lib/unicase/u-casemap.h

   1 /* Case mapping for UTF-8/UTF-16/UTF-32 strings (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 UNIT *
  19 FUNC (const UNIT *s, size_t n,
  20       casing_prefix_context_t prefix_context,
  21       casing_suffix_context_t suffix_context,
  22       const char *iso639_language,
  23       ucs4_t (*single_character_map) (ucs4_t),
  24       size_t offset_in_rule, /* offset in 'struct special_casing_rule' */
  25       uninorm_t nf,
  26       UNIT *resultbuf, size_t *lengthp)
  27 {
  28   /* The result being accumulated.  */
  29   UNIT *result;
  30   size_t length;
  31   size_t allocated;
  32
  33   /* Initialize the accumulator.  */
  34   if (nf != NULL || resultbuf == NULL)
  35     {
  36       result = NULL;
  37       allocated = 0;
  38     }
  39   else
  40     {
  41       result = resultbuf;
  42       allocated = *lengthp;
  43     }
  44   length = 0;
  45
  46   {
  47     const UNIT *s_end = s + n;
  48
  49     /* Helper for evaluating the FINAL_SIGMA condition:
  50        Last character that was not case-ignorable.  */
  51     ucs4_t last_char_except_ignorable =
  52       prefix_context.last_char_except_ignorable;
  53
  54     /* Helper for evaluating the AFTER_SOFT_DOTTED and AFTER_I conditions:
  55        Last character that was of combining class 230 ("Above") or 0.  */
  56     ucs4_t last_char_normal_or_above =
  57       prefix_context.last_char_normal_or_above;
  58
  59     while (s < s_end)
  60       {
  61         ucs4_t uc;
  62         int count = U_MBTOUC_UNSAFE (&uc, s, s_end - s);
  63
  64         ucs4_t mapped_uc[3];
  65         unsigned int mapped_count;
  66
  67         if (uc < 0x10000)
  68           {
  69             /* Look first in the special-casing table.  */
  70             char code[3];
  71
  72             code[0] = (uc >> 8) & 0xff;
  73             code[1] = uc & 0xff;
  74
  75             for (code[2] = 0; ; code[2]++)
  76               {
  77                 const struct special_casing_rule *rule =
  78                   gl_unicase_special_lookup (code, 3);
  79
  80                 if (rule == NULL)
  81                   break;
  82
  83                 /* Test if the condition applies.  */
  84                 /* Does the language apply?  */
  85                 if (rule->language[0] == '\0'
  86                     || (iso639_language != NULL
  87                         && iso639_language[0] == rule->language[0]
  88                         && iso639_language[1] == rule->language[1]))
  89                   {
  90                     /* Does the context apply?  */
  91                     int context = rule->context;
  92                     bool applies;
  93
  94                     if (context < 0)
  95                       context = - context;
  96                     switch (context)
  97                       {
  98                       case SCC_ALWAYS:
  99                         applies = true;
 100                         break;
 101
 102                       case SCC_FINAL_SIGMA:
 103                         /* "Before" condition: preceded by a sequence
 104                            consisting of a cased letter and a case-ignorable
 105                            sequence.
