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_add_indented_text(struct strbuf *buf, const char *text,
 243                                      int indent, int indent2)
 244 {
 245         if (indent < 0)
 246                 indent = 0;
 247         while (*text) {
 248                 const char *eol = strchrnul(text, '\n');
 249                 if (*eol == '\n')
 250                         eol++;
 251                 strbuf_addchars(buf, ' ', indent);
 252                 strbuf_add(buf, text, eol - text);
 253                 text = eol;
 254                 indent = indent2;
 255         }
 256 }
 257
 258 /*
 259  * Wrap the text, if necessary. The variable indent is the indent for the
 260  * first line, indent2 is the indent for all other lines.
 261  * If indent is negative, assume that already -indent columns have been
 262  * consumed (and no extra indent is necessary for the first line).
 263  */
 264 void strbuf_add_wrapped_text(struct strbuf *buf,
 265                 const char *text, int indent1, int indent2, int width)
 266 {
 267         int indent, w, assume_utf8 = 1;
 268         const char *bol, *space, *start = text;
 269         size_t orig_len = buf->len;
 270
 271         if (width <= 0) {
 272                 strbuf_add_indented_text(buf, text, indent1, indent2);
 273                 return;
 274         }
 275
 276 retry:
 277         bol = text;
 278         w = indent = indent1;
 279         space = NULL;
 280         if (indent < 0) {
 281                 w = -indent;
 282                 space = text;
 283         }
 284
 285         for (;;) {
 286                 char c;
 287                 size_t skip;
 288
 289                 while ((skip = display_mode_esc_sequence_len(text)))
 290                         text += skip;
 291
 292                 c = *text;
 293                 if (!c || isspace(c)) {
 294                         if (w <= width || !space) {
 295                                 const char *start = bol;
 296                                 if (!c && text == start)
 297                                         return;
 298                                 if (space)
 299                                         start = space;
 300                                 else
 301                                         strbuf_addchars(buf, ' ', indent);
 302                                 strbuf_add(buf, start, text - start);
 303                                 if (!c)
 304                                         return;
 305                                 space = text;
 306                                 if (c == '\t')
 307                                         w |= 0x07;
 308                                 else if (c == '\n') {
 309                                         space++;
 310                                         if (*space == '\n') {
 311                                                 strbuf_addch(buf, '\n');
 312                                                 goto new_line;
 313                                         }
 314                                         else if (!isalnum(*space))
 315                                                 goto new_line;
 316                                         else
 317                                                 strbuf_addch(buf, ' ');
 318                                 }
 319                                 w++;
 320                                 text++;
 321                         }
 322                         else {
 323 new_line:
 324                                 strbuf_addch(buf, '\n');
 325                                 text = bol = space + isspace(*space);
 326                                 space = NULL;
 327                                 w = indent = indent2;
 328                         }
 329                         continue;
 330                 }
 331                 if (assume_utf8) {
 332                         w += utf8_width(&text, NULL);
 333                         if (!text) {
 334                                 assume_utf8 = 0;
 335                                 text = start;
 336                                 strbuf_setlen(buf, orig_len);
 337                                 goto retry;
 338                         }
 339                 } else {
 340                         w++;
 341                         text++;
 342                 }
 343         }
 344 }
 345
 346 void strbuf_add_wrapped_bytes(struct strbuf *buf, const char *data, int len,
