src/icu_I18N.c

   1 /* This file is part of the YAZ toolkit.
   2  * Copyright (C) 1995-2009 Index Data
   3  * See the file LICENSE for details.
   4  */
   5
   6 /**
   7  * \file icu_I18N.c
   8  * \brief ICU utilities
   9  */
  10
  11 #if HAVE_CONFIG_H
  12 #include "config.h"
  13 #endif
  14
  15 #define USE_TIMING 0
  16 #if USE_TIMING
  17 #include <yaz/timing.h>
  18 #endif
  19
  20 #if YAZ_HAVE_ICU
  21 #include <yaz/xmalloc.h>
  22
  23 #include <yaz/icu_I18N.h>
  24
  25 #include <yaz/log.h>
  26
  27 #include <string.h>
  28 #include <stdlib.h>
  29 #include <stdio.h>
  30
  31 #include <unicode/ustring.h>  /* some more string fcns*/
  32 #include <unicode/uchar.h>    /* char names           */
  33 #include <unicode/ucol.h>
  34
  35 int icu_check_status(UErrorCode status)
  36 {
  37     if (U_FAILURE(status))
  38     {
  39         yaz_log(YLOG_WARN, "ICU: %d %s\n", status, u_errorName(status));
  40         return 0;
  41     }
  42     return 1;
  43 }
  44
  45 struct icu_buf_utf16 * icu_buf_utf16_create(size_t capacity)
  46 {
  47     struct icu_buf_utf16 * buf16
  48         = (struct icu_buf_utf16 *) xmalloc(sizeof(struct icu_buf_utf16));
  49
  50     buf16->utf16 = 0;
  51     buf16->utf16_len = 0;
  52     buf16->utf16_cap = 0;
  53
  54     if (capacity > 0)
  55     {
  56         buf16->utf16 = (UChar *) xmalloc(sizeof(UChar) * capacity);
  57         buf16->utf16[0] = (UChar) 0;
  58         buf16->utf16_cap = capacity;
  59     }
  60     return buf16;
  61 }
  62
  63 struct icu_buf_utf16 * icu_buf_utf16_clear(struct icu_buf_utf16 * buf16)
  64 {
  65     if (buf16)
  66     {
  67         if (buf16->utf16)
  68             buf16->utf16[0] = (UChar) 0;
  69         buf16->utf16_len = 0;
  70     }
  71     return buf16;
  72 }
  73
  74 struct icu_buf_utf16 * icu_buf_utf16_resize(struct icu_buf_utf16 * buf16,
  75                                             size_t capacity)
  76 {
  77     if (!buf16)
  78         return 0;
  79
  80     if (capacity >  0)
  81     {
  82         if (0 == buf16->utf16)
  83             buf16->utf16 = (UChar *) xmalloc(sizeof(UChar) * capacity);
  84         else
  85             buf16->utf16
  86                 = (UChar *) xrealloc(buf16->utf16, sizeof(UChar) * capacity);
  87
  88         icu_buf_utf16_clear(buf16);
  89         buf16->utf16_cap = capacity;
  90     }
  91     else
  92     {
  93         xfree(buf16->utf16);
  94         buf16->utf16 = 0;
  95         buf16->utf16_len = 0;
  96         buf16->utf16_cap = 0;
  97     }
  98
  99     return buf16;
 100 }
 101
 102
 103 struct icu_buf_utf16 * icu_buf_utf16_copy(struct icu_buf_utf16 * dest16,
 104                                           struct icu_buf_utf16 * src16)
 105 {
 106     if (!dest16 || !src16 || dest16 == src16)
 107         return 0;
 108
 109     if (dest16->utf16_cap < src16->utf16_len)
 110         icu_buf_utf16_resize(dest16, src16->utf16_len * 2);
 111
 112     u_strncpy(dest16->utf16, src16->utf16, src16->utf16_len);
 113     dest16->utf16_len = src16->utf16_len;
 114
 115     return dest16;
 116 }
 117
 118 void icu_buf_utf16_destroy(struct icu_buf_utf16 * buf16)
 119 {
 120     if (buf16)
 121         xfree(buf16->utf16);
 122     xfree(buf16);
 123 }
 124
 125 struct icu_buf_utf8 * icu_buf_utf8_create(size_t capacity)
 126 {
 127     struct icu_buf_utf8 * buf8
 128         = (struct icu_buf_utf8 *) xmalloc(sizeof(struct icu_buf_utf8));
 129
 130     buf8->utf8 = 0;
 131     buf8->utf8_len = 0;
 132     buf8->utf8_cap = 0;
 133
 134     if (capacity > 0)
 135     {
 136         buf8->utf8 = (uint8_t *) xmalloc(sizeof(uint8_t) * capacity);
 137         buf8->utf8[0] = (uint8_t) 0;
