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 inline void strbuf_write(struct strbuf *sb, const char *buf, int len)
 284 {
 285         if (sb)
 286                 strbuf_insert(sb, sb->len, buf, len);
 287         else
 288                 fwrite(buf, len, 1, stdout);
 289 }
 290
 291 static void print_spaces(struct strbuf *buf, int count)
 292 {
 293         static const char s[] = "                    ";
 294         while (count >= sizeof(s)) {
 295                 strbuf_write(buf, s, sizeof(s) - 1);
 296                 count -= sizeof(s) - 1;
 297         }
 298         strbuf_write(buf, s, count);
 299 }
 300
 301 static void strbuf_add_indented_text(struct strbuf *buf, const char *text,
 302                                      int indent, int indent2)
 303 {
 304         if (indent < 0)
 305                 indent = 0;
 306         while (*text) {
 307                 const char *eol = strchrnul(text, '\n');
 308                 if (*eol == '\n')
 309                         eol++;
 310                 print_spaces(buf, indent);
 311                 strbuf_write(buf, text, eol - text);
 312                 text = eol;
 313                 indent = indent2;
 314         }
 315 }
 316
 317 static size_t display_mode_esc_sequence_len(const char *s)
 318 {
 319         const char *p = s;
 320         if (*p++ != '\033')
 321                 return 0;
 322         if (*p++ != '[')
 323                 return 0;
 324         while (isdigit(*p) || *p == ';')
 325                 p++;
 326         if (*p++ != 'm')
 327                 return 0;
 328         return p - s;
 329 }
 330
 331 /*
 332  * Wrap the text, if necessary. The variable indent is the indent for the
 333  * first line, indent2 is the indent for all other lines.
 334  * If indent is negative, assume that already -indent columns have been
 335  * consumed (and no extra indent is necessary for the first line).
 336  */
 337 int strbuf_add_wrapped_text(struct strbuf *buf,
 338                 const char *text, int indent, int indent2, int width)
 339 {
 340         int w = indent, assume_utf8 = is_utf8(text);
 341         const char *bol = text, *space = NULL;
 342
 343         if (width <= 0) {
 344                 strbuf_add_indented_text(buf, text, indent, indent2);
 345                 return 1;
 346         }
 347
 348         if (indent < 0) {
 349                 w = -indent;
 350                 space = text;
 351         }
 352
 353         for (;;) {
 354                 char c;
 355                 size_t skip;
 356
 357                 while ((skip = display_mode_esc_sequence_len(text)))
 358                         text += skip;
 359
 360                 c = *text;
 361                 if (!c || isspace(c)) {
 362                         if (w < width || !space) {
 363                                 const char *start = bol;
 364                                 if (!c && text == start)
 365                                         return w;
 366                                 if (space)
 367                                         start = space;
 368                                 else
 369                                         print_spaces(buf, indent);
 370                                 strbuf_write(buf, start, text - start);
 371                                 if (!c)
 372                                         return w;
 373                                 space = text;
 374                                 if (c == '\t')
 375                                         w |= 0x07;
 376                                 else if (c == '\n') {
 377                                         space++;
 378                                         if (*space == '\n') {
 379                                                 strbuf_write(buf, "\n", 1);
 380                                                 goto new_line;
 381                                         }
 382                                         else if (!isalnum(*space))
 383                                                 goto new_line;
 384                                         else
 385                                                 strbuf_write(buf, " ", 1);
 386                                 }
 387                                 w++;
 388                                 text++;
 389                         }
 390                         else {
 391 new_line:
 392                                 strbuf_write(buf, "\n", 1);
 393                                 text = bol = space + isspace(*space);
 394                                 space = NULL;
 395                                 w = indent = indent2;
 396                         }
 397                         continue;
 398                 }
 399                 if (assume_utf8)
 400                         w += utf8_width(&text, NULL);
 401                 else {
 402                         w++;
 403                         text++;
 404                 }
 405         }
 406 }
 407
 408 int print_wrapped_text(const char *text, int indent, int indent2, int width)
 409 {
 410         return strbuf_add_wrapped_text(NULL, text, indent, indent2, width);
 411 }
 412
 413 int is_encoding_utf8(const char *name)
 414 {
 415         if (!name)
 416                 return 1;
 417         if (!strcasecmp(name, "utf-8") || !strcasecmp(name, "utf8"))
 418                 return 1;
 419         return 0;
 420 }
 421
 422 /*
 423  * Given a buffer and its encoding, return it re-encoded
 424  * with iconv.  If the conversion fails, returns NULL.
 425  */
 426 #ifndef NO_ICONV
 427 #if defined(OLD_ICONV) || (defined(__sun__) && !defined(_XPG6))
 428         typedef const char * iconv_ibp;
 429 #else
 430         typedef char * iconv_ibp;
 431 #endif
 432 char *reencode_string(const char *in, const char *out_encoding, const char *in_encoding)
 433 {
 434         iconv_t conv;
 435         size_t insz, outsz, outalloc;
 436         char *out, *outpos;
 437         iconv_ibp cp;
 438
 439         if (!in_encoding)
 440                 return NULL;
 441         conv = iconv_open(out_encoding, in_encoding);
 442         if (conv == (iconv_t) -1)
 443                 return NULL;
 444         insz = strlen(in);
 445         outsz = insz;
 446         outalloc = outsz + 1; /* for terminating NUL */
 447         out = xmalloc(outalloc);
 448         outpos = out;
 449         cp = (iconv_ibp)in;
 450
 451         while (1) {
 452                 size_t cnt = iconv(conv, &cp, &insz, &outpos, &outsz);
 453
 454                 if (cnt == -1) {
 455                         size_t sofar;
 456                         if (errno != E2BIG) {
 457                                 free(out);
 458                                 iconv_close(conv);
 459                                 return NULL;
 460                         }
 461                         /* insz has remaining number of bytes.
 462                          * since we started outsz the same as insz,
 463                          * it is likely that insz is not enough for
 464                          * converting the rest.
 465                          */
 466                         sofar = outpos - out;
 467                         outalloc = sofar + insz * 2 + 32;
 468                         out = xrealloc(out, outalloc);
 469                         outpos = out + sofar;
 470                         outsz = outalloc - sofar - 1;
 471                 }
 472                 else {
 473                         *outpos = '\0';
 474                         break;
 475                 }
 476         }
 477         iconv_close(conv);
 478         return out;
 479 }
 480 #endif