minimal_UTF8/utf8.c

//@+leo-ver=5-thin
//@+node:caminhante.20200309141148.1: * @file ./utf8.c
//@@tabwidth -2
#include "utf8.h"
//@+others
//@+node:caminhante.20200309141148.3: ** static ascii_char
static bool ascii_char (const char *source) {
  switch (source[0]) {
  case '\x09': case '\x0A': case '\x0D': return true;
  case '\x20'...'\x7E': return true;
  default: return false; }
}
//@+node:caminhante.20231212193346.1: ** static utf8_multibyte_ending
static bool utf8_multibyte_ending (const char *source) {
  switch (source[0]) {
  case '\x80'...'\xBF': return true;
  default: return false; }
}
//@+node:caminhante.20231212193009.1: ** static utf8_2bytes_notoverlong
static bool utf8_2bytes_notoverlong (const char *source) {
  switch (source[0]) {
  case '\xC2'...'\xDF': return utf8_multibyte_ending(source+1);
  default: return false; }
}
//@+node:caminhante.20231212193621.1: ** static utf8_3bytes_notoverlong
static bool utf8_3bytes_notoverlong (const char *source) {
  switch (source[0]) {
  case '\xE0':
    switch (source[1]) {
    case '\xA0'...'\xBF': return utf8_multibyte_ending(source+2);
    default: return false; }
  default: return false; }
}
//@+node:caminhante.20231212193939.1: ** static utf8_3bytes
static bool utf8_3bytes (const char *source) {
  switch (source[0]) {
  case '\xE1'...'\xEC': case '\xEE': case '\xEF':
    return utf8_multibyte_ending(source+1) && utf8_multibyte_ending(source+2);
  default: return false; }
}
//@+node:caminhante.20231212194252.1: ** static utf8_3bytes_notsurrogate
static bool utf8_3bytes_notsurrogate (const char *source) {
  switch (source[0]) {
  case '\xED':
    switch (source[1]) {
    case '\x80'...'\x9F': return utf8_multibyte_ending(source+2);
    default: return false; }
  default: return false; }
}
//@+node:caminhante.20231212194457.1: ** static utf8_4bytes_planes1to3
static bool utf8_4bytes_planes1to3 (const char *source) {
  switch (source[0]) {
  case '\xF0':
    switch (source[1]) {
    case '\x90'...'\xBF':
      return utf8_multibyte_ending(source+2) && utf8_multibyte_ending(source+3);
    default: return false; }
  default: return false; }
}
//@+node:caminhante.20231212194735.1: ** static utf8_4bytes_planes4to15
static bool utf8_4bytes_planes4to15 (const char *source) {
  switch (source[0]) {
  case '\xF1'...'\xF3':
    return utf8_multibyte_ending(source+1) && utf8_multibyte_ending(source+2) && utf8_multibyte_ending(source+3);
  default: return false; }
}
//@+node:caminhante.20231212195030.1: ** static utf8_4bytes_plane16
static bool utf8_4bytes_plane16 (const char *source) {
  switch (source[0]) {
  case '\xF4':
    switch (source[1]) {
    case '\x80'...'\x8F':
      return utf8_multibyte_ending(source+2) && utf8_multibyte_ending(source+3);
    default: return false; }
  default: return false; }
}
//@+node:caminhante.20200309141148.7: ** uchar_valid
bool uchar_valid (char* source) {
  return ascii_char(source) ||
    utf8_2bytes_notoverlong(source) ||
    utf8_3bytes_notoverlong(source) || utf8_3bytes(source) || utf8_3bytes_notsurrogate(source) ||
    utf8_4bytes_planes1to3(source) || utf8_4bytes_planes4to15(source) || utf8_4bytes_plane16(source);
}
//@+node:caminhante.20200309141148.8: ** uchar_bytes
size_t uchar_bytes (char* source) {
  if (ascii_char(source)) {return 1;}
  if (utf8_2bytes_notoverlong(source)) {return 2;}
  if (utf8_3bytes_notoverlong(source) || utf8_3bytes(source) || utf8_3bytes_notsurrogate(source)) {return 3;}
  if (utf8_4bytes_planes1to3(source) || utf8_4bytes_planes4to15(source) || utf8_4bytes_plane16(source)) {return 4;}
  return 0;
}
//@+node:caminhante.20200309141148.9: ** ustring_length
size_t ustring_length (char* source) {
  size_t length = 0, a, p=0;
  do {
    a = uchar_bytes(source+p);
    p += a;
    if (a) {length++;}
  } while (a);
  return length;
}
//@+node:caminhante.20200309141148.10: ** ustring_bytes
size_t ustring_bytes (char* source) {
  size_t a, p=0;
  do {
    a = uchar_bytes(source+p);
    p += a;
  } while (a);
  return p;
}
//@+node:caminhante.20200309141148.11: ** cstring_bytes
size_t cstring_bytes (struct uchar* source) {
  size_t a = 0;
  while(source->bytes) {
    a += source->bytes;
    source++;
  }
  return a;
}
//@+node:caminhante.20200309141148.12: ** next_uchar
struct uchar next_uchar (char* source) {
  size_t bytes = uchar_bytes(source);
  if (bytes == 0) {return (struct uchar){0};}
  struct uchar uc = (struct uchar){
    .bytes=bytes,
    .chars[0]=source[0],
    .chars[1]=(bytes>=2 ? source[1] : 0),
    .chars[2]=(bytes>=3 ? source[2] : 0),
    .chars[3]=(bytes==4 ? source[3] : 0)
  };
  return uc;
}
//@+node:caminhante.20200309141148.13: ** c_to_ustring
void c_to_ustring (char* source, struct uchar* destination) {
  char *p = source;
  struct uchar uc, *us = destination;
  do {
    uc = next_uchar(p);
    p += uc.bytes;
    *us = uc;
    us ++;
  } while (uc.bytes);
}
//@+node:caminhante.20200309141148.14: ** u_to_cstring
void u_to_cstring (struct uchar* source, char* destination) {
  struct uchar *us = source;
  char *p = destination;
  while (us->bytes) {
    memcpy(p,us->chars,us->bytes);
    p += us->bytes;
    us ++;
  }
  p[0] = '\0';
}
//@+node:caminhante.20200309141148.15: ** uchar_puts
size_t uchar_puts (int fileno, struct uchar *uc) {
  return write(fileno, uc->chars, uc->bytes);
}
//@+node:caminhante.20200309141148.16: ** ustring_puts
size_t ustring_puts (int fileno, struct uchar *ustring) {
  size_t written = 0;
  while (ustring->bytes != 0) {
    size_t a = uchar_puts(fileno,ustring);
    written +=a;
    if (a < ustring->bytes) {break;}
    ustring++;
  }
  return written;
}
//@-others
//@-leo