utf8.c

   1 #include "git-compat-util.h"
   2 #include "strbuf.h"
   3 #include "utf8.h"
   4
   5 /* This code is originally from http://www.cl.cam.ac.uk/~mgk25/ucs/ */
   6
   7 struct interval {
   8   int first;
   9   int last;
  10 };
  11
  12 /* auxiliary function for binary search in interval table */
  13 static int bisearch(ucs_char_t ucs, const struct interval *table, int max)
  14 {
  15         int min = 0;
  16         int mid;
  17
  18         if (ucs < table[0].first || ucs > table[max].last)
  19                 return 0;
  20         while (max >= min) {
  21                 mid = (min + max) / 2;
  22                 if (ucs > table[mid].last)
  23                         min = mid + 1;
  24                 else if (ucs < table[mid].first)
  25                         max = mid - 1;
  26                 else
  27                         return 1;
  28         }
  29
  30         return 0;
  31 }
  32
  33 /* The following two functions define the column width of an ISO 10646
  34  * character as follows:
  35  *
  36  *    - The null character (U+0000) has a column width of 0.
  37  *
  38  *    - Other C0/C1 control characters and DEL will lead to a return
  39  *      value of -1.
  40  *
  41  *    - Non-spacing and enclosing combining characters (general
  42  *      category code Mn or Me in the Unicode database) have a
  43  *      column width of 0.
  44  *
  45  *    - SOFT HYPHEN (U+00AD) has a column width of 1.
  46  *
  47  *    - Other format characters (general category code Cf in the Unicode
  48  *      database) and ZERO WIDTH SPACE (U+200B) have a column width of 0.
  49  *
  50  *    - Hangul Jamo medial vowels and final consonants (U+1160-U+11FF)
  51  *      have a column width of 0.
  52  *
  53  *    - Spacing characters in the East Asian Wide (W) or East Asian
  54  *      Full-width (F) category as defined in Unicode Technical
  55  *      Report #11 have a column width of 2.
  56  *
  57  *    - All remaining characters (including all printable
  58  *      ISO 8859-1 and WGL4 characters, Unicode control characters,
  59  *      etc.) have a column width of 1.
  60  *
  61  * This implementation assumes that ucs_char_t characters are encoded
  62  * in ISO 10646.
  63  */
  64
  65 static int git_wcwidth(ucs_char_t ch)
  66 {
  67         /*
  68          * Sorted list of non-overlapping intervals of non-spacing characters,
  69          * generated by
  70          *   "uniset +cat=Me +cat=Mn +cat=Cf -00AD +1160-11FF +200B c".
  71          */
  72         static const struct interval combining[] = {
  73                 { 0x0300, 0x0357 }, { 0x035D, 0x036F }, { 0x0483, 0x0486 },
  74                 { 0x0488, 0x0489 }, { 0x0591, 0x05A1 }, { 0x05A3, 0x05B9 },
  75                 { 0x05BB, 0x05BD }, { 0x05BF, 0x05BF }, { 0x05C1, 0x05C2 },
  76                 { 0x05C4, 0x05C4 }, { 0x0600, 0x0603 }, { 0x0610, 0x0615 },
  77                 { 0x064B, 0x0658 }, { 0x0670, 0x0670 }, { 0x06D6, 0x06E4 },
  78                 { 0x06E7, 0x06E8 }, { 0x06EA, 0x06ED }, { 0x070F, 0x070F },
  79                 { 0x0711, 0x0711 }, { 0x0730, 0x074A }, { 0x07A6, 0x07B0 },
  80                 { 0x0901, 0x0902 }, { 0x093C, 0x093C }, { 0x0941, 0x0948 },
