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         ucs_char_t first;
   9         ucs_char_t last;
  10 };
  11
  12 size_t display_mode_esc_sequence_len(const char *s)
  13 {
  14         const char *p = s;
  15         if (*p++ != '\033')
  16                 return 0;
  17         if (*p++ != '[')
  18                 return 0;
  19         while (isdigit(*p) || *p == ';')
  20                 p++;
  21         if (*p++ != 'm')
  22                 return 0;
  23         return p - s;
  24 }
  25
  26 /* auxiliary function for binary search in interval table */
  27 static int bisearch(ucs_char_t ucs, const struct interval *table, int max)
  28 {
  29         int min = 0;
  30         int mid;
  31
  32         if (ucs < table[0].first || ucs > table[max].last)
  33                 return 0;
  34         while (max >= min) {
  35                 mid = (min + max) / 2;
  36                 if (ucs > table[mid].last)
  37                         min = mid + 1;
  38                 else if (ucs < table[mid].first)
  39                         max = mid - 1;
  40                 else
  41                         return 1;
  42         }
  43
  44         return 0;
  45 }
  46
  47 /* The following two functions define the column width of an ISO 10646
  48  * character as follows:
  49  *
  50  *    - The null character (U+0000) has a column width of 0.
  51  *
  52  *    - Other C0/C1 control characters and DEL will lead to a return
  53  *      value of -1.
  54  *
  55  *    - Non-spacing and enclosing combining characters (general
  56  *      category code Mn or Me in the Unicode database) have a
  57  *      column width of 0.
  58  *
  59  *    - SOFT HYPHEN (U+00AD) has a column width of 1.
  60  *
  61  *    - Other format characters (general category code Cf in the Unicode
  62  *      database) and ZERO WIDTH SPACE (U+200B) have a column width of 0.
  63  *
  64  *    - Hangul Jamo medial vowels and final consonants (U+1160-U+11FF)
  65  *      have a column width of 0.
  66  *
  67  *    - Spacing characters in the East Asian Wide (W) or East Asian
  68  *      Full-width (F) category as defined in Unicode Technical
  69  *      Report #11 have a column width of 2.
  70  *
  71  *    - All remaining characters (including all printable
  72  *      ISO 8859-1 and WGL4 characters, Unicode control characters,
  73  *      etc.) have a column width of 1.
  74  *
  75  * This implementation assumes that ucs_char_t characters are encoded
  76  * in ISO 10646.
  77  */
  78
  79 static int git_wcwidth(ucs_char_t ch)
  80 {
  81         /*
  82          * Sorted list of non-overlapping intervals of non-spacing characters,
  83          */
  84 #include "unicode_width.h"
  85
  86         /* test for 8-bit control characters */
  87         if (ch == 0)
  88                 return 0;
  89         if (ch < 32 || (ch >= 0x7f && ch < 0xa0))
  90                 return -1;
  91
  92         /* binary search in table of non-spacing characters */
  93         if (bisearch(ch, zero_width, sizeof(zero_width)
  94                                 / sizeof(struct interval) - 1))
  95                 return 0;
  96
  97         /* binary search in table of double width characters */
  98         if (bisearch(ch, double_width, sizeof(double_width)
  99                                 / sizeof(struct interval) - 1))
 100                 return 2;
 101
 102         return 1;
 103 }
 104
 105 /*
 106  * Pick one ucs character starting from the location *start points at,
 107  * and return it, while updating the *start pointer to point at the
 108  * end of that character.  When remainder_p is not NULL, the location
 109  * holds the number of bytes remaining in the string that we are allowed
 110  * to pick from.  Otherwise we are allowed to pick up to the NUL that
 111  * would eventually appear in the string.  *remainder_p is also reduced
 112  * by the number of bytes we have consumed.
 113  *
 114  * If the string was not a valid UTF-8, *start pointer is set to NULL
 115  * and the return value is undefined.
