pdfrum-cmap

A CMap does two jobs at once for a composite font (ISO 32000-1 §9.7.5): it splits a show-operator's string into character codes one, two, three or four bytes wide, and it maps each code to a CID. The split is the part a byte oriented reader gets wrong — code widths vary within one string, so a CMap has to be consulted to know where the next code even begins.

use pdfrum_cmap::from_encoding_name;
use pdfrum_common::Diagnostics;
use pdfrum_object::Name;

let cmap = from_encoding_name(&Name::from("Identity-H"), &mut Diagnostics::default());
assert!(!cmap.is_vertical());
let cids: Vec<_> = cmap.decode(&[0x00, 0x41, 0x00, 0x42]).collect();
assert_eq!(cids.len(), 2);

59 named tables over 32 decoder stems ship compiled in ([predefined]), plus [parse_embedded] for a CMap the file carries as a stream. [from_encoding_name] is the /Encoding entry point: an unknown name is a diagnostic and a one-byte identity CMap, never an error, because a font whose encoding cannot be resolved still has to draw something.

Inheritance runs through three live mechanisms and they are not interchangeable. The built-in tables chain among themselves, which is what makes GB-EUC-V a thin override of GB-EUC-H. usecmap inside an embedded program is resolved by [parse_embedded]. The /UseCMap key of an /Encoding stream's dictionary is attached by [inherit_from] and supersedes the operator. The last two are §9.7.5.3's two channels, and both are child-wins: a code the child maps is the child's answer, and only a code it maps to nothing reaches the parent. The oracle implements neither.

Part of pdfrum. #![forbid(unsafe_code)].

MIT OR Apache-2.0