 138         buf8->utf8_cap = capacity;
 139     }
 140     return buf8;
 141 }
 142
 143 struct icu_buf_utf8 * icu_buf_utf8_clear(struct icu_buf_utf8 * buf8)
 144 {
 145     if (buf8)
 146     {
 147         if (buf8->utf8)
 148             buf8->utf8[0] = (uint8_t) 0;
 149         buf8->utf8_len = 0;
 150     }
 151     return buf8;
 152 }
 153
 154 struct icu_buf_utf8 * icu_buf_utf8_resize(struct icu_buf_utf8 * buf8,
 155                                           size_t capacity)
 156 {
 157     if (!buf8)
 158         return 0;
 159
 160     if (capacity >  0){
 161         if (0 == buf8->utf8)
 162             buf8->utf8 = (uint8_t *) xmalloc(sizeof(uint8_t) * capacity);
 163         else
 164             buf8->utf8
 165                 = (uint8_t *) xrealloc(buf8->utf8, sizeof(uint8_t) * capacity);
 166
 167         buf8->utf8_cap = capacity;
 168     }
 169     else {
 170         xfree(buf8->utf8);
 171         buf8->utf8 = 0;
 172         buf8->utf8_len = 0;
 173         buf8->utf8_cap = 0;
 174     }
 175
 176     return buf8;
 177 }
 178
 179 const char *icu_buf_utf8_to_cstr(struct icu_buf_utf8 *src8)
 180 {
 181     if (!src8 || src8->utf8_len == 0)
 182         return "";
 183
 184     if (src8->utf8_len == src8->utf8_cap)
 185         src8 = icu_buf_utf8_resize(src8, src8->utf8_len * 2 + 1);
 186
 187     src8->utf8[src8->utf8_len] = '\0';
 188
 189     return (const char *) src8->utf8;
 190 }
 191
 192 void icu_buf_utf8_destroy(struct icu_buf_utf8 * buf8)
 193 {
 194     if (buf8)
 195         xfree(buf8->utf8);
 196     xfree(buf8);
 197 }
 198
 199 UErrorCode icu_utf16_from_utf8_cstr(struct icu_buf_utf16 * dest16,
 200                                     const char * src8cstr,
 201                                     UErrorCode * status)
 202 {
 203     size_t src8cstr_len = 0;
 204     int32_t utf16_len = 0;
 205
 206     *status = U_ZERO_ERROR;
 207     src8cstr_len = strlen(src8cstr);
 208
 209     u_strFromUTF8(dest16->utf16, dest16->utf16_cap,
 210                   &utf16_len,
 211                   src8cstr, src8cstr_len, status);
 212
 213     /* check for buffer overflow, resize and retry */
 214     if (*status == U_BUFFER_OVERFLOW_ERROR)
 215     {
 216         icu_buf_utf16_resize(dest16, utf16_len * 2);
 217         *status = U_ZERO_ERROR;
 218         u_strFromUTF8(dest16->utf16, dest16->utf16_cap,
 219                       &utf16_len,
 220                       src8cstr, src8cstr_len, status);
 221     }
 222
 223     if (U_SUCCESS(*status)
 224         && utf16_len <= dest16->utf16_cap)
 225         dest16->utf16_len = utf16_len;
 226     else
 227         icu_buf_utf16_clear(dest16);
 228
 229     return *status;
 230 }
 231
 232 UErrorCode icu_utf16_to_utf8(struct icu_buf_utf8 * dest8,
 233                              struct icu_buf_utf16 * src16,
 234                              UErrorCode * status)
 235 {
 236     int32_t utf8_len = 0;
 237
 238     u_strToUTF8((char *) dest8->utf8, dest8->utf8_cap,
 239                 &utf8_len,
 240                 src16->utf16, src16->utf16_len, status);
 241
 242     /* check for buffer overflow, resize and retry */
 243     if (*status == U_BUFFER_OVERFLOW_ERROR)
 244     {
 245         icu_buf_utf8_resize(dest8, utf8_len * 2);
 246         *status = U_ZERO_ERROR;
 247         u_strToUTF8((char *) dest8->utf8, dest8->utf8_cap,
 248                     &utf8_len,
 249                     src16->utf16, src16->utf16_len, status);
 250     }
 251
 252     if (U_SUCCESS(*status)
 253         && utf8_len <= dest8->utf8_cap)
 254         dest8->utf8_len = utf8_len;
 255     else
 256         icu_buf_utf8_clear(dest8);
 257
 258     return *status;
 259 }
 260
 261
 262