 116  */
 117 static ucs_char_t pick_one_utf8_char(const char **start, size_t *remainder_p)
 118 {
 119         unsigned char *s = (unsigned char *)*start;
 120         ucs_char_t ch;
 121         size_t remainder, incr;
 122
 123         /*
 124          * A caller that assumes NUL terminated text can choose
 125          * not to bother with the remainder length.  We will
 126          * stop at the first NUL.
 127          */
 128         remainder = (remainder_p ? *remainder_p : 999);
 129
 130         if (remainder < 1) {
 131                 goto invalid;
 132         } else if (*s < 0x80) {
 133                 /* 0xxxxxxx */
 134                 ch = *s;
 135                 incr = 1;
 136         } else if ((s[0] & 0xe0) == 0xc0) {
 137                 /* 110XXXXx 10xxxxxx */
 138                 if (remainder < 2 ||
 139                     (s[1] & 0xc0) != 0x80 ||
 140                     (s[0] & 0xfe) == 0xc0)
 141                         goto invalid;
 142                 ch = ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
 143                 incr = 2;
 144         } else if ((s[0] & 0xf0) == 0xe0) {
 145                 /* 1110XXXX 10Xxxxxx 10xxxxxx */
 146                 if (remainder < 3 ||
 147                     (s[1] & 0xc0) != 0x80 ||
 148                     (s[2] & 0xc0) != 0x80 ||
 149                     /* overlong? */
 150                     (s[0] == 0xe0 && (s[1] & 0xe0) == 0x80) ||
 151                     /* surrogate? */
 152                     (s[0] == 0xed && (s[1] & 0xe0) == 0xa0) ||
 153                     /* U+FFFE or U+FFFF? */
 154                     (s[0] == 0xef && s[1] == 0xbf &&
 155                      (s[2] & 0xfe) == 0xbe))
 156                         goto invalid;
 157                 ch = ((s[0] & 0x0f) << 12) |
 158                         ((s[1] & 0x3f) << 6) | (s[2] & 0x3f);
 159                 incr = 3;
 160         } else if ((s[0] & 0xf8) == 0xf0) {
 161                 /* 11110XXX 10XXxxxx 10xxxxxx 10xxxxxx */
 162                 if (remainder < 4 ||
 163                     (s[1] & 0xc0) != 0x80 ||
 164                     (s[2] & 0xc0) != 0x80 ||
 165                     (s[3] & 0xc0) != 0x80 ||
 166                     /* overlong? */
 167                     (s[0] == 0xf0 && (s[1] & 0xf0) == 0x80) ||
 168                     /* > U+10FFFF? */
 169                     (s[0] == 0xf4 && s[1] > 0x8f) || s[0] > 0xf4)
 170                         goto invalid;
 171                 ch = ((s[0] & 0x07) << 18) | ((s[1] & 0x3f) << 12) |
 172                         ((s[2] & 0x3f) << 6) | (s[3] & 0x3f);
 173                 incr = 4;
 174         } else {
 175 invalid:
 176                 *start = NULL;
 177                 return 0;
 178         }
 179
 180         *start += incr;
 181         if (remainder_p)
 182                 *remainder_p = remainder - incr;
 183         return ch;
 184 }
 185
 186 /*
 187  * This function returns the number of columns occupied by the character
 188  * pointed to by the variable start. The pointer is updated to point at
 189  * the next character. When remainder_p is not NULL, it points at the
 190  * location that stores the number of remaining bytes we can use to pick
 191  * a character (see pick_one_utf8_char() above).
 192  */
 193 int utf8_width(const char **start, size_t *remainder_p)
 194 {
 195         ucs_char_t ch = pick_one_utf8_char(start, remainder_p);
 196         if (!*start)
 197                 return 0;
 198         return git_wcwidth(ch);
 199 }
 200
 201 /*
 202  * Returns the total number of columns required by a null-terminated
 203  * string, assuming that the string is utf8.  Returns strlen() instead
 204  * if the string does not look like a valid utf8 string.
