utf8.c

   1 #include "git-compat-util.h"
   2 #include "utf8.h"
   3
   4 /* This code is originally from http://www.cl.cam.ac.uk/~mgk25/ucs/ */
   5
   6 struct interval {
   7   int first;
   8   int last;
   9 };
  10
  11 /* auxiliary function for binary search in interval table */
  12 static int bisearch(ucs_char_t ucs, const struct interval *table, int max)
  13 {
  14         int min = 0;
  15         int mid;
  16
  17         if (ucs < table[0].first || ucs > table[max].last)
  18                 return 0;
  19         while (max >= min) {
  20                 mid = (min + max) / 2;
  21                 if (ucs > table[mid].last)
  22                         min = mid + 1;
  23                 else if (ucs < table[mid].first)
  24                         max = mid - 1;
  25                 else
  26                         return 1;
  27         }
  28
  29         return 0;
  30 }
  31
  32 /* The following two functions define the column width of an ISO 10646
  33  * character as follows:
  34  *
  35  *    - The null character (U+0000) has a column width of 0.
  36  *
  37  *    - Other C0/C1 control characters and DEL will lead to a return
  38  *      value of -1.
  39  *
  40  *    - Non-spacing and enclosing combining characters (general
  41  *      category code Mn or Me in the Unicode database) have a
  42  *      column width of 0.
  43  *
  44  *    - SOFT HYPHEN (U+00AD) has a column width of 1.
  45  *
  46  *    - Other format characters (general category code Cf in the Unicode
  47  *      database) and ZERO WIDTH SPACE (U+200B) have a column width of 0.
  48  *
  49  *    - Hangul Jamo medial vowels and final consonants (U+1160-U+11FF)
  50  *      have a column width of 0.
  51  *
  52  *    - Spacing characters in the East Asian Wide (W) or East Asian
  53  *      Full-width (F) category as defined in Unicode Technical
  54  *      Report #11 have a column width of 2.
  55  *
  56  *    - All remaining characters (including all printable
  57  *      ISO 8859-1 and WGL4 characters, Unicode control characters,
  58  *      etc.) have a column width of 1.
  59  *
  60  * This implementation assumes that ucs_char_t characters are encoded
  61  * in ISO 10646.
  62  */
  63
  64 static int git_wcwidth(ucs_char_t ch)
  65 {
  66         /*
  67          * Sorted list of non-overlapping intervals of non-spacing characters,
  68          * generated by
  69          *   "uniset +cat=Me +cat=Mn +cat=Cf -00AD +1160-11FF +200B c".
  70          */
  71         static const struct interval combining[] = {
  72                 { 0x0300, 0x0357 }, { 0x035D, 0x036F }, { 0x0483, 0x0486 },
  73                 { 0x0488, 0x0489 }, { 0x0591, 0x05A1 }, { 0x05A3, 0x05B9 },
  74                 { 0x05BB, 0x05BD }, { 0x05BF, 0x05BF }, { 0x05C1, 0x05C2 },
  75                 { 0x05C4, 0x05C4 }, { 0x0600, 0x0603 }, { 0x0610, 0x0615 },
  76                 { 0x064B, 0x0658 }, { 0x0670, 0x0670 }, { 0x06D6, 0x06E4 },
  77                 { 0x06E7, 0x06E8 }, { 0x06EA, 0x06ED }, { 0x070F, 0x070F },
  78                 { 0x0711, 0x0711 }, { 0x0730, 0x074A }, { 0x07A6, 0x07B0 },
  79                 { 0x0901, 0x0902 }, { 0x093C, 0x093C }, { 0x0941, 0x0948 },