 263 struct icu_casemap * icu_casemap_create(char action, UErrorCode *status)
 264 {
 265     struct icu_casemap * casemap
 266         = (struct icu_casemap *) xmalloc(sizeof(struct icu_casemap));
 267     casemap->action = action;
 268
 269     switch(casemap->action)
 270     {
 271     case 'l':
 272     case 'L':
 273     case 'u':
 274     case 'U':
 275     case 't':
 276     case 'T':
 277     case 'f':
 278     case 'F':
 279         break;
 280     default:
 281         icu_casemap_destroy(casemap);
 282         return 0;
 283     }
 284     return casemap;
 285 }
 286
 287 void icu_casemap_destroy(struct icu_casemap * casemap)
 288 {
 289     xfree(casemap);
 290 }
 291
 292 int icu_casemap_casemap(struct icu_casemap * casemap,
 293                         struct icu_buf_utf16 * dest16,
 294                         struct icu_buf_utf16 * src16,
 295                         UErrorCode *status,
 296                         const char *locale)
 297 {
 298     if(!casemap)
 299         return 0;
 300
 301     return icu_utf16_casemap(dest16, src16, locale,
 302                              casemap->action, status);
 303 }
 304
 305 int icu_utf16_casemap(struct icu_buf_utf16 * dest16,
 306                       struct icu_buf_utf16 * src16,
 307                       const char *locale, char action,
 308                       UErrorCode *status)
 309 {
 310     int32_t dest16_len = 0;
 311
 312     if (!src16->utf16_len)
 313     {           /* guarding for empty source string */
 314         if (dest16->utf16)
 315             dest16->utf16[0] = (UChar) 0;
 316         dest16->utf16_len = 0;
 317         return U_ZERO_ERROR;
 318     }
 319
 320     switch(action)
 321     {
 322     case 'l':
 323     case 'L':
 324         dest16_len = u_strToLower(dest16->utf16, dest16->utf16_cap,
 325                                   src16->utf16, src16->utf16_len,
 326                                   locale, status);
 327         break;
 328     case 'u':
 329     case 'U':
 330         dest16_len = u_strToUpper(dest16->utf16, dest16->utf16_cap,
 331                                   src16->utf16, src16->utf16_len,
 332                                   locale, status);
 333         break;
 334     case 't':
 335     case 'T':
 336         dest16_len = u_strToTitle(dest16->utf16, dest16->utf16_cap,
 337                                   src16->utf16, src16->utf16_len,
 338                                   0, locale, status);
 339         break;
 340     case 'f':
 341     case 'F':
 342         dest16_len = u_strFoldCase(dest16->utf16, dest16->utf16_cap,
 343                                    src16->utf16, src16->utf16_len,
 344                                    U_FOLD_CASE_DEFAULT, status);
 345         break;
 346
 347     default:
 348         return U_UNSUPPORTED_ERROR;
 349         break;
 350     }
 351
 352     /* check for buffer overflow, resize and retry */
 353     if (*status == U_BUFFER_OVERFLOW_ERROR
 354         && dest16 != src16        /* do not resize if in-place conversion */
 355         )
 356     {
 357         icu_buf_utf16_resize(dest16, dest16_len * 2);
 358         *status = U_ZERO_ERROR;
 359
 360         switch(action) {
 361         case 'l':
 362         case 'L':
 363             dest16_len = u_strToLower(dest16->utf16, dest16->utf16_cap,
 364                                       src16->utf16, src16->utf16_len,
 365                                       locale, status);
 366             break;
 367         case 'u':
 368         case 'U':
 369             dest16_len = u_strToUpper(dest16->utf16, dest16->utf16_cap,
 370                                       src16->utf16, src16->utf16_len,