 205  */
 206 int utf8_strnwidth(const char *string, int len, int skip_ansi)
 207 {
 208         int width = 0;
 209         const char *orig = string;
 210
 211         if (len == -1)
 212                 len = strlen(string);
 213         while (string && string < orig + len) {
 214                 int skip;
 215                 while (skip_ansi &&
 216                        (skip = display_mode_esc_sequence_len(string)) != 0)
 217                         string += skip;
 218                 width += utf8_width(&string, NULL);
 219         }
 220         return string ? width : len;
 221 }
 222
 223 int utf8_strwidth(const char *string)
 224 {
 225         return utf8_strnwidth(string, -1, 0);
 226 }
 227
 228 int is_utf8(const char *text)
 229 {
 230         while (*text) {
 231                 if (*text == '\n' || *text == '\t' || *text == '\r') {
 232                         text++;
 233                         continue;
 234                 }
 235                 utf8_width(&text, NULL);
 236                 if (!text)
 237                         return 0;
 238         }
 239         return 1;
 240 }
 241
 242 static void strbuf_addchars(struct strbuf *sb, int c, size_t n)
 243 {
 244         strbuf_grow(sb, n);
 245         memset(sb->buf + sb->len, c, n);
 246         strbuf_setlen(sb, sb->len + n);
 247 }
 248
 249 static void strbuf_add_indented_text(struct strbuf *buf, const char *text,
 250                                      int indent, int indent2)
 251 {
 252         if (indent < 0)
 253                 indent = 0;
 254         while (*text) {
 255                 const char *eol = strchrnul(text, '\n');
 256                 if (*eol == '\n')
 257                         eol++;
 258                 strbuf_addchars(buf, ' ', indent);
 259                 strbuf_add(buf, text, eol - text);
 260                 text = eol;
 261                 indent = indent2;
 262         }
 263 }
 264
 265 /*
 266  * Wrap the text, if necessary. The variable indent is the indent for the
 267  * first line, indent2 is the indent for all other lines.
 268  * If indent is negative, assume that already -indent columns have been
 269  * consumed (and no extra indent is necessary for the first line).
 270  */
 271 void strbuf_add_wrapped_text(struct strbuf *buf,
 272                 const char *text, int indent1, int indent2, int width)
 273 {
 274         int indent, w, assume_utf8 = 1;
 275         const char *bol, *space, *start = text;
 276         size_t orig_len = buf->len;
 277
 278         if (width <= 0) {
 279                 strbuf_add_indented_text(buf, text, indent1, indent2);
 280                 return;
 281         }
 282
 283 retry:
 284         bol = text;
 285         w = indent = indent1;
 286         space = NULL;
 287         if (indent < 0) {
 288                 w = -indent;
 289                 space = text;
 290         }
 291
 292         for (;;) {
 293                 char c;
 294                 size_t skip;
 295
 296                 while ((skip = display_mode_esc_sequence_len(text)))
 297                         text += skip;
 298
 299                 c = *text;
 300                 if (!c || isspace(c)) {
 301                         if (w <= width || !space) {
 302                                 const char *start = bol;
 303                                 if (!c && text == start)
 304                                         return;
 305                                 if (space)
 306                                         start = space;
 307                                 else
 308                                         strbuf_addchars(buf, ' ', indent);
 309                                 strbuf_add(buf, start, text - start);
 310                                 if (!c)
 311                                         return;
 312                                 space = text;
 313                                 if (c == '\t')
 314                                         w |= 0x07;
 315                                 else if (c == '\n') {
 316                                         space++;
 317                                         if (*space == '\n') {
 318                                                 strbuf_addch(buf, '\n');
 319                                                 goto new_line;
 320                                         }
 321                                         else if (!isalnum(*space))
 322                                                 goto new_line;
 323                                         else