  81                 { 0x094D, 0x094D }, { 0x0951, 0x0954 }, { 0x0962, 0x0963 },
  82                 { 0x0981, 0x0981 }, { 0x09BC, 0x09BC }, { 0x09C1, 0x09C4 },
  83                 { 0x09CD, 0x09CD }, { 0x09E2, 0x09E3 }, { 0x0A01, 0x0A02 },
  84                 { 0x0A3C, 0x0A3C }, { 0x0A41, 0x0A42 }, { 0x0A47, 0x0A48 },
  85                 { 0x0A4B, 0x0A4D }, { 0x0A70, 0x0A71 }, { 0x0A81, 0x0A82 },
  86                 { 0x0ABC, 0x0ABC }, { 0x0AC1, 0x0AC5 }, { 0x0AC7, 0x0AC8 },
  87                 { 0x0ACD, 0x0ACD }, { 0x0AE2, 0x0AE3 }, { 0x0B01, 0x0B01 },
  88                 { 0x0B3C, 0x0B3C }, { 0x0B3F, 0x0B3F }, { 0x0B41, 0x0B43 },
  89                 { 0x0B4D, 0x0B4D }, { 0x0B56, 0x0B56 }, { 0x0B82, 0x0B82 },
  90                 { 0x0BC0, 0x0BC0 }, { 0x0BCD, 0x0BCD }, { 0x0C3E, 0x0C40 },
  91                 { 0x0C46, 0x0C48 }, { 0x0C4A, 0x0C4D }, { 0x0C55, 0x0C56 },
  92                 { 0x0CBC, 0x0CBC }, { 0x0CBF, 0x0CBF }, { 0x0CC6, 0x0CC6 },
  93                 { 0x0CCC, 0x0CCD }, { 0x0D41, 0x0D43 }, { 0x0D4D, 0x0D4D },
  94                 { 0x0DCA, 0x0DCA }, { 0x0DD2, 0x0DD4 }, { 0x0DD6, 0x0DD6 },
  95                 { 0x0E31, 0x0E31 }, { 0x0E34, 0x0E3A }, { 0x0E47, 0x0E4E },
  96                 { 0x0EB1, 0x0EB1 }, { 0x0EB4, 0x0EB9 }, { 0x0EBB, 0x0EBC },
  97                 { 0x0EC8, 0x0ECD }, { 0x0F18, 0x0F19 }, { 0x0F35, 0x0F35 },
  98                 { 0x0F37, 0x0F37 }, { 0x0F39, 0x0F39 }, { 0x0F71, 0x0F7E },
  99                 { 0x0F80, 0x0F84 }, { 0x0F86, 0x0F87 }, { 0x0F90, 0x0F97 },
 100                 { 0x0F99, 0x0FBC }, { 0x0FC6, 0x0FC6 }, { 0x102D, 0x1030 },
 101                 { 0x1032, 0x1032 }, { 0x1036, 0x1037 }, { 0x1039, 0x1039 },
 102                 { 0x1058, 0x1059 }, { 0x1160, 0x11FF }, { 0x1712, 0x1714 },
 103                 { 0x1732, 0x1734 }, { 0x1752, 0x1753 }, { 0x1772, 0x1773 },
 104                 { 0x17B4, 0x17B5 }, { 0x17B7, 0x17BD }, { 0x17C6, 0x17C6 },
 105                 { 0x17C9, 0x17D3 }, { 0x17DD, 0x17DD }, { 0x180B, 0x180D },
 106                 { 0x18A9, 0x18A9 }, { 0x1920, 0x1922 }, { 0x1927, 0x1928 },
 107                 { 0x1932, 0x1932 }, { 0x1939, 0x193B }, { 0x200B, 0x200F },
 108                 { 0x202A, 0x202E }, { 0x2060, 0x2063 }, { 0x206A, 0x206F },
 109                 { 0x20D0, 0x20EA }, { 0x302A, 0x302F }, { 0x3099, 0x309A },
 110                 { 0xFB1E, 0xFB1E }, { 0xFE00, 0xFE0F }, { 0xFE20, 0xFE23 },
 111                 { 0xFEFF, 0xFEFF }, { 0xFFF9, 0xFFFB }, { 0x1D167, 0x1D169 },
 112                 { 0x1D173, 0x1D182 }, { 0x1D185, 0x1D18B },
 113                 { 0x1D1AA, 0x1D1AD }, { 0xE0001, 0xE0001 },
 114                 { 0xE0020, 0xE007F }, { 0xE0100, 0xE01EF }
 115         };
 116
 117         /* test for 8-bit control characters */
 118         if (ch == 0)
 119                 return 0;
 120         if (ch < 32 || (ch >= 0x7f && ch < 0xa0))
 121                 return -1;
 122
 123         /* binary search in table of non-spacing characters */
 124         if (bisearch(ch, combining, sizeof(combining)
 125                                 / sizeof(struct interval) - 1))
 126                 return 0;
 127
 128         /*
 129          * If we arrive here, ch is neither a combining nor a C0/C1
 130          * control character.