 106                            "After" condition: not followed by a sequence
 107                            consisting of a case-ignorable sequence and then a
 108                            cased letter.  */
 109                         /* Test the "before" condition.  */
 110                         applies = uc_is_cased (last_char_except_ignorable);
 111                         /* Test the "after" condition.  */
 112                         if (applies)
 113                           {
 114                             const UNIT *s2 = s + count;
 115                             for (;;)
 116                               {
 117                                 if (s2 < s_end)
 118                                   {
 119                                     ucs4_t uc2;
 120                                     int count2 = U_MBTOUC_UNSAFE (&uc2, s2, s_end - s2);
 121                                     if (uc_is_cased (uc2))
 122                                       {
 123                                         applies = false;
 124                                         break;
 125                                       }
 126                                     if (!uc_is_case_ignorable (uc2))
 127                                       break;
 128                                     s2 += count2;
 129                                   }
 130                                 else
 131                                   {
 132                                     applies = ((suffix_context.bits & SCC_FINAL_SIGMA_MASK) == 0);
 133                                     break;
 134                                   }
 135                               }
 136                           }
 137                         break;
 138
 139                       case SCC_AFTER_SOFT_DOTTED:
 140                         /* "Before" condition: There is a Soft_Dotted character
 141                            before it, with no intervening character of
 142                            combining class 0 or 230 (Above).  */
 143                         /* Test the "before" condition.  */
 144                         applies = uc_is_property_soft_dotted (last_char_normal_or_above);
 145                         break;
 146
 147                       case SCC_MORE_ABOVE:
 148                         /* "After" condition: followed by a character of
 149                            combining class 230 (Above) with no intervening
 150                            character of combining class 0 or 230 (Above).  */
 151                         /* Test the "after" condition.  */
 152                         {
 153                           const UNIT *s2 = s + count;
 154                           applies = false;
 155                           for (;;)
 156                             {
 157                               if (s2 < s_end)
 158                                 {
 159                                   ucs4_t uc2;
 160                                   int count2 = U_MBTOUC_UNSAFE (&uc2, s2, s_end - s2);
 161                                   int ccc = uc_combining_class (uc2);
 162                                   if (ccc == UC_CCC_A)
 163                                     {
 164                                       applies = true;
 165                                       break;
 166                                     }
 167                                   if (ccc == UC_CCC_NR)
 168                                     break;
 169                                   s2 += count2;
 170                                 }
 171                               else
 172                                 {
 173                                   applies = ((suffix_context.bits & SCC_MORE_ABOVE_MASK) != 0);
 174                                   break;
 175                                 }
 176                             }
 177                         }
 178                         break;
 179
 180                       case SCC_BEFORE_DOT:
 181                         /* "After" condition: followed by COMBINING DOT ABOVE
 182                            (U+0307). Any sequence of characters with a
 183                            combining class that is neither 0 nor 230 may
 184                            intervene between the current character and the
 185                            combining dot above.  */
 186                         /* Test the "after" condition.  */
 187                         {
 188                           const UNIT *s2 = s + count;
 189                           applies = false;
 190                           for (;;)
 191                             {
 192                               if (s2 < s_end)
 193                                 {
 194                                   ucs4_t uc2;
 195                                   int count2 = U_MBTOUC_UNSAFE (&uc2, s2, s_end - s2);
 196                                   if (uc2 == 0x0307) /* COMBINING DOT ABOVE */
 197                                     {
 198                                       applies = true;
 199                                       break;
 200                                     }
 201                                   {
 202                                     int ccc = uc_combining_class (uc2);
 203                                     if (ccc == UC_CCC_A || ccc == UC_CCC_NR)
 204                                       break;
 205                                   }
 206                                   s2 += count2;
 207                                 }
 208                               else
 209                                 {
 210                                   applies = ((suffix_context.bits & SCC_BEFORE_DOT_MASK) != 0);
 211                                   break;
 212                                 }
 213                             }
 214                         }
 215                         break;
 216
 217                       case SCC_AFTER_I:
 218                         /* "Before" condition: There is an uppercase I before
 219                            it, and there is no intervening character of
 220                            combining class 0 or 230 (Above).  */
 221                         /* Test the "before" condition.  */
 222                         applies = (last_char_normal_or_above == 'I');
 223                         break;
 224
 225                       default:
 226                         abort ();
 227                       }
 228                     if (rule->context < 0)
 229                       applies = !applies;
 230
 231                     if (applies)
 232                       {
 233                         /* The rule applies.