 324                                                 strbuf_addch(buf, ' ');
 325                                 }
 326                                 w++;
 327                                 text++;
 328                         }
 329                         else {
 330 new_line:
 331                                 strbuf_addch(buf, '\n');
 332                                 text = bol = space + isspace(*space);
 333                                 space = NULL;
 334                                 w = indent = indent2;
 335                         }
 336                         continue;
 337                 }
 338                 if (assume_utf8) {
 339                         w += utf8_width(&text, NULL);
 340                         if (!text) {
 341                                 assume_utf8 = 0;
 342                                 text = start;
 343                                 strbuf_setlen(buf, orig_len);
 344                                 goto retry;
 345                         }
 346                 } else {
 347                         w++;
 348                         text++;
 349                 }
 350         }
 351 }
 352
 353 void strbuf_add_wrapped_bytes(struct strbuf *buf, const char *data, int len,
 354                              int indent, int indent2, int width)
 355 {
 356         char *tmp = xstrndup(data, len);
 357         strbuf_add_wrapped_text(buf, tmp, indent, indent2, width);
 358         free(tmp);
 359 }
 360
 361 void strbuf_utf8_replace(struct strbuf *sb_src, int pos, int width,
 362                          const char *subst)
 363 {
 364         struct strbuf sb_dst = STRBUF_INIT;
 365         char *src = sb_src->buf;
 366         char *end = src + sb_src->len;
 367         char *dst;
 368         int w = 0, subst_len = 0;
 369
 370         if (subst)
 371                 subst_len = strlen(subst);
 372         strbuf_grow(&sb_dst, sb_src->len + subst_len);
 373         dst = sb_dst.buf;
 374
 375         while (src < end) {
 376                 char *old;
 377                 size_t n;
 378
 379                 while ((n = display_mode_esc_sequence_len(src))) {
 380                         memcpy(dst, src, n);
 381                         src += n;
 382                         dst += n;
 383                 }
 384
 385                 if (src >= end)
 386                         break;
 387
 388                 old = src;
 389                 n = utf8_width((const char**)&src, NULL);
 390                 if (!src)       /* broken utf-8, do nothing */
 391                         return;
 392                 if (n && w >= pos && w < pos + width) {
 393                         if (subst) {
 394                                 memcpy(dst, subst, subst_len);
 395                                 dst += subst_len;
 396                                 subst = NULL;
 397                         }
 398                         w += n;
 399                         continue;
 400                 }
 401                 memcpy(dst, old, src - old);
 402                 dst += src - old;
 403                 w += n;
 404         }
 405         strbuf_setlen(&sb_dst, dst - sb_dst.buf);
 406         strbuf_swap(sb_src, &sb_dst);
 407         strbuf_release(&sb_dst);
 408 }
 409
 410 int is_encoding_utf8(const char *name)
 411 {
 412         if (!name)
 413                 return 1;
 414         if (!strcasecmp(name, "utf-8") || !strcasecmp(name, "utf8"))
 415                 return 1;
 416         return 0;
 417 }
 418
 419 int same_encoding(const char *src, const char *dst)
 420 {
 421         if (is_encoding_utf8(src) && is_encoding_utf8(dst))
 422                 return 1;
 423         return !strcasecmp(src, dst);
 424 }
 425
 426 /*
 427  * Wrapper for fprintf and returns the total number of columns required
 428  * for the printed string, assuming that the string is utf8.
 429  */
 430 int utf8_fprintf(FILE *stream, const char *format, ...)
 431 {
 432         struct strbuf buf = STRBUF_INIT;
 433         va_list arg;
 434         int columns;
 435
 436         va_start(arg, format);
 437         strbuf_vaddf(&buf, format, arg);
 438         va_end(arg);
 439
 440         columns = fputs(buf.buf, stream);
 441         if (0 <= columns) /* keep the error from the I/O */
 442                 columns = utf8_strwidth(buf.buf);
 443         strbuf_release(&buf);
 444         return columns;
 445 }
 446
 447 /*
 448  * Given a buffer and its encoding, return it re-encoded
 449  * with iconv.  If the conversion fails, returns NULL.
