lib/unicase/u-ct-totitle.h

   1 /* Titlecase mapping for UTF-8/UTF-16/UTF-32 substrings (locale dependent).
   2    Copyright (C) 2009 Free Software Foundation, Inc.
   3    Written by Bruno Haible <bruno@clisp.org>, 2009.
   4
   5    This program is free software: you can redistribute it and/or modify it
   6    under the terms of the GNU Lesser General Public License as published
   7    by the Free Software Foundation; either version 3 of the License, or
   8    (at your option) any later version.
   9
  10    This program is distributed in the hope that it will be useful,
  11    but WITHOUT ANY WARRANTY; without even the implied warranty of
  12    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13    Lesser General Public License for more details.
  14
  15    You should have received a copy of the GNU Lesser General Public License
  16    along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
  17
  18 /* Quoting the Unicode standard, section "Default Case Algorithms":
  19      Find the word boundaries in X according to Unicode Standard Annex #29,
  20      “Text Boundaries.” For each word boundary, find the first cased character
  21      F following the word boundary. If F exists, map F to Titlecase_Mapping(F);
  22      then map all characters C between F and the following word boundary to
  23      Lowercase_Mapping(C).  */
  24
  25 UNIT *
  26 FUNC (const UNIT *s, size_t n,
  27       casing_prefix_context_t prefix_context,
  28       casing_suffix_context_t suffix_context,
  29       const char *iso639_language,
  30       uninorm_t nf,
  31       UNIT *resultbuf, size_t *lengthp)
  32 {
  33   /* The result being accumulated.  */
  34   UNIT *result;
  35   size_t length;
  36   size_t allocated;
  37   /* An array containing the word break positions.  */
  38   char *wordbreaks;
  39
  40   /* Initialize the accumulator.  */
  41   if (nf != NULL || resultbuf == NULL)
  42     {
  43       result = NULL;
  44       allocated = 0;
  45     }
  46   else
  47     {
  48       result = resultbuf;
  49       allocated = *lengthp;
  50     }
  51   length = 0;
  52
  53   /* Initialize the word breaks array.  */
  54   if (n > 0)
  55     {
  56       wordbreaks = (char *) malloc (n);
  57       if (wordbreaks == NULL)
  58         {
  59           errno = ENOMEM;
  60           goto fail2;
  61         }
  62       U_WORDBREAKS (s, n, wordbreaks);
  63     }
  64   else
  65     wordbreaks = NULL;
  66
  67   {
  68     const UNIT *s_end = s + n;
  69     const char *wp = wordbreaks;
  70
  71     /* When considering the string as segmented by word boundaries: For each
  72        such segment:
  73         - In the first part, we are searching for the first cased character.
  74           In this state, in_word_first_part = true, and no conversion takes
  75           place.
  76         - In the second part, we are converting every character: the first
  77           among these characters to title case, the other ones to lower case.
  78           In this state, in_word_first_part = false.  */
  79     bool in_word_first_part = true;
  80
  81     /* Helper for evaluating the FINAL_SIGMA condition:
  82        Last character that was not case-ignorable.  */
  83     ucs4_t last_char_except_ignorable =
  84       prefix_context.last_char_except_ignorable;
  85
  86     /* Helper for evaluating the AFTER_SOFT_DOTTED and AFTER_I conditions:
  87        Last character that was of combining class 230 ("Above") or 0.  */
  88     ucs4_t last_char_normal_or_above =
  89       prefix_context.last_char_normal_or_above;
  90
  91     while (s < s_end)
  92       {
  93         /* Fetch the next character.  */
  94         ucs4_t uc;
  95         int count = U_MBTOUC_UNSAFE (&uc, s, s_end - s);
  96
  97         ucs4_t (*single_character_map) (ucs4_t);
  98         size_t offset_in_rule; /* offset in 'struct special_casing_rule' */
  99
 100         ucs4_t mapped_uc[3];
 101         unsigned int mapped_count;
 102
 103         if (*wp)
 104           /* Crossing a word boundary.  */
 105           in_word_first_part = true;
 106
 107         /* Determine single_character_map, offset_in_rule.
 108            There are three possibilities:
 109              - uc should not be converted.
