glib/gunidecomp.c

   1 /* decomp.c - Character decomposition.
   2  *
   3  *  Copyright (C) 1999, 2000 Tom Tromey
   4  *  Copyright 2000 Red Hat, Inc.
   5  *
   6  * This library is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2 of the License, or (at your option) any later version.
  10  *
  11  * This library is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public License
  17  * along with this library; if not, see <http://www.gnu.org/licenses/>.
  18  */
  19
  20 /**
  21  * SECTION:unicode
  22  * @Title: Unicode Manipulation
  23  * @Short_description: functions operating on Unicode characters and
  24  *     UTF-8 strings
  25  * @See_also: g_locale_to_utf8(), g_locale_from_utf8()
  26  *
  27  * This section describes a number of functions for dealing with
  28  * Unicode characters and strings. There are analogues of the
  29  * traditional `ctype.h` character classification and case conversion
  30  * functions, UTF-8 analogues of some string utility functions,
  31  * functions to perform normalization, case conversion and collation
  32  * on UTF-8 strings and finally functions to convert between the UTF-8,
  33  * UTF-16 and UCS-4 encodings of Unicode.
  34  *
  35  * The implementations of the Unicode functions in GLib are based
  36  * on the Unicode Character Data tables, which are available from
  37  * [www.unicode.org](http://www.unicode.org/).
  38  * GLib 2.8 supports Unicode 4.0, GLib 2.10 supports Unicode 4.1,
  39  * GLib 2.12 supports Unicode 5.0, GLib 2.16.3 supports Unicode 5.1,
  40  * GLib 2.30 supports Unicode 6.0.
  41  */
  42
  43 #include "config.h"
  44
  45 #include <stdlib.h>
  46
  47 #include "gunicode.h"
  48 #include "gunidecomp.h"
  49 #include "gmem.h"
  50 #include "gunicomp.h"
  51 #include "gunicodeprivate.h"
  52
  53
  54 #define CC_PART1(Page, Char) \
  55   ((combining_class_table_part1[Page] >= G_UNICODE_MAX_TABLE_INDEX) \
  56    ? (combining_class_table_part1[Page] - G_UNICODE_MAX_TABLE_INDEX) \
  57    : (cclass_data[combining_class_table_part1[Page]][Char]))
  58
  59 #define CC_PART2(Page, Char) \
  60   ((combining_class_table_part2[Page] >= G_UNICODE_MAX_TABLE_INDEX) \
  61    ? (combining_class_table_part2[Page] - G_UNICODE_MAX_TABLE_INDEX) \
  62    : (cclass_data[combining_class_table_part2[Page]][Char]))
  63
  64 #define COMBINING_CLASS(Char) \
  65   (((Char) <= G_UNICODE_LAST_CHAR_PART1) \
  66    ? CC_PART1 ((Char) >> 8, (Char) & 0xff) \
  67    : (((Char) >= 0xe0000 && (Char) <= G_UNICODE_LAST_CHAR) \
  68       ? CC_PART2 (((Char) - 0xe0000) >> 8, (Char) & 0xff) \
  69       : 0))
  70
  71 /**
  72  * g_unichar_combining_class:
  73  * @uc: a Unicode character
  74  *
  75  * Determines the canonical combining class of a Unicode character.
  76  *
  77  * Returns: the combining class of the character
  78  *
  79  * Since: 2.14
  80  **/
  81 gint
  82 g_unichar_combining_class (gunichar uc)
  83 {
  84   return COMBINING_CLASS (uc);
  85 }
  86
  87 /* constants for hangul syllable [de]composition */
  88 #define SBase 0xAC00
  89 #define LBase 0x1100
  90 #define VBase 0x1161
  91 #define TBase 0x11A7
  92 #define LCount 19
  93 #define VCount 21
  94 #define TCount 28
  95 #define NCount (VCount * TCount)
  96 #define SCount (LCount * NCount)
  97
  98 /**
  99  * g_unicode_canonical_ordering:
 100  * @string: a UCS-4 encoded string.
 101  * @len: the maximum length of @string to use.
 102  *
 103  * Computes the canonical ordering of a string in-place.
 104  * This rearranges decomposed characters in the string
 105  * according to their combining classes.  See the Unicode
 106  * manual for more information.