 347                              int indent, int indent2, int width)
 348 {
 349         char *tmp = xstrndup(data, len);
 350         strbuf_add_wrapped_text(buf, tmp, indent, indent2, width);
 351         free(tmp);
 352 }
 353
 354 void strbuf_utf8_replace(struct strbuf *sb_src, int pos, int width,
 355                          const char *subst)
 356 {
 357         struct strbuf sb_dst = STRBUF_INIT;
 358         char *src = sb_src->buf;
 359         char *end = src + sb_src->len;
 360         char *dst;
 361         int w = 0, subst_len = 0;
 362
 363         if (subst)
 364                 subst_len = strlen(subst);
 365         strbuf_grow(&sb_dst, sb_src->len + subst_len);
 366         dst = sb_dst.buf;
 367
 368         while (src < end) {
 369                 char *old;
 370                 size_t n;
 371
 372                 while ((n = display_mode_esc_sequence_len(src))) {
 373                         memcpy(dst, src, n);
 374                         src += n;
 375                         dst += n;
 376                 }
 377
 378                 if (src >= end)
 379                         break;
 380
 381                 old = src;
 382                 n = utf8_width((const char**)&src, NULL);
 383                 if (!src)       /* broken utf-8, do nothing */
 384                         return;
 385                 if (n && w >= pos && w < pos + width) {
 386                         if (subst) {
 387                                 memcpy(dst, subst, subst_len);
 388                                 dst += subst_len;
 389                                 subst = NULL;
 390                         }
 391                         w += n;
 392                         continue;
 393                 }
 394                 memcpy(dst, old, src - old);
 395                 dst += src - old;
 396                 w += n;
 397         }
 398         strbuf_setlen(&sb_dst, dst - sb_dst.buf);
 399         strbuf_swap(sb_src, &sb_dst);
 400         strbuf_release(&sb_dst);
 401 }
 402
 403 int is_encoding_utf8(const char *name)
 404 {
 405         if (!name)
 406                 return 1;
 407         if (!strcasecmp(name, "utf-8") || !strcasecmp(name, "utf8"))
 408                 return 1;
 409         return 0;
 410 }
 411
 412 int same_encoding(const char *src, const char *dst)
 413 {
 414         if (is_encoding_utf8(src) && is_encoding_utf8(dst))
 415                 return 1;
 416         return !strcasecmp(src, dst);
 417 }
 418
 419 /*
 420  * Wrapper for fprintf and returns the total number of columns required
 421  * for the printed string, assuming that the string is utf8.
 422  */
 423 int utf8_fprintf(FILE *stream, const char *format, ...)
 424 {
 425         struct strbuf buf = STRBUF_INIT;
 426         va_list arg;
 427         int columns;
 428
 429         va_start(arg, format);
 430         strbuf_vaddf(&buf, format, arg);
 431         va_end(arg);
 432
 433         columns = fputs(buf.buf, stream);
 434         if (0 <= columns) /* keep the error from the I/O */
 435                 columns = utf8_strwidth(buf.buf);
 436         strbuf_release(&buf);
 437         return columns;
 438 }
 439
 440 /*
 441  * Given a buffer and its encoding, return it re-encoded
 442  * with iconv.  If the conversion fails, returns NULL.
 443  */
 444 #ifndef NO_ICONV
 445 #if defined(OLD_ICONV) || (defined(__sun__) && !defined(_XPG6))
 446         typedef const char * iconv_ibp;
 447 #else
 448         typedef char * iconv_ibp;
 449 #endif
 450 char *reencode_string_iconv(const char *in, size_t insz, iconv_t conv, int *outsz_p)
 451 {
 452         size_t outsz, outalloc;
 453         char *out, *outpos;
 454         iconv_ibp cp;
 455
 456         outsz = insz;
 457         outalloc = outsz + 1; /* for terminating NUL */
 458         out = xmalloc(outalloc);
 459         outpos = out;
 460         cp = (iconv_ibp)in;
 461
 462         while (1) {
 463                 size_t cnt = iconv(conv, &cp, &insz, &outpos, &outsz);
 464
 465                 if (cnt == (size_t) -1) {
 466                         size_t sofar;
 467                         if (errno != E2BIG) {
 468                                 free(out);
 469                                 return NULL;
 470                         }
 471                         /* insz has remaining number of bytes.
 472                          * since we started outsz the same as insz,
 473                          * it is likely that insz is not enough for
 474                          * converting the rest.