 371                                       locale, status);
 372             break;
 373         case 't':
 374         case 'T':
 375             dest16_len = u_strToTitle(dest16->utf16, dest16->utf16_cap,
 376                                       src16->utf16, src16->utf16_len,
 377                                       0, locale, status);
 378             break;
 379         case 'f':
 380         case 'F':
 381             dest16_len = u_strFoldCase(dest16->utf16, dest16->utf16_cap,
 382                                        src16->utf16, src16->utf16_len,
 383                                        U_FOLD_CASE_DEFAULT, status);
 384             break;
 385
 386         default:
 387             return U_UNSUPPORTED_ERROR;
 388             break;
 389         }
 390     }
 391
 392     if (U_SUCCESS(*status)
 393         && dest16_len <= dest16->utf16_cap)
 394         dest16->utf16_len = dest16_len;
 395     else
 396     {
 397         if (dest16->utf16)
 398             dest16->utf16[0] = (UChar) 0;
 399         dest16->utf16_len = 0;
 400     }
 401
 402     return *status;
 403 }
 404
 405 void icu_sortkey8_from_utf16(UCollator *coll,
 406                              struct icu_buf_utf8 * dest8,
 407                              struct icu_buf_utf16 * src16,
 408                              UErrorCode * status)
 409 {
 410     int32_t sortkey_len = 0;
 411
 412     sortkey_len = ucol_getSortKey(coll, src16->utf16, src16->utf16_len,
 413                                   dest8->utf8, dest8->utf8_cap);
 414
 415     /* check for buffer overflow, resize and retry */
 416     if (sortkey_len > dest8->utf8_cap)
 417     {
 418         icu_buf_utf8_resize(dest8, sortkey_len * 2);
 419         sortkey_len = ucol_getSortKey(coll, src16->utf16, src16->utf16_len,
 420                                       dest8->utf8, dest8->utf8_cap);
 421     }
 422
 423     if (U_SUCCESS(*status)
 424         && sortkey_len > 0)
 425         dest8->utf8_len = sortkey_len;
 426     else
 427         icu_buf_utf8_clear(dest8);
 428 }
 429
 430 struct icu_tokenizer * icu_tokenizer_create(const char *locale, char action,
 431                                             UErrorCode *status)
 432 {
 433     struct icu_tokenizer * tokenizer
 434         = (struct icu_tokenizer *) xmalloc(sizeof(struct icu_tokenizer));
 435
 436     tokenizer->action = action;
 437     tokenizer->bi = 0;
 438     tokenizer->buf16 = 0;
 439     tokenizer->token_count = 0;
 440     tokenizer->token_id = 0;
 441     tokenizer->token_start = 0;
 442     tokenizer->token_end = 0;
 443
 444     switch(tokenizer->action)
 445     {
 446     case 'l':
 447     case 'L':
 448         tokenizer->bi = ubrk_open(UBRK_LINE, locale, 0, 0, status);
 449         break;
 450     case 's':
 451     case 'S':
 452         tokenizer->bi = ubrk_open(UBRK_SENTENCE, locale, 0, 0, status);
 453         break;
 454     case 'w':
 455     case 'W':
 456         tokenizer->bi = ubrk_open(UBRK_WORD, locale, 0, 0, status);
 457         break;
 458     case 'c':
 459     case 'C':
 460         tokenizer->bi = ubrk_open(UBRK_CHARACTER, locale, 0, 0, status);
 461         break;
 462     case 't':
 463     case 'T':
 464         tokenizer->bi = ubrk_open(UBRK_TITLE, locale, 0, 0, status);
 465         break;
 466     default:
 467         *status = U_UNSUPPORTED_ERROR;
 468         return 0;
 469         break;
 470     }
 471
 472     /* ICU error stuff is a very  funny business */
 473     if (U_SUCCESS(*status))
 474         return tokenizer;
 475
 476     /* freeing if failed */
 477     icu_tokenizer_destroy(tokenizer);
 478     return 0;
 479 }
 480
 481 void icu_tokenizer_destroy(struct icu_tokenizer * tokenizer)
