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
   9 /**
  10  * Initial state for UTF-8 parser.
  11  */
  12 extern const uint16_t utf8_initial_state;
  13 /**
  14  * Parse a byte.
  15  *
  16  * Parameter ch: The character to parse. -1 for end of string.
  17  * Parameter state: The state. Mutated.
  18  * Returns: The codepoint, or -1 if no codepoint emitted.
  19  *
  20  * Note: When called with EOF, max 1 codepoint can be emitted.
  21  */
  22 int32_t utf8_parse_byte(int ch, uint16_t& state) throw();
  23 /**
  24  * Return length of string in UTF-8 codepoints.
  25  *
  26  * Parameter str: The string.
  27  * Returns: The length in codepoints.
  28  */
  29 size_t utf8_strlen(const std::string& str) throw();
  30
  31 /**
  32  * Transform UTF-8 into UTF-32.
  33  */
  34 std::u32string to_u32string(const std::string& utf8);
  35
  36 /**
  37  * Transform UTF-32 into UTF-8.
  38  */
  39 std::string to_u8string(const std::u32string& utf32);
  40
  41 /**
  42  * Iterator copy from UTF-8 to UTF-32
  43  */
  44 template<typename srcitr, typename dstitr>
  45 inline void copy_from_utf8(srcitr begin, srcitr end, dstitr target)
  46 {
  47         uint16_t state = utf8_initial_state;
  48         for(srcitr i = begin; i != end; i++) {
  49                 int32_t x = utf8_parse_byte((unsigned char)*i, state);
  50                 if(x >= 0) {
  51                         *target = x;
  52                         ++target;
  53                 }
  54         }
  55         int32_t x = utf8_parse_byte(-1, state);
  56         if(x >= 0) {
  57                 *target = x;
  58                 ++target;
  59         }
  60 }
  61
  62 #endif