 475                          */
 476                         sofar = outpos - out;
 477                         outalloc = sofar + insz * 2 + 32;
 478                         out = xrealloc(out, outalloc);
 479                         outpos = out + sofar;
 480                         outsz = outalloc - sofar - 1;
 481                 }
 482                 else {
 483                         *outpos = '\0';
 484                         if (outsz_p)
 485                                 *outsz_p = outpos - out;
 486                         break;
 487                 }
 488         }
 489         return out;
 490 }
 491
 492 static const char *fallback_encoding(const char *name)
 493 {
 494         /*
 495          * Some platforms do not have the variously spelled variants of
 496          * UTF-8, so let's fall back to trying the most official
 497          * spelling. We do so only as a fallback in case the platform
 498          * does understand the user's spelling, but not our official
 499          * one.
 500          */
 501         if (is_encoding_utf8(name))
 502                 return "UTF-8";
 503
 504         /*
 505          * Even though latin-1 is still seen in e-mail
 506          * headers, some platforms only install ISO-8859-1.
 507          */
 508         if (!strcasecmp(name, "latin-1"))
 509                 return "ISO-8859-1";
 510
 511         return name;
 512 }
 513
 514 char *reencode_string_len(const char *in, int insz,
 515                           const char *out_encoding, const char *in_encoding,
 516                           int *outsz)
 517 {
 518         iconv_t conv;
 519         char *out;
 520
 521         if (!in_encoding)
 522                 return NULL;
 523
 524         conv = iconv_open(out_encoding, in_encoding);
 525         if (conv == (iconv_t) -1) {
 526                 in_encoding = fallback_encoding(in_encoding);
 527                 out_encoding = fallback_encoding(out_encoding);
 528
 529                 conv = iconv_open(out_encoding, in_encoding);
 530                 if (conv == (iconv_t) -1)
 531                         return NULL;
 532         }
 533
 534         out = reencode_string_iconv(in, insz, conv, outsz);
 535         iconv_close(conv);
 536         return out;
 537 }
 538 #endif
 539
 540 /*
 541  * Returns first character length in bytes for multi-byte `text` according to
 542  * `encoding`.
 543  *
 544  * - The `text` pointer is updated to point at the next character.
 545  * - When `remainder_p` is not NULL, on entry `*remainder_p` is how much bytes
 546  *   we can consume from text, and on exit `*remainder_p` is reduced by returned
 547  *   character length. Otherwise `text` is treated as limited by NUL.
 548  */
 549 int mbs_chrlen(const char **text, size_t *remainder_p, const char *encoding)
 550 {
 551         int chrlen;
 552         const char *p = *text;
 553         size_t r = (remainder_p ? *remainder_p : SIZE_MAX);
 554
 555         if (r < 1)
 556                 return 0;
 557
 558         if (is_encoding_utf8(encoding)) {
 559                 pick_one_utf8_char(&p, &r);
 560
 561                 chrlen = p ? (p - *text)
 562                            : 1 /* not valid UTF-8 -> raw byte sequence */;
 563         }
 564         else {
 565                 /*
 566                  * TODO use iconv to decode one char and obtain its chrlen
 567                  * for now, let's treat encodings != UTF-8 as one-byte
 568                  */
 569                 chrlen = 1;
 570         }
 571
 572         *text += chrlen;
 573         if (remainder_p)
 574                 *remainder_p -= chrlen;
 575
 576         return chrlen;
 577 }
 578
 579 /*
 580  * Pick the next char from the stream, ignoring codepoints an HFS+ would.
 581  * Note that this is _not_ complete by any means. It's just enough
 582  * to make is_hfs_dotgit() work, and should not be used otherwise.