 131          */
 132
 133         return 1 +
 134                 (ch >= 0x1100 &&
 135                     /* Hangul Jamo init. consonants */
 136                  (ch <= 0x115f ||
 137                   ch == 0x2329 || ch == 0x232a ||
 138                   /* CJK ... Yi */
 139                   (ch >= 0x2e80 && ch <= 0xa4cf &&
 140                    ch != 0x303f) ||
 141                   /* Hangul Syllables */
 142                   (ch >= 0xac00 && ch <= 0xd7a3) ||
 143                   /* CJK Compatibility Ideographs */
 144                   (ch >= 0xf900 && ch <= 0xfaff) ||
 145                   /* CJK Compatibility Forms */
 146                   (ch >= 0xfe30 && ch <= 0xfe6f) ||
 147                   /* Fullwidth Forms */
 148                   (ch >= 0xff00 && ch <= 0xff60) ||
 149                   (ch >= 0xffe0 && ch <= 0xffe6) ||
 150                   (ch >= 0x20000 && ch <= 0x2fffd) ||
 151                   (ch >= 0x30000 && ch <= 0x3fffd)));
 152 }
 153
 154 /*
 155  * Pick one ucs character starting from the location *start points at,
 156  * and return it, while updating the *start pointer to point at the
 157  * end of that character.  When remainder_p is not NULL, the location
 158  * holds the number of bytes remaining in the string that we are allowed
 159  * to pick from.  Otherwise we are allowed to pick up to the NUL that
 160  * would eventually appear in the string.  *remainder_p is also reduced
 161  * by the number of bytes we have consumed.
 162  *
 163  * If the string was not a valid UTF-8, *start pointer is set to NULL
 164  * and the return value is undefined.
 165  */
 166 static ucs_char_t pick_one_utf8_char(const char **start, size_t *remainder_p)
 167 {
 168         unsigned char *s = (unsigned char *)*start;
 169         ucs_char_t ch;
 170         size_t remainder, incr;
 171
 172         /*
 173          * A caller that assumes NUL terminated text can choose
 174          * not to bother with the remainder length.  We will
 175          * stop at the first NUL.
 176          */
 177         remainder = (remainder_p ? *remainder_p : 999);
 178
 179         if (remainder < 1) {
 180                 goto invalid;
 181         } else if (*s < 0x80) {
 182                 /* 0xxxxxxx */
 183                 ch = *s;
 184                 incr = 1;
 185         } else if ((s[0] & 0xe0) == 0xc0) {
 186                 /* 110XXXXx 10xxxxxx */
 187                 if (remainder < 2 ||
 188                     (s[1] & 0xc0) != 0x80 ||
 189                     (s[0] & 0xfe) == 0xc0)
 190                         goto invalid;
 191                 ch = ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
 192                 incr = 2;
 193         } else if ((s[0] & 0xf0) == 0xe0) {
 194                 /* 1110XXXX 10Xxxxxx 10xxxxxx */
 195                 if (remainder < 3 ||
 196                     (s[1] & 0xc0) != 0x80 ||
 197                     (s[2] & 0xc0) != 0x80 ||
 198                     /* overlong? */
 199                     (s[0] == 0xe0 && (s[1] & 0xe0) == 0x80) ||
 200                     /* surrogate? */
 201                     (s[0] == 0xed && (s[1] & 0xe0) == 0xa0) ||
 202                     /* U+FFFE or U+FFFF? */
 203                     (s[0] == 0xef && s[1] == 0xbf &&
 204                      (s[2] & 0xfe) == 0xbe))
 205                         goto invalid;
 206                 ch = ((s[0] & 0x0f) << 12) |
 207                         ((s[1] & 0x3f) << 6) | (s[2] & 0x3f);
 208                 incr = 3;
 209         } else if ((s[0] & 0xf8) == 0xf0) {
 210                 /* 11110XXX 10XXxxxx 10xxxxxx 10xxxxxx */
 211                 if (remainder < 4 ||
 212                     (s[1] & 0xc0) != 0x80 ||
 213                     (s[2] & 0xc0) != 0x80 ||
 214                     (s[3] & 0xc0) != 0x80 ||