  80                 { 0x094D, 0x094D }, { 0x0951, 0x0954 }, { 0x0962, 0x0963 },
  81                 { 0x0981, 0x0981 }, { 0x09BC, 0x09BC }, { 0x09C1, 0x09C4 },
  82                 { 0x09CD, 0x09CD }, { 0x09E2, 0x09E3 }, { 0x0A01, 0x0A02 },
  83                 { 0x0A3C, 0x0A3C }, { 0x0A41, 0x0A42 }, { 0x0A47, 0x0A48 },
  84                 { 0x0A4B, 0x0A4D }, { 0x0A70, 0x0A71 }, { 0x0A81, 0x0A82 },
  85                 { 0x0ABC, 0x0ABC }, { 0x0AC1, 0x0AC5 }, { 0x0AC7, 0x0AC8 },
  86                 { 0x0ACD, 0x0ACD }, { 0x0AE2, 0x0AE3 }, { 0x0B01, 0x0B01 },
  87                 { 0x0B3C, 0x0B3C }, { 0x0B3F, 0x0B3F }, { 0x0B41, 0x0B43 },
  88                 { 0x0B4D, 0x0B4D }, { 0x0B56, 0x0B56 }, { 0x0B82, 0x0B82 },
  89                 { 0x0BC0, 0x0BC0 }, { 0x0BCD, 0x0BCD }, { 0x0C3E, 0x0C40 },
  90                 { 0x0C46, 0x0C48 }, { 0x0C4A, 0x0C4D }, { 0x0C55, 0x0C56 },
  91                 { 0x0CBC, 0x0CBC }, { 0x0CBF, 0x0CBF }, { 0x0CC6, 0x0CC6 },
  92                 { 0x0CCC, 0x0CCD }, { 0x0D41, 0x0D43 }, { 0x0D4D, 0x0D4D },
  93                 { 0x0DCA, 0x0DCA }, { 0x0DD2, 0x0DD4 }, { 0x0DD6, 0x0DD6 },
  94                 { 0x0E31, 0x0E31 }, { 0x0E34, 0x0E3A }, { 0x0E47, 0x0E4E },
  95                 { 0x0EB1, 0x0EB1 }, { 0x0EB4, 0x0EB9 }, { 0x0EBB, 0x0EBC },
  96                 { 0x0EC8, 0x0ECD }, { 0x0F18, 0x0F19 }, { 0x0F35, 0x0F35 },
  97                 { 0x0F37, 0x0F37 }, { 0x0F39, 0x0F39 }, { 0x0F71, 0x0F7E },
  98                 { 0x0F80, 0x0F84 }, { 0x0F86, 0x0F87 }, { 0x0F90, 0x0F97 },
  99                 { 0x0F99, 0x0FBC }, { 0x0FC6, 0x0FC6 }, { 0x102D, 0x1030 },
 100                 { 0x1032, 0x1032 }, { 0x1036, 0x1037 }, { 0x1039, 0x1039 },
 101                 { 0x1058, 0x1059 }, { 0x1160, 0x11FF }, { 0x1712, 0x1714 },
 102                 { 0x1732, 0x1734 }, { 0x1752, 0x1753 }, { 0x1772, 0x1773 },
 103                 { 0x17B4, 0x17B5 }, { 0x17B7, 0x17BD }, { 0x17C6, 0x17C6 },
 104                 { 0x17C9, 0x17D3 }, { 0x17DD, 0x17DD }, { 0x180B, 0x180D },
 105                 { 0x18A9, 0x18A9 }, { 0x1920, 0x1922 }, { 0x1927, 0x1928 },
 106                 { 0x1932, 0x1932 }, { 0x1939, 0x193B }, { 0x200B, 0x200F },
 107                 { 0x202A, 0x202E }, { 0x2060, 0x2063 }, { 0x206A, 0x206F },
 108                 { 0x20D0, 0x20EA }, { 0x302A, 0x302F }, { 0x3099, 0x309A },
 109                 { 0xFB1E, 0xFB1E }, { 0xFE00, 0xFE0F }, { 0xFE20, 0xFE23 },
 110                 { 0xFEFF, 0xFEFF }, { 0xFFF9, 0xFFFB }, { 0x1D167, 0x1D169 },
 111                 { 0x1D173, 0x1D182 }, { 0x1D185, 0x1D18B },
 112                 { 0x1D1AA, 0x1D1AD }, { 0xE0001, 0xE0001 },
 113                 { 0xE0020, 0xE007F }, { 0xE0100, 0xE01EF }
 114         };
 115
 116         /* test for 8-bit control characters */
 117         if (ch == 0)
 118                 return 0;
 119         if (ch < 32 || (ch >= 0x7f && ch < 0xa0))
 120                 return -1;
 121
 122         /* binary search in table of non-spacing characters */
 123         if (bisearch(ch, combining, sizeof(combining)
 124                                 / sizeof(struct interval) - 1))
 125                 return 0;
 126
 127         /*
 128          * If we arrive here, ch is neither a combining nor a C0/C1
 129          * control character.
