Struct EmbeddedFontPlan 

pub(crate) struct EmbeddedFontPlan {
    pub id: EmbeddedFontId,
    pub subset_bytes: Vec<u8>,
    pub remapper: GlyphRemapper,
    pub used_gids: Vec<u16>,
    pub gid_to_text: BTreeMap<u16, String>,
}

Per-face plan: which glyph IDs were used, the subset bytes, the GlyphRemapper that maps original → subset GIDs, and the gid-to-source-text map used to build /ToUnicode.

Fields

id: EmbeddedFontId

subset_bytes: Vec<u8>

remapper: GlyphRemapper

used_gids: Vec<u16>

Original GIDs used in content streams, including GID 0 (.notdef) when unsupported codepoints were shaped. This is wider than gid_to_text: .notdef needs a PDF width but no /ToUnicode mapping.

gid_to_text: BTreeMap<u16, String>

Original GID → source text for that glyph’s cluster. For ligatures (1 glyph, N codepoints) the value is the multi-char cluster string. For 1:1 mappings (typical LTR) it’s a single-char string. For one-codepoint-many-glyphs decompositions (rare), the first glyph carries the codepoint and later glyphs in the same cluster carry empty strings. Co-populated with used_gids by accumulate_glyphs; keep those sources of truth in sync.