 482 {
 483     if (tokenizer) {
 484         if (tokenizer->bi)
 485             ubrk_close(tokenizer->bi);
 486         xfree(tokenizer);
 487     }
 488 }
 489
 490 int icu_tokenizer_attach(struct icu_tokenizer * tokenizer,
 491                          struct icu_buf_utf16 * src16,
 492                          UErrorCode *status)
 493 {
 494     if (!tokenizer || !tokenizer->bi || !src16)
 495         return 0;
 496
 497     tokenizer->buf16 = src16;
 498     tokenizer->token_count = 0;
 499     tokenizer->token_id = 0;
 500     tokenizer->token_start = 0;
 501     tokenizer->token_end = 0;
 502
 503     ubrk_setText(tokenizer->bi, src16->utf16, src16->utf16_len, status);
 504
 505     if (U_FAILURE(*status))
 506         return 0;
 507
 508     return 1;
 509 }
 510
 511 int32_t icu_tokenizer_next_token(struct icu_tokenizer * tokenizer,
 512                          struct icu_buf_utf16 * tkn16,
 513                          UErrorCode *status)
 514 {
 515     int32_t tkn_start = 0;
 516     int32_t tkn_end = 0;
 517     int32_t tkn_len = 0;
 518
 519     if (!tokenizer || !tokenizer->bi
 520         || !tokenizer->buf16 || !tokenizer->buf16->utf16_len)
 521         return 0;
 522     /*
 523     never change tokenizer->buf16 and keep always invariant
 524     0 <= tokenizer->token_start
 525        <= tokenizer->token_end
 526        <= tokenizer->buf16->utf16_len
 527     returns length of token
 528     */
 529
 530     if (0 == tokenizer->token_end) /* first call */
 531         tkn_start = ubrk_first(tokenizer->bi);
 532     else /* successive calls */
 533         tkn_start = tokenizer->token_end;
 534
 535     /* get next position */
 536     tkn_end = ubrk_next(tokenizer->bi);
 537
 538     /* repairing invariant at end of ubrk, which is UBRK_DONE = -1 */
 539     if (UBRK_DONE == tkn_end)
 540         tkn_end = tokenizer->buf16->utf16_len;
 541
 542     /* copy out if everything is well */
 543     if (U_FAILURE(*status))
 544         return 0;
 545
 546     /* everything OK, now update internal state */
 547     tkn_len = tkn_end - tkn_start;
 548
 549     if (0 < tkn_len)
 550     {
 551         tokenizer->token_count++;
 552         tokenizer->token_id++;
 553     } else {
 554         tokenizer->token_id = 0;
 555     }
 556     tokenizer->token_start = tkn_start;
 557     tokenizer->token_end = tkn_end;
 558
 559     /* copying into token buffer if it exists */
 560     if (tkn16){
 561         if (tkn16->utf16_cap < tkn_len)
 562             icu_buf_utf16_resize(tkn16, (size_t) tkn_len * 2);
 563
 564         u_strncpy(tkn16->utf16, &(tokenizer->buf16->utf16)[tkn_start],
 565                   tkn_len);
 566
 567         tkn16->utf16_len = tkn_len;
 568     }
 569
 570     return tkn_len;
 571 }
 572
 573 int32_t icu_tokenizer_token_id(struct icu_tokenizer * tokenizer)
 574 {
 575     return tokenizer->token_id;
 576 }
 577
 578 int32_t icu_tokenizer_token_start(struct icu_tokenizer * tokenizer)
 579 {
 580     return tokenizer->token_start;
 581 }
 582
 583 int32_t icu_tokenizer_token_end(struct icu_tokenizer * tokenizer)
 584 {
 585     return tokenizer->token_end;
 586 }
 587
 588 int32_t icu_tokenizer_token_length(struct icu_tokenizer * tokenizer)
 589 {
 590     return (tokenizer->token_end - tokenizer->token_start);
 591 }
 592
 593 int32_t icu_tokenizer_token_count(struct icu_tokenizer * tokenizer)
 594 {
 595     return tokenizer->token_count;
 596 }
 597
 598 struct icu_transform * icu_transform_create(const char *id, char action,
 599                                             const char *rules,
 600                                             UErrorCode *status)
 601 {
 602     struct icu_buf_utf16 *id16 = icu_buf_utf16_create(0);