 107  **/
 108 void
 109 g_unicode_canonical_ordering (gunichar *string,
 110                               gsize     len)
 111 {
 112   gsize i;
 113   int swap = 1;
 114
 115   while (swap)
 116     {
 117       int last;
 118       swap = 0;
 119       last = COMBINING_CLASS (string[0]);
 120       for (i = 0; i < len - 1; ++i)
 121         {
 122           int next = COMBINING_CLASS (string[i + 1]);
 123           if (next != 0 && last > next)
 124             {
 125               gsize j;
 126               /* Percolate item leftward through string.  */
 127               for (j = i + 1; j > 0; --j)
 128                 {
 129                   gunichar t;
 130                   if (COMBINING_CLASS (string[j - 1]) <= next)
 131                     break;
 132                   t = string[j];
 133                   string[j] = string[j - 1];
 134                   string[j - 1] = t;
 135                   swap = 1;
 136                 }
 137               /* We're re-entering the loop looking at the old
 138                  character again.  */
 139               next = last;
 140             }
 141           last = next;
 142         }
 143     }
 144 }
 145
 146 /* http://www.unicode.org/unicode/reports/tr15/#Hangul
 147  * r should be null or have sufficient space. Calling with r == NULL will
 148  * only calculate the result_len; however, a buffer with space for three
 149  * characters will always be big enough. */
 150 static void
 151 decompose_hangul (gunichar s,
 152                   gunichar *r,
 153                   gsize *result_len)
 154 {
 155   gint SIndex = s - SBase;
 156   gint TIndex = SIndex % TCount;
 157
 158   if (r)
 159     {
 160       r[0] = LBase + SIndex / NCount;
 161       r[1] = VBase + (SIndex % NCount) / TCount;
 162     }
 163
 164   if (TIndex)
 165     {
 166       if (r)
 167         r[2] = TBase + TIndex;
 168       *result_len = 3;
 169     }
 170   else
 171     *result_len = 2;
 172 }
 173
 174 /* returns a pointer to a null-terminated UTF-8 string */
 175 static const gchar *
 176 find_decomposition (gunichar ch,
 177                     gboolean compat)
 178 {
 179   int start = 0;
 180   int end = G_N_ELEMENTS (decomp_table);
 181
 182   if (ch >= decomp_table[start].ch &&
 183       ch <= decomp_table[end - 1].ch)
 184     {
 185       while (TRUE)
 186         {
 187           int half = (start + end) / 2;
 188           if (ch == decomp_table[half].ch)
 189             {
 190               int offset;
 191
 192               if (compat)
 193                 {
 194                   offset = decomp_table[half].compat_offset;
 195                   if (offset == G_UNICODE_NOT_PRESENT_OFFSET)
 196                     offset = decomp_table[half].canon_offset;
 197                 }
 198               else
 199                 {
 200                   offset = decomp_table[half].canon_offset;
 201                   if (offset == G_UNICODE_NOT_PRESENT_OFFSET)
 202                     return NULL;
 203                 }
 204
 205               return &(decomp_expansion_string[offset]);
 206             }
 207           else if (half == start)
 208             break;
 209           else if (ch > decomp_table[half].ch)
 210             start = half;
 211           else
 212             end = half;
 213         }
 214     }
 215
 216   return NULL;
 217 }
 218
 219 /**
 220  * g_unicode_canonical_decomposition:
 221  * @ch: a Unicode character.
 222  * @result_len: location to store the length of the return value.
 223  *
 224  * Computes the canonical decomposition of a Unicode character.
 225  *
 226  * Returns: a newly allocated string of Unicode characters.
 227  *   @result_len is set to the resulting length of the string.
 228  *
 229  * Deprecated: 2.30: Use the more flexible g_unichar_fully_decompose()
 230  *   instead.