 110              - uc should be titlecased.
 111              - uc should be lowercased.  */
 112         if (in_word_first_part)
 113           {
 114             if (uc_is_cased (uc))
 115               {
 116                 /* uc is to be titlecased.  */
 117                 single_character_map = uc_totitle;
 118                 offset_in_rule = offsetof (struct special_casing_rule, title[0]);
 119                 in_word_first_part = false;
 120               }
 121             else
 122               {
 123                 /* uc is not converted.  */
 124                 single_character_map = NULL;
 125                 offset_in_rule = 0;
 126               }
 127           }
 128         else
 129           {
 130             /* uc is to be lowercased.  */
 131             single_character_map = uc_tolower;
 132             offset_in_rule = offsetof (struct special_casing_rule, lower[0]);
 133           }
 134
 135         /* Actually map uc.  */
 136         if (single_character_map == NULL)
 137           {
 138             mapped_uc[0] = uc;
 139             mapped_count = 1;
 140             goto found_mapping;
 141           }
 142
 143         if (uc < 0x10000)
 144           {
 145             /* Look first in the special-casing table.  */
 146             char code[3];
 147
 148             code[0] = (uc >> 8) & 0xff;
 149             code[1] = uc & 0xff;
 150
 151             for (code[2] = 0; ; code[2]++)
 152               {
 153                 const struct special_casing_rule *rule =
 154                   gl_unicase_special_lookup (code, 3);
 155
 156                 if (rule == NULL)
 157                   break;
 158
 159                 /* Test if the condition applies.  */
 160                 /* Does the language apply?  */
 161                 if (rule->language[0] == '\0'
 162                     || (iso639_language != NULL
 163                         && iso639_language[0] == rule->language[0]
 164                         && iso639_language[1] == rule->language[1]))
 165                   {
 166                     /* Does the context apply?  */
 167                     int context = rule->context;
 168                     bool applies;
 169
 170                     if (context < 0)
 171                       context = - context;
 172                     switch (context)
 173                       {
 174                       case SCC_ALWAYS:
 175                         applies = true;
 176                         break;
 177
 178                       case SCC_FINAL_SIGMA:
 179                         /* "Before" condition: preceded by a sequence
 180                            consisting of a cased letter and a case-ignorable
 181                            sequence.
 182                            "After" condition: not followed by a sequence
 183                            consisting of a case-ignorable sequence and then a
 184                            cased letter.  */
 185                         /* Test the "before" condition.  */
 186                         applies = uc_is_cased (last_char_except_ignorable);
 187                         /* Test the "after" condition.  */
 188                         if (applies)
 189                           {
 190                             const UNIT *s2 = s + count;
 191                             for (;;)
 192                               {
 193                                 if (s2 < s_end)
 194                                   {
 195                                     ucs4_t uc2;
 196                                     int count2 = U_MBTOUC_UNSAFE (&uc2, s2, s_end - s2);
 197                                     if (uc_is_cased (uc2))
 198                                       {
 199                                         applies = false;
 200                                         break;
 201                                       }
 202                                     if (!uc_is_case_ignorable (uc2))
 203                                       break;
 204                                     s2 += count2;
 205                                   }
 206                                 else
 207                                   {
 208                                     applies = ((suffix_context.bits & SCC_FINAL_SIGMA_MASK) == 0);
 209                                     break;
 210                                   }
 211                               }
 212                           }
 213                         break;
 214
 215                       case SCC_AFTER_SOFT_DOTTED:
 216                         /* "Before" condition: There is a Soft_Dotted character
 217                            before it, with no intervening character of
 218                            combining class 0 or 230 (Above).  */
 219                         /* Test the "before" condition.  */
 220                         applies = uc_is_property_soft_dotted (last_char_normal_or_above);
 221                         break;
 222
 223                       case SCC_MORE_ABOVE:
 224                         /* "After" condition: followed by a character of
 225                            combining class 230 (Above) with no intervening
 226                            character of combining class 0 or 230 (Above).  */
 227                         /* Test the "after" condition.  */
 228                         {