 234                            Look up the mapping (0 to 3 characters).  */
 235                         const unsigned short *mapped_in_rule =
 236                           (const unsigned short *)((const char *)rule + offset_in_rule);
 237
 238                         if (mapped_in_rule[0] == 0)
 239                           mapped_count = 0;
 240                         else
 241                           {
 242                             mapped_uc[0] = mapped_in_rule[0];
 243                             if (mapped_in_rule[1] == 0)
 244                               mapped_count = 1;
 245                             else
 246                               {
 247                                 mapped_uc[1] = mapped_in_rule[1];
 248                                 if (mapped_in_rule[2] == 0)
 249                                   mapped_count = 2;
 250                                 else
 251                                   {
 252                                     mapped_uc[2] = mapped_in_rule[2];
 253                                     mapped_count = 3;
 254                                   }
 255                               }
 256                           }
 257                         goto found_mapping;
 258                       }
 259                   }
 260
 261                 /* Optimization: Save a hash table lookup in the next round.  */
 262                 if (!rule->has_next)
 263                   break;
 264               }
 265           }
 266
 267         /* No special-cased mapping.  So use the locale and context independent
 268            mapping.  */
 269         mapped_uc[0] = single_character_map (uc);
 270         mapped_count = 1;
 271
 272        found_mapping:
 273         /* Found the mapping: uc maps to mapped_uc[0..mapped_count-1].  */
 274         {
 275           unsigned int i;
 276
 277           for (i = 0; i < mapped_count; i++)
 278             {
 279               ucs4_t muc = mapped_uc[i];
 280
 281               /* Append muc to the result accumulator.  */
 282               if (length < allocated)
 283                 {
 284                   int ret = U_UCTOMB (result + length, muc, allocated - length);
 285                   if (ret == -1)
 286                     {
 287                       errno = EINVAL;
 288                       goto fail;
 289                     }
 290                   if (ret >= 0)
 291                     {
 292                       length += ret;
 293                       goto done_appending;
 294                     }
 295                 }
 296               {
 297                 size_t old_allocated = allocated;
 298                 size_t new_allocated = 2 * old_allocated;
 299                 if (new_allocated < 64)
 300                   new_allocated = 64;
 301                 if (new_allocated < old_allocated) /* integer overflow? */
 302                   abort ();
 303                 {
 304                   UNIT *larger_result;
 305                   if (result == NULL)
 306                     {
 307                       larger_result = (UNIT *) malloc (new_allocated * sizeof (UNIT));
 308                       if (larger_result == NULL)
 309                         {
 310                           errno = ENOMEM;
 311                           goto fail;
 312                         }
 313                     }
 314                   else if (result == resultbuf)
 315                     {
 316                       larger_result = (UNIT *) malloc (new_allocated * sizeof (UNIT));
 317                       if (larger_result == NULL)
 318                         {
 319                           errno = ENOMEM;
 320                           goto fail;
 321                         }
 322                       U_CPY (larger_result, resultbuf, length);
 323                     }
 324                   else
 325                     {
 326                       larger_result =
 327                         (UNIT *) realloc (result, new_allocated * sizeof (UNIT));
 328                       if (larger_result == NULL)
 329                         {
 330                           errno = ENOMEM;
 331                           goto fail;
 332                         }
 333                     }
 334                   result = larger_result;
 335                   allocated = new_allocated;
 336                   {
 337                     int ret = U_UCTOMB (result + length, muc, allocated - length);
 338                     if (ret == -1)
 339                       {
 340                         errno = EINVAL;
 341                         goto fail;
 342                       }
 343                     if (ret < 0)
 344                       abort ();
 345                     length += ret;
 346                     goto done_appending;
 347                   }
 348                 }
 349               }
 350              done_appending: ;
 351             }
 352         }
 353
 354         if (!uc_is_case_ignorable (uc))
 355           last_char_except_ignorable = uc;
 356
 357         {
 358           int ccc = uc_combining_class (uc);
 359           if (ccc == UC_CCC_A || ccc == UC_CCC_NR)
 360             last_char_normal_or_above = uc;
 361         }
 362
 363         s += count;
 364       }
 365   }
 366
 367   if (nf != NULL)
 368     {
 369       /* Finally, normalize the result.  */
 370       UNIT *normalized_result;
 371
 372       normalized_result = U_NORMALIZE (nf, result, length, resultbuf, lengthp);
 373       if (normalized_result == NULL)
 374         goto fail;
 375
 376       free (result);
 377       return normalized_result;
 378     }
 379
 380   if (length == 0)
 381     {
 382       if (result == NULL)
 383         {
 384           /* Return a non-NULL value.  NULL means error.  */
 385           result = (UNIT *) malloc (1);
 386           if (result == NULL)
 387             {
 388               errno = ENOMEM;
 389               goto fail;
 390             }
 391         }
 392     }
 393   else if (result != resultbuf && length < allocated)
 394     {
 395       /* Shrink the allocated memory if possible.  */
 396       UNIT *memory;
 397
 398       memory = (UNIT *) realloc (result, length * sizeof (UNIT));
 399       if (memory != NULL)
 400         result = memory;
 401     }
 402
 403   *lengthp = length;
 404   return result;
 405
 406  fail:
 407   if (result != resultbuf)
 408     {
 409       int saved_errno = errno;
 410       free (result);
 411       errno = saved_errno;
 412     }
 413   return NULL;
 414 }