 130          */
 131
 132         return 1 +
 133                 (ch >= 0x1100 &&
 134                     /* Hangul Jamo init. consonants */
 135                  (ch <= 0x115f ||
 136                   ch == 0x2329 || ch == 0x232a ||
 137                   /* CJK ... Yi */
 138                   (ch >= 0x2e80 && ch <= 0xa4cf &&
 139                    ch != 0x303f) ||
 140                   /* Hangul Syllables */
 141                   (ch >= 0xac00 && ch <= 0xd7a3) ||
 142                   /* CJK Compatibility Ideographs */
 143                   (ch >= 0xf900 && ch <= 0xfaff) ||
 144                   /* CJK Compatibility Forms */
 145                   (ch >= 0xfe30 && ch <= 0xfe6f) ||
 146                   /* Fullwidth Forms */
 147                   (ch >= 0xff00 && ch <= 0xff60) ||
 148                   (ch >= 0xffe0 && ch <= 0xffe6) ||
 149                   (ch >= 0x20000 && ch <= 0x2fffd) ||
 150                   (ch >= 0x30000 && ch <= 0x3fffd)));
 151 }
 152
 153 /*
 154  * Pick one ucs character starting from the location *start points at,
 155  * and return it, while updating the *start pointer to point at the
 156  * end of that character.  When remainder_p is not NULL, the location
 157  * holds the number of bytes remaining in the string that we are allowed
 158  * to pick from.  Otherwise we are allowed to pick up to the NUL that
 159  * would eventually appear in the string.  *remainder_p is also reduced
 160  * by the number of bytes we have consumed.
 161  *
 162  * If the string was not a valid UTF-8, *start pointer is set to NULL
 163  * and the return value is undefined.
 164  */
 165 ucs_char_t pick_one_utf8_char(const char **start, size_t *remainder_p)
 166 {
 167         unsigned char *s = (unsigned char *)*start;
 168         ucs_char_t ch;
 169         size_t remainder, incr;
 170
 171         /*
 172          * A caller that assumes NUL terminated text can choose
 173          * not to bother with the remainder length.  We will
 174          * stop at the first NUL.
 175          */
 176         remainder = (remainder_p ? *remainder_p : 999);
 177
 178         if (remainder < 1) {
 179                 goto invalid;
 180         } else if (*s < 0x80) {
 181                 /* 0xxxxxxx */
 182                 ch = *s;
 183                 incr = 1;
 184         } else if ((s[0] & 0xe0) == 0xc0) {
 185                 /* 110XXXXx 10xxxxxx */
 186                 if (remainder < 2 ||
 187                     (s[1] & 0xc0) != 0x80 ||
 188                     (s[0] & 0xfe) == 0xc0)
 189                         goto invalid;
 190                 ch = ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
 191                 incr = 2;
 192         } else if ((s[0] & 0xf0) == 0xe0) {
 193                 /* 1110XXXX 10Xxxxxx 10xxxxxx */
 194                 if (remainder < 3 ||
 195                     (s[1] & 0xc0) != 0x80 ||
 196                     (s[2] & 0xc0) != 0x80 ||
 197                     /* overlong? */
 198                     (s[0] == 0xe0 && (s[1] & 0xe0) == 0x80) ||
 199                     /* surrogate? */
 200                     (s[0] == 0xed && (s[1] & 0xe0) == 0xa0) ||
 201                     /* U+FFFE or U+FFFF? */
 202                     (s[0] == 0xef && s[1] == 0xbf &&
 203                      (s[2] & 0xfe) == 0xbe))
 204                         goto invalid;
 205                 ch = ((s[0] & 0x0f) << 12) |
 206                         ((s[1] & 0x3f) << 6) | (s[2] & 0x3f);
 207                 incr = 3;
 208         } else if ((s[0] & 0xf8) == 0xf0) {
 209                 /* 11110XXX 10XXxxxx 10xxxxxx 10xxxxxx */
 210                 if (remainder < 4 ||
 211                     (s[1] & 0xc0) != 0x80 ||
 212                     (s[2] & 0xc0) != 0x80 ||
 213                     (s[3] & 0xc0) != 0x80 ||
 214                     /* overlong? */
 215                     (s[0] == 0xf0 && (s[1] & 0xf0) == 0x80) ||
 216                     /* > U+10FFFF? */
 217                     (s[0] == 0xf4 && s[1] > 0x8f) || s[0] > 0xf4)
 218                         goto invalid;
 219                 ch = ((s[0] & 0x07) << 18) | ((s[1] & 0x3f) << 12) |
 220                         ((s[2] & 0x3f) << 6) | (s[3] & 0x3f);
 221                 incr = 4;
 222         } else {
 223 invalid:
 224                 *start = NULL;
 225                 return 0;
 226         }
 227
 228         *start += incr;
 229         if (remainder_p)
 230                 *remainder_p = remainder - incr;
 231         return ch;
 232 }
 233
 234 /*
 235  * This function returns the number of columns occupied by the character
 236  * pointed to by the variable start. The pointer is updated to point at
 237  * the next character. When remainder_p is not NULL, it points at the
 238  * location that stores the number of remaining bytes we can use to pick
 239  * a character (see pick_one_utf8_char() above).