 229                           const UNIT *s2 = s + count;
 230                           applies = false;
 231                           for (;;)
 232                             {
 233                               if (s2 < s_end)
 234                                 {
 235                                   ucs4_t uc2;
 236                                   int count2 = U_MBTOUC_UNSAFE (&uc2, s2, s_end - s2);
 237                                   int ccc = uc_combining_class (uc2);
 238                                   if (ccc == UC_CCC_A)
 239                                     {
 240                                       applies = true;
 241                                       break;
 242                                     }
 243                                   if (ccc == UC_CCC_NR)
 244                                     break;
 245                                   s2 += count2;
 246                                 }
 247                               else
 248                                 {
 249                                   applies = ((suffix_context.bits & SCC_MORE_ABOVE_MASK) != 0);
 250                                   break;
 251                                 }
 252                             }
 253                         }
 254                         break;
 255
 256                       case SCC_BEFORE_DOT:
 257                         /* "After" condition: followed by COMBINING DOT ABOVE
 258                            (U+0307). Any sequence of characters with a
 259                            combining class that is neither 0 nor 230 may
 260                            intervene between the current character and the
 261                            combining dot above.  */
 262                         /* Test the "after" condition.  */
 263                         {
 264                           const UNIT *s2 = s + count;
 265                           applies = false;
 266                           for (;;)
 267                             {
 268                               if (s2 < s_end)
 269                                 {
 270                                   ucs4_t uc2;
 271                                   int count2 = U_MBTOUC_UNSAFE (&uc2, s2, s_end - s2);
 272                                   if (uc2 == 0x0307) /* COMBINING DOT ABOVE */
 273                                     {
 274                                       applies = true;
 275                                       break;
 276                                     }
 277                                   {
 278                                     int ccc = uc_combining_class (uc2);
 279                                     if (ccc == UC_CCC_A || ccc == UC_CCC_NR)
 280                                       break;
 281                                   }
 282                                   s2 += count2;
 283                                 }
 284                               else
 285                                 {
 286                                   applies = ((suffix_context.bits & SCC_BEFORE_DOT_MASK) != 0);
 287                                   break;
 288                                 }
 289                             }
 290                         }
 291                         break;
 292
 293                       case SCC_AFTER_I:
 294                         /* "Before" condition: There is an uppercase I before
 295                            it, and there is no intervening character of
 296                            combining class 0 or 230 (Above).  */
 297                         /* Test the "before" condition.  */
 298                         applies = (last_char_normal_or_above == 'I');
 299                         break;
 300
 301                       default:
 302                         abort ();
 303                       }
 304                     if (rule->context < 0)
 305                       applies = !applies;
 306
 307                     if (applies)
 308                       {
 309                         /* The rule applies.
 310                            Look up the mapping (0 to 3 characters).  */
 311                         const unsigned short *mapped_in_rule =
 312                           (const unsigned short *)((const char *)rule + offset_in_rule);
 313
 314                         if (mapped_in_rule[0] == 0)
 315                           mapped_count = 0;
 316                         else
 317                           {
 318                             mapped_uc[0] = mapped_in_rule[0];
 319                             if (mapped_in_rule[1] == 0)
 320                               mapped_count = 1;
 321                             else
 322                               {
 323                                 mapped_uc[1] = mapped_in_rule[1];
 324                                 if (mapped_in_rule[2] == 0)
 325                                   mapped_count = 2;
 326                                 else
 327                                   {
 328                                     mapped_uc[2] = mapped_in_rule[2];
 329                                     mapped_count = 3;
 330                                   }
 331                               }
 332                           }
 333                         goto found_mapping;
 334                       }
 335                   }
 336
 337                 /* Optimization: Save a hash table lookup in the next round.  */