 215                     /* overlong? */
 216                     (s[0] == 0xf0 && (s[1] & 0xf0) == 0x80) ||
 217                     /* > U+10FFFF? */
 218                     (s[0] == 0xf4 && s[1] > 0x8f) || s[0] > 0xf4)
 219                         goto invalid;
 220                 ch = ((s[0] & 0x07) << 18) | ((s[1] & 0x3f) << 12) |
 221                         ((s[2] & 0x3f) << 6) | (s[3] & 0x3f);
 222                 incr = 4;
 223         } else {
 224 invalid:
 225                 *start = NULL;
 226                 return 0;
 227         }
 228
 229         *start += incr;
 230         if (remainder_p)
 231                 *remainder_p = remainder - incr;
 232         return ch;
 233 }
 234
 235 /*
 236  * This function returns the number of columns occupied by the character
 237  * pointed to by the variable start. The pointer is updated to point at
 238  * the next character. When remainder_p is not NULL, it points at the
 239  * location that stores the number of remaining bytes we can use to pick
 240  * a character (see pick_one_utf8_char() above).
 241  */
 242 int utf8_width(const char **start, size_t *remainder_p)
 243 {
 244         ucs_char_t ch = pick_one_utf8_char(start, remainder_p);
 245         if (!*start)
 246                 return 0;
 247         return git_wcwidth(ch);
 248 }
 249
 250 /*
 251  * Returns the total number of columns required by a null-terminated
 252  * string, assuming that the string is utf8.  Returns strlen() instead
 253  * if the string does not look like a valid utf8 string.
 254  */
 255 int utf8_strwidth(const char *string)
 256 {
 257         int width = 0;
 258         const char *orig = string;
 259
 260         while (1) {
 261                 if (!string)
 262                         return strlen(orig);
 263                 if (!*string)
 264                         return width;
 265                 width += utf8_width(&string, NULL);
 266         }
 267 }
 268
 269 int is_utf8(const char *text)
 270 {
 271         while (*text) {
 272                 if (*text == '\n' || *text == '\t' || *text == '\r') {
 273                         text++;
 274                         continue;
 275                 }
 276                 utf8_width(&text, NULL);
 277                 if (!text)
 278                         return 0;
 279         }
 280         return 1;
 281 }
 282
 283 static void strbuf_addchars(struct strbuf *sb, int c, size_t n)
 284 {
 285         strbuf_grow(sb, n);
 286         memset(sb->buf + sb->len, c, n);
 287         strbuf_setlen(sb, sb->len + n);
 288 }
 289
 290 static void strbuf_add_indented_text(struct strbuf *buf, const char *text,
 291                                      int indent, int indent2)
 292 {
 293         if (indent < 0)
 294                 indent = 0;
 295         while (*text) {
 296                 const char *eol = strchrnul(text, '\n');
 297                 if (*eol == '\n')
 298                         eol++;
 299                 strbuf_addchars(buf, ' ', indent);
 300                 strbuf_add(buf, text, eol - text);
 301                 text = eol;
 302                 indent = indent2;
 303         }
 304 }
 305
 306 static size_t display_mode_esc_sequence_len(const char *s)
 307 {
 308         const char *p = s;
 309         if (*p++ != '\033')
 310                 return 0;
 311         if (*p++ != '[')
 312                 return 0;
 313         while (isdigit(*p) || *p == ';')
 314                 p++;
 315         if (*p++ != 'm')
 316                 return 0;
 317         return p - s;
 318 }
 319
 320 /*
 321  * Wrap the text, if necessary. The variable indent is the indent for the
 322  * first line, indent2 is the indent for all other lines.
 323  * If indent is negative, assume that already -indent columns have been
 324  * consumed (and no extra indent is necessary for the first line).
