remap) indices, never the original ones. Getting that pairing wrong produces a file that opens cleanly and renders the wrong glyphs, so the two always travel together in UnicodeEncoder.

Structs

UnicodeEncoder

Accumulates glyph usage for one page, then writes the font.

UnicodeFont

A caller-supplied font, validated as embeddable.

Enums

Outlines

Which outline format the font program carries.