CffSubsetter

Subsets a CFF font program (the outlines inside an OpenType .otf) down to the glyphs a document uses, and re-emits it as a CID-keyed CFF.

The output is always CID-keyed with a charset that maps each new glyph id to its original glyph id (used as the CID). Under a PDF Type0 / CIDFontType0 / Identity-H font that means the content stream can keep emitting the original glyph id as the 2-byte code (identical to the TrueType path) and the reader resolves code → CID(=original gid) → new gid via this charset. So /W and /ToUnicode stay keyed by the original gid too.

Only the CharStrings INDEX is subset; all global and local subroutines are kept verbatim (their biased references stay valid), as do all FontDicts. For a CJK CFF the charstrings dominate, so this is the bulk of the size win without the complexity (and risk) of renumbering subrs. The deprecated seac-via- endchar accent composition is not followed, so a glyph built that way would lose its components. CID-keyed fonts don't use it.