 231  **/
 232 gunichar *
 233 g_unicode_canonical_decomposition (gunichar ch,
 234                                    gsize   *result_len)
 235 {
 236   const gchar *decomp;
 237   const gchar *p;
 238   gunichar *r;
 239
 240   /* Hangul syllable */
 241   if (ch >= SBase && ch < SBase + SCount)
 242     {
 243       decompose_hangul (ch, NULL, result_len);
 244       r = g_malloc (*result_len * sizeof (gunichar));
 245       decompose_hangul (ch, r, result_len);
 246     }
 247   else if ((decomp = find_decomposition (ch, FALSE)) != NULL)
 248     {
 249       /* Found it.  */
 250       int i;
 251
 252       *result_len = g_utf8_strlen (decomp, -1);
 253       r = g_malloc (*result_len * sizeof (gunichar));
 254
 255       for (p = decomp, i = 0; *p != '\0'; p = g_utf8_next_char (p), i++)
 256         r[i] = g_utf8_get_char (p);
 257     }
 258   else
 259     {
 260       /* Not in our table.  */
 261       r = g_malloc (sizeof (gunichar));
 262       *r = ch;
 263       *result_len = 1;
 264     }
 265
 266   return r;
 267 }
 268
 269 /* L,V => LV and LV,T => LVT  */
 270 static gboolean
 271 combine_hangul (gunichar a,
 272                 gunichar b,
 273                 gunichar *result)
 274 {
 275   gint LIndex = a - LBase;
 276   gint SIndex = a - SBase;
 277
 278   gint VIndex = b - VBase;
 279   gint TIndex = b - TBase;
 280
 281   if (0 <= LIndex && LIndex < LCount
 282       && 0 <= VIndex && VIndex < VCount)
 283     {
 284       *result = SBase + (LIndex * VCount + VIndex) * TCount;
 285       return TRUE;
 286     }
 287   else if (0 <= SIndex && SIndex < SCount && (SIndex % TCount) == 0
 288            && 0 < TIndex && TIndex < TCount)
 289     {
 290       *result = a + TIndex;
 291       return TRUE;
 292     }
 293
 294   return FALSE;
 295 }
 296
 297 #define CI(Page, Char) \
 298   ((compose_table[Page] >= G_UNICODE_MAX_TABLE_INDEX) \
 299    ? (compose_table[Page] - G_UNICODE_MAX_TABLE_INDEX) \
 300    : (compose_data[compose_table[Page]][Char]))
 301
 302 #define COMPOSE_INDEX(Char) \
 303      (((Char >> 8) > (COMPOSE_TABLE_LAST)) ? 0 : CI((Char) >> 8, (Char) & 0xff))
 304
 305 static gboolean
 306 combine (gunichar  a,
 307          gunichar  b,
 308          gunichar *result)
 309 {
 310   gushort index_a, index_b;
 311
 312   if (combine_hangul (a, b, result))
 313     return TRUE;
 314
 315   index_a = COMPOSE_INDEX(a);
 316
 317   if (index_a >= COMPOSE_FIRST_SINGLE_START && index_a < COMPOSE_SECOND_START)
 318     {
 319       if (b == compose_first_single[index_a - COMPOSE_FIRST_SINGLE_START][0])
 320         {
 321           *result = compose_first_single[index_a - COMPOSE_FIRST_SINGLE_START][1];
 322           return TRUE;
 323         }
 324       else
 325         return FALSE;
 326     }
 327
 328   index_b = COMPOSE_INDEX(b);
 329
 330   if (index_b >= COMPOSE_SECOND_SINGLE_START)
 331     {
 332       if (a == compose_second_single[index_b - COMPOSE_SECOND_SINGLE_START][0])
 333         {
 334           *result = compose_second_single[index_b - COMPOSE_SECOND_SINGLE_START][1];
 335           return TRUE;
 336         }
 337       else
 338         return FALSE;
 339     }
 340
 341   if (index_a >= COMPOSE_FIRST_START && index_a < COMPOSE_FIRST_SINGLE_START &&
 342       index_b >= COMPOSE_SECOND_START && index_b < COMPOSE_SECOND_SINGLE_START)
 343     {
 344       gunichar res = compose_array[index_a - COMPOSE_FIRST_START][index_b - COMPOSE_SECOND_START];
 345
 346       if (res)
 347         {
 348           *result = res;
 349           return TRUE;
 350         }
 351     }
 352
 353   return FALSE;
 354 }
 355
 356 gunichar *
 357 _g_utf8_normalize_wc (const gchar    *str,
 358                       gssize          max_len,
 359                       GNormalizeMode  mode)
 360 {
 361   gsize n_wc;
 362   gunichar *wc_buffer;
 363   const char *p;
 364   gsize last_start;
 365   gboolean do_compat = (mode == G_NORMALIZE_NFKC ||
 366                         mode == G_NORMALIZE_NFKD);
