include/library/utf8.hpp

   1 #ifndef _library__utf8__hpp__included__
   2 #define _library__utf8__hpp__included__
   3
   4 #include <iostream>
   5 #include <cstdint>
   6 #include <cstdlib>
   7 #include <string>
   8 #include <functional>
   9
  10 namespace utf8
  11 {
  12 /**
  13  * Initial state for UTF-8 parser.
  14  */
  15 extern const uint16_t initial_state;
  16 /**
  17  * Parse a byte.
  18  *
  19  * Parameter ch: The character to parse. -1 for end of string.
  20  * Parameter state: The state. Mutated.
  21  * Returns: The codepoint, or -1 if no codepoint emitted.
  22  *
  23  * Note: When called with EOF, max 1 codepoint can be emitted.
  24  */
  25 int32_t parse_byte(int ch, uint16_t& state) throw();
  26 /**
  27  * Return length of string in UTF-8 codepoints.
  28  *
  29  * Parameter str: The string.
  30  * Returns: The length in codepoints.
  31  */
  32 size_t strlen(const std::string& str) throw();
  33
  34 /**
  35  * Transform UTF-8 into UTF-32.
  36  */
  37 std::u32string to32(const std::string& utf8);
  38
  39 /**
  40  * Transform UTF-32 into UTF-8.
  41  */
  42 std::string to8(const std::u32string& utf32);
  43
  44 /**
  45  * Iterator copy from UTF-8 to UTF-32
  46  */
  47 template<typename srcitr, typename dstitr>
  48 inline void to32i(srcitr begin, srcitr end, dstitr target)
  49 {
  50         uint16_t state = initial_state;
  51         for(srcitr i = begin; i != end; i++) {
  52                 int32_t x = parse_byte((unsigned char)*i, state);
  53                 if(x >= 0) {
  54                         *target = x;
  55                         ++target;
  56                 }
  57         }
  58         int32_t x = parse_byte(-1, state);
  59         if(x >= 0) {
  60                 *target = x;
  61                 ++target;
  62         }
  63 }
  64
  65 }
  66 #endif