 325  */
 326 void strbuf_add_wrapped_text(struct strbuf *buf,
 327                 const char *text, int indent1, int indent2, int width)
 328 {
 329         int indent, w, assume_utf8 = 1;
 330         const char *bol, *space, *start = text;
 331         size_t orig_len = buf->len;
 332
 333         if (width <= 0) {
 334                 strbuf_add_indented_text(buf, text, indent1, indent2);
 335                 return;
 336         }
 337
 338 retry:
 339         bol = text;
 340         w = indent = indent1;
 341         space = NULL;
 342         if (indent < 0) {
 343                 w = -indent;
 344                 space = text;
 345         }
 346
 347         for (;;) {
 348                 char c;
 349                 size_t skip;
 350
 351                 while ((skip = display_mode_esc_sequence_len(text)))
 352                         text += skip;
 353
 354                 c = *text;
 355                 if (!c || isspace(c)) {
 356                         if (w <= width || !space) {
 357                                 const char *start = bol;
 358                                 if (!c && text == start)
 359                                         return;
 360                                 if (space)
 361                                         start = space;
 362                                 else
 363                                         strbuf_addchars(buf, ' ', indent);
 364                                 strbuf_add(buf, start, text - start);
 365                                 if (!c)
 366                                         return;
 367                                 space = text;
 368                                 if (c == '\t')
 369                                         w |= 0x07;
 370                                 else if (c == '\n') {
 371                                         space++;
 372                                         if (*space == '\n') {
 373                                                 strbuf_addch(buf, '\n');
 374                                                 goto new_line;
 375                                         }
 376                                         else if (!isalnum(*space))
 377                                                 goto new_line;
 378                                         else
 379                                                 strbuf_addch(buf, ' ');
 380                                 }
 381                                 w++;
 382                                 text++;
 383                         }
 384                         else {
 385 new_line:
 386                                 strbuf_addch(buf, '\n');
 387                                 text = bol = space + isspace(*space);
 388                                 space = NULL;
 389                                 w = indent = indent2;
 390                         }
 391                         continue;
 392                 }
 393                 if (assume_utf8) {
 394                         w += utf8_width(&text, NULL);
 395                         if (!text) {
 396                                 assume_utf8 = 0;
 397                                 text = start;
 398                                 strbuf_setlen(buf, orig_len);
 399                                 goto retry;
 400                         }
 401                 } else {
 402                         w++;
 403                         text++;
 404                 }
 405         }
 406 }
 407
 408 void strbuf_add_wrapped_bytes(struct strbuf *buf, const char *data, int len,
 409                              int indent, int indent2, int width)
 410 {
 411         char *tmp = xstrndup(data, len);
 412         strbuf_add_wrapped_text(buf, tmp, indent, indent2, width);
 413         free(tmp);
 414 }
 415
 416 int is_encoding_utf8(const char *name)
 417 {
 418         if (!name)
 419                 return 1;
 420         if (!strcasecmp(name, "utf-8") || !strcasecmp(name, "utf8"))
 421                 return 1;
 422         return 0;
 423 }
 424
 425 int same_encoding(const char *src, const char *dst)
 426 {
 427         if (is_encoding_utf8(src) && is_encoding_utf8(dst))
 428                 return 1;
 429         return !strcasecmp(src, dst);
 430 }
 431
 432 /*
 433  * Wrapper for fprintf and returns the total number of columns required
 434  * for the printed string, assuming that the string is utf8.
 435  */
 436 int utf8_fprintf(FILE *stream, const char *format, ...)
 437 {
 438         struct strbuf buf = STRBUF_INIT;
 439         va_list arg;
 440         int columns;
 441
 442         va_start(arg, format);
 443         strbuf_vaddf(&buf, format, arg);
 444         va_end(arg);
 445
 446         columns = fputs(buf.buf, stream);
 447         if (0 <= columns) /* keep the error from the I/O */
 448                 columns = utf8_strwidth(buf.buf);
 449         strbuf_release(&buf);
 450         return columns;
 451 }
 452
 453 /*
 454  * Given a buffer and its encoding, return it re-encoded
 455  * with iconv.  If the conversion fails, returns NULL.
