hfs_unicode.hpp

Go to the documentation of this file.

// hfs_unicode.hpp -- UTF-16BE -> UTF-8 + partial NFD -> NFC normalisation.
//
// HFS+ stores filenames as UTF-16 BIG-ENDIAN code units in NFD
// (canonical decomposed) form. For example, "città" is stored as
//   c, i, t, t, à                 NO (that's NFC)
//   c, i, t, t, a, U+0300         YES (NFD: 'a' + COMBINING GRAVE)
//
// We need:
//   1. Convert each UCS-2 code unit (surrogates included) to a UTF-8
//      string the host filesystem can use.
//   2. Recompose common NFD sequences back to NFC. We CANNOT pull in
//      ICU or libutf8proc as a dependency, so we implement a finite
//      table of the most common compositions:
//        * Latin-1 letter + combining accent  -> precomposed code point
//        * (a few hundred entries cover practically all European names)
//      Anything outside that table stays in NFD, which is still valid
//      Unicode, just visually equivalent to NFC.
//
// IMPORTANT: this is INTENTIONALLY a partial implementation. Hangul,
// CJK ideographs, mathematical alphanumerics, and IPA all decompose in
// ways our small table does not cover. Their NFD form will go through
// unchanged. The host filesystem will display them correctly (modern
// terminals and file browsers handle NFD), so this is a cosmetic
// limitation, not a correctness issue.
 
#pragma once
 
#include <cstdint>
#include <string>
#include <vector>
 
namespace hfs {
 
// Decode N UTF-16-BE code units (2*N bytes) into a UTF-8 string.
// Surrogate pairs are handled. Lone surrogates are replaced with U+FFFD.
//
// `src` points to the first byte; `code_units` is the number of UTF-16
// code units (NOT bytes). For an HFSUniStr255 record, `code_units` is
// the `length` field.
std::string utf16be_to_utf8(const unsigned char* src, size_t code_units);
 
// Apply a best-effort NFD -> NFC composition pass over a UTF-8 string.
// Recognises a finite table of (base, combining) -> precomposed pairs
// covering most European Latin scripts (Italian, French, Spanish,
// German, Portuguese, Polish/Czech basics, Nordic). Everything not in
// the table is left as-is.
std::string nfd_to_nfc(const std::string& utf8);
 
// Convenience: utf16be -> utf8 -> nfd_to_nfc, then strip characters
// that are illegal in host filesystem paths (NUL, '/', and on Windows
// the additional set "<>:\"|?*" plus reserved names) and trim trailing
// dots / spaces (Windows-incompatible).
//
// On non-Windows hosts only NUL and '/' are stripped. Replacement
// character is '_'.
std::string hfs_filename_to_host(const unsigned char* src,
                                  size_t code_units);
 
} // namespace hfs