 450  */
 451 #ifndef NO_ICONV
 452 #if defined(OLD_ICONV) || (defined(__sun__) && !defined(_XPG6))
 453         typedef const char * iconv_ibp;
 454 #else
 455         typedef char * iconv_ibp;
 456 #endif
 457 char *reencode_string_iconv(const char *in, size_t insz, iconv_t conv, int *outsz_p)
 458 {
 459         size_t outsz, outalloc;
 460         char *out, *outpos;
 461         iconv_ibp cp;
 462
 463         outsz = insz;
 464         outalloc = outsz + 1; /* for terminating NUL */
 465         out = xmalloc(outalloc);
 466         outpos = out;
 467         cp = (iconv_ibp)in;
 468
 469         while (1) {
 470                 size_t cnt = iconv(conv, &cp, &insz, &outpos, &outsz);
 471
 472                 if (cnt == (size_t) -1) {
 473                         size_t sofar;
 474                         if (errno != E2BIG) {
 475                                 free(out);
 476                                 return NULL;
 477                         }
 478                         /* insz has remaining number of bytes.
 479                          * since we started outsz the same as insz,
 480                          * it is likely that insz is not enough for
 481                          * converting the rest.
 482                          */
 483                         sofar = outpos - out;
 484                         outalloc = sofar + insz * 2 + 32;
 485                         out = xrealloc(out, outalloc);
 486                         outpos = out + sofar;
 487                         outsz = outalloc - sofar - 1;
 488                 }
 489                 else {
 490                         *outpos = '\0';
 491                         if (outsz_p)
 492                                 *outsz_p = outpos - out;
 493                         break;
 494                 }
 495         }
 496         return out;
 497 }
 498
 499 char *reencode_string_len(const char *in, int insz,
 500                           const char *out_encoding, const char *in_encoding,
 501                           int *outsz)
 502 {
 503         iconv_t conv;
 504         char *out;
 505
 506         if (!in_encoding)
 507                 return NULL;
 508
 509         conv = iconv_open(out_encoding, in_encoding);
 510         if (conv == (iconv_t) -1) {
 511                 /*
 512                  * Some platforms do not have the variously spelled variants of
 513                  * UTF-8, so let's fall back to trying the most official
 514                  * spelling. We do so only as a fallback in case the platform
 515                  * does understand the user's spelling, but not our official
 516                  * one.
 517                  */
 518                 if (is_encoding_utf8(in_encoding))
 519                         in_encoding = "UTF-8";
 520                 if (is_encoding_utf8(out_encoding))
 521                         out_encoding = "UTF-8";
 522                 conv = iconv_open(out_encoding, in_encoding);
 523                 if (conv == (iconv_t) -1)
 524                         return NULL;
 525         }
 526
 527         out = reencode_string_iconv(in, insz, conv, outsz);
 528         iconv_close(conv);
 529         return out;
 530 }
 531 #endif
 532
 533 /*
 534  * Returns first character length in bytes for multi-byte `text` according to
 535  * `encoding`.
 536  *
 537  * - The `text` pointer is updated to point at the next character.
 538  * - When `remainder_p` is not NULL, on entry `*remainder_p` is how much bytes
 539  *   we can consume from text, and on exit `*remainder_p` is reduced by returned
 540  *   character length. Otherwise `text` is treated as limited by NUL.
 541  */
 542 int mbs_chrlen(const char **text, size_t *remainder_p, const char *encoding)
 543 {
 544         int chrlen;
 545         const char *p = *text;
 546         size_t r = (remainder_p ? *remainder_p : SIZE_MAX);
 547
 548         if (r < 1)
 549                 return 0;
 550
 551         if (is_encoding_utf8(encoding)) {
 552                 pick_one_utf8_char(&p, &r);
 553
 554                 chrlen = p ? (p - *text)
 555                            : 1 /* not valid UTF-8 -> raw byte sequence */;
 556         }
 557         else {
 558                 /*
 559                  * TODO use iconv to decode one char and obtain its chrlen
 560                  * for now, let's treat encodings != UTF-8 as one-byte
 561                  */
 562                 chrlen = 1;
 563         }
 564
 565         *text += chrlen;
 566         if (remainder_p)
 567                 *remainder_p -= chrlen;
 568
 569         return chrlen;
 570 }