 240  */
 241 int utf8_width(const char **start, size_t *remainder_p)
 242 {
 243         ucs_char_t ch = pick_one_utf8_char(start, remainder_p);
 244         if (!*start)
 245                 return 0;
 246         return git_wcwidth(ch);
 247 }
 248
 249 int is_utf8(const char *text)
 250 {
 251         while (*text) {
 252                 if (*text == '\n' || *text == '\t' || *text == '\r') {
 253                         text++;
 254                         continue;
 255                 }
 256                 utf8_width(&text, NULL);
 257                 if (!text)
 258                         return 0;
 259         }
 260         return 1;
 261 }
 262
 263 static void print_spaces(int count)
 264 {
 265         static const char s[] = "                    ";
 266         while (count >= sizeof(s)) {
 267                 fwrite(s, sizeof(s) - 1, 1, stdout);
 268                 count -= sizeof(s) - 1;
 269         }
 270         fwrite(s, count, 1, stdout);
 271 }
 272
 273 /*
 274  * Wrap the text, if necessary. The variable indent is the indent for the
 275  * first line, indent2 is the indent for all other lines.
 276  * If indent is negative, assume that already -indent columns have been
 277  * consumed (and no extra indent is necessary for the first line).
 278  */
 279 int print_wrapped_text(const char *text, int indent, int indent2, int width)
 280 {
 281         int w = indent, assume_utf8 = is_utf8(text);
 282         const char *bol = text, *space = NULL;
 283
 284         if (indent < 0) {
 285                 w = -indent;
 286                 space = text;
 287         }
 288
 289         for (;;) {
 290                 char c = *text;
 291                 if (!c || isspace(c)) {
 292                         if (w < width || !space) {
 293                                 const char *start = bol;
 294                                 if (space)
 295                                         start = space;
 296                                 else
 297                                         print_spaces(indent);
 298                                 fwrite(start, text - start, 1, stdout);
 299                                 if (!c)
 300                                         return w;
 301                                 else if (c == '\t')
 302                                         w |= 0x07;
 303                                 space = text;
 304                                 w++;
 305                                 text++;
 306                         }
 307                         else {
 308                                 putchar('\n');
 309                                 text = bol = space + isspace(*space);
 310                                 space = NULL;
 311                                 w = indent = indent2;
 312                         }
 313                         continue;
 314                 }
 315                 if (assume_utf8)
 316                         w += utf8_width(&text, NULL);
 317                 else {
 318                         w++;
 319                         text++;
 320                 }
 321         }
 322 }
 323
 324 int is_encoding_utf8(const char *name)
 325 {
 326         if (!name)
 327                 return 1;
 328         if (!strcasecmp(name, "utf-8") || !strcasecmp(name, "utf8"))
 329                 return 1;
 330         return 0;
 331 }
 332
 333 /*
 334  * Given a buffer and its encoding, return it re-encoded
 335  * with iconv.  If the conversion fails, returns NULL.
 336  */
 337 #ifndef NO_ICONV
 338 #ifdef OLD_ICONV
 339         typedef const char * iconv_ibp;
 340 #else
 341         typedef char * iconv_ibp;
 342 #endif
 343 char *reencode_string(const char *in, const char *out_encoding, const char *in_encoding)
 344 {
 345         iconv_t conv;
 346         size_t insz, outsz, outalloc;
 347         char *out, *outpos;
 348         iconv_ibp cp;
 349
 350         if (!in_encoding)
 351                 return NULL;
 352         conv = iconv_open(out_encoding, in_encoding);
 353         if (conv == (iconv_t) -1)
 354                 return NULL;
 355         insz = strlen(in);
 356         outsz = insz;
 357         outalloc = outsz + 1; /* for terminating NUL */
 358         out = xmalloc(outalloc);
 359         outpos = out;
 360         cp = (iconv_ibp)in;
 361
 362         while (1) {
 363                 size_t cnt = iconv(conv, &cp, &insz, &outpos, &outsz);
 364
 365                 if (cnt == -1) {
 366                         size_t sofar;
 367                         if (errno != E2BIG) {
 368                                 free(out);
 369                                 iconv_close(conv);
 370                                 return NULL;
 371                         }
 372                         /* insz has remaining number of bytes.
 373                          * since we started outsz the same as insz,
 374                          * it is likely that insz is not enough for
 375                          * converting the rest.
 376                          */
 377                         sofar = outpos - out;
 378                         outalloc = sofar + insz * 2 + 32;
 379                         out = xrealloc(out, outalloc);
 380                         outpos = out + sofar;
 381                         outsz = outalloc - sofar - 1;
 382                 }
 383                 else {
 384                         *outpos = '\0';
 385                         break;
 386                 }
 387         }
 388         iconv_close(conv);
 389         return out;
 390 }
 391 #else
 392 char *reencode_string(const char *in, const char *out_encoding, const char *in_encoding)
 393 {
 394         if (!in_encoding)
 395                 return NULL;
 396         return xstrdup(in);
 397 }
 398 #endif