 367   gboolean do_compose = (mode == G_NORMALIZE_NFC ||
 368                          mode == G_NORMALIZE_NFKC);
 369
 370   n_wc = 0;
 371   p = str;
 372   while ((max_len < 0 || p < str + max_len) && *p)
 373     {
 374       const gchar *decomp;
 375       gunichar wc = g_utf8_get_char (p);
 376
 377       if (wc >= SBase && wc < SBase + SCount)
 378         {
 379           gsize result_len;
 380           decompose_hangul (wc, NULL, &result_len);
 381           n_wc += result_len;
 382         }
 383       else
 384         {
 385           decomp = find_decomposition (wc, do_compat);
 386
 387           if (decomp)
 388             n_wc += g_utf8_strlen (decomp, -1);
 389           else
 390             n_wc++;
 391         }
 392
 393       p = g_utf8_next_char (p);
 394     }
 395
 396   wc_buffer = g_new (gunichar, n_wc + 1);
 397
 398   last_start = 0;
 399   n_wc = 0;
 400   p = str;
 401   while ((max_len < 0 || p < str + max_len) && *p)
 402     {
 403       gunichar wc = g_utf8_get_char (p);
 404       const gchar *decomp;
 405       int cc;
 406       gsize old_n_wc = n_wc;
 407
 408       if (wc >= SBase && wc < SBase + SCount)
 409         {
 410           gsize result_len;
 411           decompose_hangul (wc, wc_buffer + n_wc, &result_len);
 412           n_wc += result_len;
 413         }
 414       else
 415         {
 416           decomp = find_decomposition (wc, do_compat);
 417
 418           if (decomp)
 419             {
 420               const char *pd;
 421               for (pd = decomp; *pd != '\0'; pd = g_utf8_next_char (pd))
 422                 wc_buffer[n_wc++] = g_utf8_get_char (pd);
 423             }
 424           else
 425             wc_buffer[n_wc++] = wc;
 426         }
 427
 428       if (n_wc > 0)
 429         {
 430           cc = COMBINING_CLASS (wc_buffer[old_n_wc]);
 431
 432           if (cc == 0)
 433             {
 434               g_unicode_canonical_ordering (wc_buffer + last_start, n_wc - last_start);
 435               last_start = old_n_wc;
 436             }
 437         }
 438
 439       p = g_utf8_next_char (p);
 440     }
 441
 442   if (n_wc > 0)
 443     {
 444       g_unicode_canonical_ordering (wc_buffer + last_start, n_wc - last_start);
 445       last_start = n_wc;
 446     }
 447
 448   wc_buffer[n_wc] = 0;
 449
 450   /* All decomposed and reordered */
 451
 452   if (do_compose && n_wc > 0)
 453     {
 454       gsize i, j;
 455       int last_cc = 0;
 456       last_start = 0;
 457
 458       for (i = 0; i < n_wc; i++)
 459         {
 460           int cc = COMBINING_CLASS (wc_buffer[i]);
 461
 462           if (i > 0 &&
 463               (last_cc == 0 || last_cc < cc) &&
 464               combine (wc_buffer[last_start], wc_buffer[i],
 465                        &wc_buffer[last_start]))
 466             {
 467               for (j = i + 1; j < n_wc; j++)
 468                 wc_buffer[j-1] = wc_buffer[j];
 469               n_wc--;
 470               i--;
 471
 472               if (i == last_start)
 473                 last_cc = 0;
 474               else
 475                 last_cc = COMBINING_CLASS (wc_buffer[i-1]);
 476
 477               continue;
 478             }
 479
 480           if (cc == 0)
 481             last_start = i;
 482
 483           last_cc = cc;
 484         }
 485     }
 486
 487   wc_buffer[n_wc] = 0;
 488
 489   return wc_buffer;
 490 }
 491
 492 /**
 493  * g_utf8_normalize:
 494  * @str: a UTF-8 encoded string.
 495  * @len: length of @str, in bytes, or -1 if @str is nul-terminated.
 496  * @mode: the type of normalization to perform.
 497  *
 498  * Converts a string into canonical form, standardizing
 499  * such issues as whether a character with an accent
 500  * is represented as a base character and combining
 501  * accent or as a single precomposed character. The
 502  * string has to be valid UTF-8, otherwise %NULL is
 503  * returned. You should generally call g_utf8_normalize()
 504  * before comparing two Unicode strings.