 603     struct icu_buf_utf16 *rules16 = icu_buf_utf16_create(0);
 604
 605     struct icu_transform * transform
 606         = (struct icu_transform *) xmalloc(sizeof(struct icu_transform));
 607
 608     transform->action = action;
 609     transform->trans = 0;
 610
 611     if (id)
 612         icu_utf16_from_utf8_cstr(id16, id, status);
 613     if (rules)
 614         icu_utf16_from_utf8_cstr(rules16, rules, status);
 615
 616     switch(transform->action)
 617     {
 618     case 'f':
 619     case 'F':
 620         transform->trans
 621             = utrans_openU(id16->utf16,
 622                            id16->utf16_len,
 623                            UTRANS_FORWARD,
 624                            rules16->utf16,
 625                            rules16->utf16_len,
 626                            &transform->parse_error, status);
 627         break;
 628     case 'r':
 629     case 'R':
 630         transform->trans
 631             = utrans_openU(id16->utf16,
 632                            id16->utf16_len,
 633                            UTRANS_REVERSE ,
 634                            rules16->utf16,
 635                            rules16->utf16_len,
 636                            &transform->parse_error, status);
 637         break;
 638     default:
 639         *status = U_UNSUPPORTED_ERROR;
 640         break;
 641     }
 642     icu_buf_utf16_destroy(rules16);
 643     icu_buf_utf16_destroy(id16);
 644
 645     if (U_SUCCESS(*status))
 646         return transform;
 647
 648     /* freeing if failed */
 649     icu_transform_destroy(transform);
 650     return 0;
 651 }
 652
 653 void icu_transform_destroy(struct icu_transform * transform)
 654 {
 655     if (transform)
 656     {
 657         if (transform->trans)
 658             utrans_close(transform->trans);
 659         xfree(transform);
 660     }
 661 }
 662
 663 int icu_transform_trans(struct icu_transform * transform,
 664                         struct icu_buf_utf16 * dest16,
 665                         struct icu_buf_utf16 * src16,
 666                         UErrorCode *status)
 667 {
 668     if (!transform || !transform->trans
 669         || !src16  || !dest16)
 670         return 0;
 671
 672     if (!src16->utf16_len)
 673     {           /* guarding for empty source string */
 674         icu_buf_utf16_clear(dest16);
 675         return 0;
 676     }
 677
 678     if (!icu_buf_utf16_copy(dest16, src16))
 679         return 0;
 680
 681     utrans_transUChars (transform->trans,
 682                         dest16->utf16, &(dest16->utf16_len),
 683                         dest16->utf16_cap,
 684                         0, &(src16->utf16_len), status);
 685
 686     if (U_FAILURE(*status))
 687         icu_buf_utf16_clear(dest16);
 688
 689     return dest16->utf16_len;
 690 }
 691
 692 struct icu_chain_step * icu_chain_step_create(struct icu_chain * chain,
 693                                               enum icu_chain_step_type type,
 694                                               const uint8_t * rule,
 695                                               struct icu_buf_utf16 * buf16,
 696                                               UErrorCode *status)
 697 {
 698     struct icu_chain_step * step = 0;
 699
 700     if(!chain || !type || !rule)
 701         return 0;
 702
 703     step = (struct icu_chain_step *) xmalloc(sizeof(struct icu_chain_step));
 704
 705     step->type = type;
 706
 707     step->buf16 = buf16;
 708
 709     /* create auxilary objects */
 710     switch(step->type)
 711     {
 712     case ICU_chain_step_type_display:
 713         break;
 714     case ICU_chain_step_type_casemap:
 715         step->u.casemap = icu_casemap_create(rule[0], status);
 716         break;
 717     case ICU_chain_step_type_transform:
 718         /* rule omitted. Only ID used */
 719         step->u.transform = icu_transform_create((const char *) rule, 'f',
 720                                                  0, status);
 721         break;
 722     case ICU_chain_step_type_tokenize:
 723         step->u.tokenizer = icu_tokenizer_create((char *) chain->locale,
 724                                                  (char) rule[0], status);
 725         break;
 726     case ICU_chain_step_type_transliterate:
 727         /* we pass a dummy ID to utrans_openU.. */
 728         step->u.transform = icu_transform_create("custom", 'f',
 729                                                  (const char *) rule, status);
 730         break;
 731     default:
 732         break;
 733     }
 734     return step;
 735 }
 736
 737
 738 void icu_chain_step_destroy(struct icu_chain_step * step)
 739 {
 740     if (!step)
 741         return;
 742
 743     icu_chain_step_destroy(step->previous);
 744
 745     switch(step->type)
 746     {
 747     case ICU_chain_step_type_display:
 748         break;
 749     case ICU_chain_step_type_casemap:
 750         icu_casemap_destroy(step->u.casemap);
 751         icu_buf_utf16_destroy(step->buf16);
 752         break;
 753     case ICU_chain_step_type_transform:
 754     case ICU_chain_step_type_transliterate:
 755         icu_transform_destroy(step->u.transform);
 756         icu_buf_utf16_destroy(step->buf16);
 757         break;
 758     case ICU_chain_step_type_tokenize:
 759         icu_tokenizer_destroy(step->u.tokenizer);
 760         icu_buf_utf16_destroy(step->buf16);
 761         break;
 762     default:
 763         break;
 764     }
 765     xfree(step);
 766 }
 767
 768 struct icu_chain * icu_chain_create(const char *locale,  int sort,
 769                                     UErrorCode * status)
 770 {
 771     struct icu_chain * chain
 772         = (struct icu_chain *) xmalloc(sizeof(struct icu_chain));
 773
 774     *status = U_ZERO_ERROR;
 775
 776     chain->locale = xstrdup(locale);
 777
 778     chain->sort = sort;
 779
 780     chain->coll = ucol_open((const char *) chain->locale, status);
 781
 782     if (U_FAILURE(*status))
 783         return 0;
 784
 785     chain->token_count = 0;
 786
 787     chain->src8cstr = 0;
 788
 789     chain->display8 = icu_buf_utf8_create(0);
 790     chain->norm8 = icu_buf_utf8_create(0);
 791     chain->sort8 = icu_buf_utf8_create(0);
 792
 793     chain->src16 = icu_buf_utf16_create(0);
 794
 795     chain->steps = 0;
 796
 797     return chain;
 798 }
 799
 800 void icu_chain_destroy(struct icu_chain * chain)
 801 {
 802     if (chain)
 803     {
 804         if (chain->coll)
 805             ucol_close(chain->coll);
 806
 807         icu_buf_utf8_destroy(chain->display8);
 808         icu_buf_utf8_destroy(chain->norm8);
 809         icu_buf_utf8_destroy(chain->sort8);
 810
 811         icu_buf_utf16_destroy(chain->src16);
 812
 813         icu_chain_step_destroy(chain->steps);
 814         xfree(chain->locale);
 815         xfree(chain);
 816     }
 817 }
 818
 819 struct icu_chain * icu_chain_xml_config(const xmlNode *xml_node,
 820                                         int sort,
 821                                         UErrorCode * status)
 822 {
 823     xmlNode *node = 0;
 824     struct icu_chain * chain = 0;
 825
 826     *status = U_ZERO_ERROR;
 827
 828     if (!xml_node ||xml_node->type != XML_ELEMENT_NODE)
 829         return 0;
 830
 831     {
 832         xmlChar * xml_locale = xmlGetProp((xmlNode *) xml_node,
 833                                           (xmlChar *) "locale");