 338                 if (!rule->has_next)
 339                   break;
 340               }
 341           }
 342
 343         /* No special-cased mapping.  So use the locale and context independent
 344            mapping.  */
 345         mapped_uc[0] = single_character_map (uc);
 346         mapped_count = 1;
 347
 348        found_mapping:
 349         /* Found the mapping: uc maps to mapped_uc[0..mapped_count-1].  */
 350         {
 351           unsigned int i;
 352
 353           for (i = 0; i < mapped_count; i++)
 354             {
 355               ucs4_t muc = mapped_uc[i];
 356
 357               /* Append muc to the result accumulator.  */
 358               if (length < allocated)
 359                 {
 360                   int ret = U_UCTOMB (result + length, muc, allocated - length);
 361                   if (ret == -1)
 362                     {
 363                       errno = EINVAL;
 364                       goto fail1;
 365                     }
 366                   if (ret >= 0)
 367                     {
 368                       length += ret;
 369                       goto done_appending;
 370                     }
 371                 }
 372               {
 373                 size_t old_allocated = allocated;
 374                 size_t new_allocated = 2 * old_allocated;
 375                 if (new_allocated < 64)
 376                   new_allocated = 64;
 377                 if (new_allocated < old_allocated) /* integer overflow? */
 378                   abort ();
 379                 {
 380                   UNIT *larger_result;
 381                   if (result == NULL)
 382                     {
 383                       larger_result = (UNIT *) malloc (new_allocated * sizeof (UNIT));
 384                       if (larger_result == NULL)
 385                         {
 386                           errno = ENOMEM;
 387                           goto fail1;
 388                         }
 389                     }
 390                   else if (result == resultbuf)
 391                     {
 392                       larger_result = (UNIT *) malloc (new_allocated * sizeof (UNIT));
 393                       if (larger_result == NULL)
 394                         {
 395                           errno = ENOMEM;
 396                           goto fail1;
 397                         }
 398                       U_CPY (larger_result, resultbuf, length);
 399                     }
 400                   else
 401                     {
 402                       larger_result =
 403                         (UNIT *) realloc (result, new_allocated * sizeof (UNIT));
 404                       if (larger_result == NULL)
 405                         {
 406                           errno = ENOMEM;
 407                           goto fail1;
 408                         }
 409                     }
 410                   result = larger_result;
 411                   allocated = new_allocated;
 412                   {
 413                     int ret = U_UCTOMB (result + length, muc, allocated - length);
 414                     if (ret == -1)
 415                       {
 416                         errno = EINVAL;
 417                         goto fail1;
 418                       }
 419                     if (ret < 0)
 420                       abort ();
 421                     length += ret;
 422                     goto done_appending;
 423                   }
 424                 }
 425               }
 426              done_appending: ;
 427             }
 428         }
 429
 430         if (!uc_is_case_ignorable (uc))
 431           last_char_except_ignorable = uc;
 432
 433         {
 434           int ccc = uc_combining_class (uc);
 435           if (ccc == UC_CCC_A || ccc == UC_CCC_NR)
 436             last_char_normal_or_above = uc;
 437         }
 438
 439         s += count;
 440         wp += count;
 441       }
 442   }
 443
 444   free (wordbreaks);
 445
 446   if (nf != NULL)
 447     {
 448       /* Finally, normalize the result.  */
 449       UNIT *normalized_result;
 450
 451       normalized_result = U_NORMALIZE (nf, result, length, resultbuf, lengthp);
 452       if (normalized_result == NULL)
 453         goto fail2;
 454
 455       free (result);
 456       return normalized_result;
 457     }
 458
 459   if (length == 0)
 460     {
 461       if (result == NULL)
 462         {
 463           /* Return a non-NULL value.  NULL means error.  */
 464           result = (UNIT *) malloc (1);
 465           if (result == NULL)
 466             {
 467               errno = ENOMEM;
 468               goto fail2;
 469             }
 470         }
 471     }
 472   else if (result != resultbuf && length < allocated)
 473     {
 474       /* Shrink the allocated memory if possible.  */
 475       UNIT *memory;
 476
 477       memory = (UNIT *) realloc (result, length * sizeof (UNIT));
 478       if (memory != NULL)
 479         result = memory;
 480     }
 481
 482   *lengthp = length;
 483   return result;
 484
 485  fail1:
 486   {
 487     int saved_errno = errno;
 488     free (wordbreaks);
 489     errno = saved_errno;
 490   }
 491  fail2:
 492   if (result != resultbuf)
 493     {
 494       int saved_errno = errno;
 495       free (result);
 496       errno = saved_errno;
 497     }
 498   return NULL;
 499 }