 505  *
 506  * The normalization mode %G_NORMALIZE_DEFAULT only
 507  * standardizes differences that do not affect the
 508  * text content, such as the above-mentioned accent
 509  * representation. %G_NORMALIZE_ALL also standardizes
 510  * the "compatibility" characters in Unicode, such
 511  * as SUPERSCRIPT THREE to the standard forms
 512  * (in this case DIGIT THREE). Formatting information
 513  * may be lost but for most text operations such
 514  * characters should be considered the same.
 515  *
 516  * %G_NORMALIZE_DEFAULT_COMPOSE and %G_NORMALIZE_ALL_COMPOSE
 517  * are like %G_NORMALIZE_DEFAULT and %G_NORMALIZE_ALL,
 518  * but returned a result with composed forms rather
 519  * than a maximally decomposed form. This is often
 520  * useful if you intend to convert the string to
 521  * a legacy encoding or pass it to a system with
 522  * less capable Unicode handling.
 523  *
 524  * Returns: a newly allocated string, that is the
 525  *   normalized form of @str, or %NULL if @str is not
 526  *   valid UTF-8.
 527  **/
 528 gchar *
 529 g_utf8_normalize (const gchar    *str,
 530                   gssize          len,
 531                   GNormalizeMode  mode)
 532 {
 533   gunichar *result_wc = _g_utf8_normalize_wc (str, len, mode);
 534   gchar *result;
 535
 536   result = g_ucs4_to_utf8 (result_wc, -1, NULL, NULL, NULL);
 537   g_free (result_wc);
 538
 539   return result;
 540 }
 541
 542 static gboolean
 543 decompose_hangul_step (gunichar  ch,
 544                        gunichar *a,
 545                        gunichar *b)
 546 {
 547   gint SIndex, TIndex;
 548
 549   if (ch < SBase || ch >= SBase + SCount)
 550     return FALSE;  /* not a hangul syllable */
 551
 552   SIndex = ch - SBase;
 553   TIndex = SIndex % TCount;
 554
 555   if (TIndex)
 556     {
 557       /* split LVT -> LV,T */
 558       *a = ch - TIndex;
 559       *b = TBase + TIndex;
 560     }
 561   else
 562     {
 563       /* split LV -> L,V */
 564       *a = LBase + SIndex / NCount;
 565       *b = VBase + (SIndex % NCount) / TCount;
 566     }
 567
 568   return TRUE;
 569 }
 570
 571 /**
 572  * g_unichar_decompose:
 573  * @ch: a Unicode character
 574  * @a: return location for the first component of @ch
 575  * @b: return location for the second component of @ch
 576  *
 577  * Performs a single decomposition step of the
 578  * Unicode canonical decomposition algorithm.
 579  *
 580  * This function does not include compatibility
 581  * decompositions. It does, however, include algorithmic
 582  * Hangul Jamo decomposition, as well as 'singleton'
 583  * decompositions which replace a character by a single
 584  * other character. In the case of singletons *@b will
 585  * be set to zero.
 586  *
 587  * If @ch is not decomposable, *@a is set to @ch and *@b
 588  * is set to zero.
 589  *
 590  * Note that the way Unicode decomposition pairs are
 591  * defined, it is guaranteed that @b would not decompose
 592  * further, but @a may itself decompose.  To get the full
 593  * canonical decomposition for @ch, one would need to
 594  * recursively call this function on @a.  Or use
 595  * g_unichar_fully_decompose().
 596  *
 597  * See
 598  * [UAX#15](http://unicode.org/reports/tr15/)
 599  * for details.
 600  *
 601  * Returns: %TRUE if the character could be decomposed
 602  *
 603  * Since: 2.30
 604  */
 605 gboolean
 606 g_unichar_decompose (gunichar  ch,
 607                      gunichar *a,
 608                      gunichar *b)
 609 {
 610   gint start = 0;
 611   gint end = G_N_ELEMENTS (decomp_step_table);
 612
 613   if (decompose_hangul_step (ch, a, b))
 614     return TRUE;
 615
 616   /* TODO use bsearch() */
 617   if (ch >= decomp_step_table[start].ch &&
 618       ch <= decomp_step_table[end - 1].ch)
 619     {
 620       while (TRUE)
 621         {
 622           gint half = (start + end) / 2;
 623           const decomposition_step *p = &(decomp_step_table[half]);
 624           if (ch == p->ch)
 625             {
 626               *a = p->a;
 627               *b = p->b;
 628               return TRUE;
 629             }
 630           else if (half == start)
 631             break;
 632           else if (ch > p->ch)
 633             start = half;
 634           else
 635             end = half;
 636         }
 637     }
 638
 639   *a = ch;
 640   *b = 0;
 641
 642   return FALSE;
 643 }
 644
 645 /**
 646  * g_unichar_compose:
 647  * @a: a Unicode character
 648  * @b: a Unicode character
 649  * @ch: return location for the composed character
 650  *
 651  * Performs a single composition step of the
 652  * Unicode canonical composition algorithm.