 583  */
 584 static ucs_char_t next_hfs_char(const char **in)
 585 {
 586         while (1) {
 587                 ucs_char_t out = pick_one_utf8_char(in, NULL);
 588                 /*
 589                  * check for malformed utf8. Technically this
 590                  * gets converted to a percent-sequence, but
 591                  * returning 0 is good enough for is_hfs_dotgit
 592                  * to realize it cannot be .git
 593                  */
 594                 if (!*in)
 595                         return 0;
 596
 597                 /* these code points are ignored completely */
 598                 switch (out) {
 599                 case 0x200c: /* ZERO WIDTH NON-JOINER */
 600                 case 0x200d: /* ZERO WIDTH JOINER */
 601                 case 0x200e: /* LEFT-TO-RIGHT MARK */
 602                 case 0x200f: /* RIGHT-TO-LEFT MARK */
 603                 case 0x202a: /* LEFT-TO-RIGHT EMBEDDING */
 604                 case 0x202b: /* RIGHT-TO-LEFT EMBEDDING */
 605                 case 0x202c: /* POP DIRECTIONAL FORMATTING */
 606                 case 0x202d: /* LEFT-TO-RIGHT OVERRIDE */
 607                 case 0x202e: /* RIGHT-TO-LEFT OVERRIDE */
 608                 case 0x206a: /* INHIBIT SYMMETRIC SWAPPING */
 609                 case 0x206b: /* ACTIVATE SYMMETRIC SWAPPING */
 610                 case 0x206c: /* INHIBIT ARABIC FORM SHAPING */
 611                 case 0x206d: /* ACTIVATE ARABIC FORM SHAPING */
 612                 case 0x206e: /* NATIONAL DIGIT SHAPES */
 613                 case 0x206f: /* NOMINAL DIGIT SHAPES */
 614                 case 0xfeff: /* ZERO WIDTH NO-BREAK SPACE */
 615                         continue;
 616                 }
 617
 618                 return out;
 619         }
 620 }
 621
 622 int is_hfs_dotgit(const char *path)
 623 {
 624         ucs_char_t c;
 625
 626         c = next_hfs_char(&path);
 627         if (c != '.')
 628                 return 0;
 629         c = next_hfs_char(&path);
 630
 631         /*
 632          * there's a great deal of other case-folding that occurs
 633          * in HFS+, but this is enough to catch anything that will
 634          * convert to ".git"
 635          */
 636         if (c != 'g' && c != 'G')
 637                 return 0;
 638         c = next_hfs_char(&path);
 639         if (c != 'i' && c != 'I')
 640                 return 0;
 641         c = next_hfs_char(&path);
 642         if (c != 't' && c != 'T')
 643                 return 0;
 644         c = next_hfs_char(&path);
 645         if (c && !is_dir_sep(c))
 646                 return 0;
 647
 648         return 1;
 649 }
 650
 651 const char utf8_bom[] = "\357\273\277";
 652
 653 int skip_utf8_bom(char **text, size_t len)
 654 {
 655         if (len < strlen(utf8_bom) ||
 656             memcmp(*text, utf8_bom, strlen(utf8_bom)))
 657                 return 0;
 658         *text += strlen(utf8_bom);
 659         return 1;
 660 }
 661
 662 void strbuf_utf8_align(struct strbuf *buf, align_type position, unsigned int width,
 663                        const char *s)
 664 {
 665         int slen = strlen(s);
 666         int display_len = utf8_strnwidth(s, slen, 0);
 667         int utf8_compensation = slen - display_len;
 668
 669         if (display_len >= width) {
 670                 strbuf_addstr(buf, s);
 671                 return;
 672         }
 673
 674         if (position == ALIGN_LEFT)
 675                 strbuf_addf(buf, "%-*s", width + utf8_compensation, s);
 676         else if (position == ALIGN_MIDDLE) {
 677                 int left = (width - display_len) / 2;
 678                 strbuf_addf(buf, "%*s%-*s", left, "", width - left + utf8_compensation, s);
 679         } else if (position == ALIGN_RIGHT)
 680                 strbuf_addf(buf, "%*s", width + utf8_compensation, s);
 681 }