 456  */
 457 #ifndef NO_ICONV
 458 #if defined(OLD_ICONV) || (defined(__sun__) && !defined(_XPG6))
 459         typedef const char * iconv_ibp;
 460 #else
 461         typedef char * iconv_ibp;
 462 #endif
 463 char *reencode_string_iconv(const char *in, size_t insz, iconv_t conv)
 464 {
 465         size_t outsz, outalloc;
 466         char *out, *outpos;
 467         iconv_ibp cp;
 468
 469         outsz = insz;
 470         outalloc = outsz + 1; /* for terminating NUL */
 471         out = xmalloc(outalloc);
 472         outpos = out;
 473         cp = (iconv_ibp)in;
 474
 475         while (1) {
 476                 size_t cnt = iconv(conv, &cp, &insz, &outpos, &outsz);
 477
 478                 if (cnt == -1) {
 479                         size_t sofar;
 480                         if (errno != E2BIG) {
 481                                 free(out);
 482                                 return NULL;
 483                         }
 484                         /* insz has remaining number of bytes.
 485                          * since we started outsz the same as insz,
 486                          * it is likely that insz is not enough for
 487                          * converting the rest.
 488                          */
 489                         sofar = outpos - out;
 490                         outalloc = sofar + insz * 2 + 32;
 491                         out = xrealloc(out, outalloc);
 492                         outpos = out + sofar;
 493                         outsz = outalloc - sofar - 1;
 494                 }
 495                 else {
 496                         *outpos = '\0';
 497                         break;
 498                 }
 499         }
 500         return out;
 501 }
 502
 503 char *reencode_string(const char *in, const char *out_encoding, const char *in_encoding)
 504 {
 505         iconv_t conv;
 506         char *out;
 507
 508         if (!in_encoding)
 509                 return NULL;
 510
 511         conv = iconv_open(out_encoding, in_encoding);
 512         if (conv == (iconv_t) -1) {
 513                 /*
 514                  * Some platforms do not have the variously spelled variants of
 515                  * UTF-8, so let's fall back to trying the most official
 516                  * spelling. We do so only as a fallback in case the platform
 517                  * does understand the user's spelling, but not our official
 518                  * one.
 519                  */
 520                 if (is_encoding_utf8(in_encoding))
 521                         in_encoding = "UTF-8";
 522                 if (is_encoding_utf8(out_encoding))
 523                         out_encoding = "UTF-8";
 524                 conv = iconv_open(out_encoding, in_encoding);
 525                 if (conv == (iconv_t) -1)
 526                         return NULL;
 527         }
 528
 529         out = reencode_string_iconv(in, strlen(in), conv);
 530         iconv_close(conv);
 531         return out;
 532 }
 533 #endif
 534
 535 /*
 536  * Returns first character length in bytes for multi-byte `text` according to
 537  * `encoding`.
 538  *
 539  * - The `text` pointer is updated to point at the next character.
 540  * - When `remainder_p` is not NULL, on entry `*remainder_p` is how much bytes
 541  *   we can consume from text, and on exit `*remainder_p` is reduced by returned
 542  *   character length. Otherwise `text` is treated as limited by NUL.
 543  */
 544 int mbs_chrlen(const char **text, size_t *remainder_p, const char *encoding)
 545 {
 546         int chrlen;
 547         const char *p = *text;
 548         size_t r = (remainder_p ? *remainder_p : SIZE_MAX);
 549
 550         if (r < 1)
 551                 return 0;
 552
 553         if (is_encoding_utf8(encoding)) {
 554                 pick_one_utf8_char(&p, &r);
 555
 556                 chrlen = p ? (p - *text)
 557                            : 1 /* not valid UTF-8 -> raw byte sequence */;
 558         }
 559         else {
 560                 /*
 561                  * TODO use iconv to decode one char and obtain its chrlen
 562                  * for now, let's treat encodings != UTF-8 as one-byte
 563                  */
 564                 chrlen = 1;
 565         }
 566
 567         *text += chrlen;
 568         if (remainder_p)
 569                 *remainder_p -= chrlen;
 570
 571         return chrlen;
 572 }