 653  *
 654  * This function includes algorithmic Hangul Jamo composition,
 655  * but it is not exactly the inverse of g_unichar_decompose().
 656  * No composition can have either of @a or @b equal to zero.
 657  * To be precise, this function composes if and only if
 658  * there exists a Primary Composite P which is canonically
 659  * equivalent to the sequence <@a,@b>.  See the Unicode
 660  * Standard for the definition of Primary Composite.
 661  *
 662  * If @a and @b do not compose a new character, @ch is set to zero.
 663  *
 664  * See
 665  * [UAX#15](http://unicode.org/reports/tr15/)
 666  * for details.
 667  *
 668  * Returns: %TRUE if the characters could be composed
 669  *
 670  * Since: 2.30
 671  */
 672 gboolean
 673 g_unichar_compose (gunichar  a,
 674                    gunichar  b,
 675                    gunichar *ch)
 676 {
 677   if (combine (a, b, ch))
 678     return TRUE;
 679
 680   *ch = 0;
 681   return FALSE;
 682 }
 683
 684 /**
 685  * g_unichar_fully_decompose:
 686  * @ch: a Unicode character.
 687  * @compat: whether perform canonical or compatibility decomposition
 688  * @result: (nullable): location to store decomposed result, or %NULL
 689  * @result_len: length of @result
 690  *
 691  * Computes the canonical or compatibility decomposition of a
 692  * Unicode character.  For compatibility decomposition,
 693  * pass %TRUE for @compat; for canonical decomposition
 694  * pass %FALSE for @compat.
 695  *
 696  * The decomposed sequence is placed in @result.  Only up to
 697  * @result_len characters are written into @result.  The length
 698  * of the full decomposition (irrespective of @result_len) is
 699  * returned by the function.  For canonical decomposition,
 700  * currently all decompositions are of length at most 4, but
 701  * this may change in the future (very unlikely though).
 702  * At any rate, Unicode does guarantee that a buffer of length
 703  * 18 is always enough for both compatibility and canonical
 704  * decompositions, so that is the size recommended. This is provided
 705  * as %G_UNICHAR_MAX_DECOMPOSITION_LENGTH.
 706  *
 707  * See
 708  * [UAX#15](http://unicode.org/reports/tr15/)
 709  * for details.
 710  *
 711  * Returns: the length of the full decomposition.
 712  *
 713  * Since: 2.30
 714  **/
 715 gsize
 716 g_unichar_fully_decompose (gunichar  ch,
 717                            gboolean  compat,
 718                            gunichar *result,
 719                            gsize     result_len)
 720 {
 721   const gchar *decomp;
 722   const gchar *p;
 723
 724   /* Hangul syllable */
 725   if (ch >= SBase && ch < SBase + SCount)
 726     {
 727       gsize len, i;
 728       gunichar buffer[3];
 729       decompose_hangul (ch, result ? buffer : NULL, &len);
 730       if (result)
 731         for (i = 0; i < len && i < result_len; i++)
 732           result[i] = buffer[i];
 733       return len;
 734     }
 735   else if ((decomp = find_decomposition (ch, compat)) != NULL)
 736     {
 737       /* Found it.  */
 738       gsize len, i;
 739
 740       len = g_utf8_strlen (decomp, -1);
 741
 742       for (p = decomp, i = 0; i < len && i < result_len; p = g_utf8_next_char (p), i++)
 743         result[i] = g_utf8_get_char (p);
 744
 745       return len;
 746     }
 747
 748   /* Does not decompose */
 749   if (result && result_len >= 1)
 750     *result = ch;
 751   return 1;
 752 }