 834
 835         if (xml_locale)
 836         {
 837             chain = icu_chain_create((const char *) xml_locale, sort, status);
 838             xmlFree(xml_locale);
 839         }
 840
 841     }
 842     if (!chain)
 843         return 0;
 844
 845     for (node = xml_node->children; node; node = node->next)
 846     {
 847         xmlChar *xml_rule;
 848         struct icu_chain_step * step = 0;
 849
 850         if (node->type != XML_ELEMENT_NODE)
 851             continue;
 852
 853         xml_rule = xmlGetProp(node, (xmlChar *) "rule");
 854
 855         if (!strcmp((const char *) node->name, "casemap"))
 856             step = icu_chain_insert_step(chain, ICU_chain_step_type_casemap,
 857                                          (const uint8_t *) xml_rule, status);
 858         else if (!strcmp((const char *) node->name, "transform"))
 859             step = icu_chain_insert_step(chain, ICU_chain_step_type_transform,
 860                                          (const uint8_t *) xml_rule, status);
 861         else if (!strcmp((const char *) node->name, "transliterate"))
 862             step = icu_chain_insert_step(chain, ICU_chain_step_type_transliterate,
 863                                          (const uint8_t *) xml_rule, status);
 864         else if (!strcmp((const char *) node->name, "tokenize"))
 865             step = icu_chain_insert_step(chain, ICU_chain_step_type_tokenize,
 866                                          (const uint8_t *) xml_rule, status);
 867         else if (!strcmp((const char *) node->name, "display"))
 868             step = icu_chain_insert_step(chain, ICU_chain_step_type_display,
 869                                          (const uint8_t *) "", status);
 870         else if (!strcmp((const char *) node->name, "normalize"))
 871         {
 872             yaz_log(YLOG_WARN, "Element %s is deprecated. "
 873                     "Use transform instead", node->name);
 874             step = icu_chain_insert_step(chain, ICU_chain_step_type_transform,
 875                                          (const uint8_t *) xml_rule, status);
 876         }
 877         else if (!strcmp((const char *) node->name, "index")
 878                  || !strcmp((const char *) node->name, "sortkey"))
 879         {
 880             yaz_log(YLOG_WARN, "Element %s is no longer needed. "
 881                     "Remove it from the configuration", node->name);
 882         }
 883         else
 884         {
 885             yaz_log(YLOG_WARN, "Unknown element %s", node->name);
 886             icu_chain_destroy(chain);
 887             return 0;
 888         }
 889         xmlFree(xml_rule);
 890         if (step && U_FAILURE(*status))
 891         {
 892             icu_chain_destroy(chain);
 893             return 0;
 894         }
 895     }
 896     return chain;
 897 }
 898
 899 struct icu_chain_step * icu_chain_insert_step(struct icu_chain * chain,
 900                                               enum icu_chain_step_type type,
 901                                               const uint8_t * rule,
 902                                               UErrorCode *status)
 903 {
 904     struct icu_chain_step * step = 0;
 905     struct icu_buf_utf16 * src16 = 0;
 906     struct icu_buf_utf16 * buf16 = 0;
 907
 908     if (!chain || !type || !rule)
 909         return 0;
 910
 911     /* assign utf16 src buffers as needed */
 912     if (chain->steps && chain->steps->buf16)
 913         src16 = chain->steps->buf16;
 914     else if (chain->src16)
 915         src16 = chain->src16;
 916     else
 917         return 0;
 918
 919     /* create utf16 destination buffers as needed, or */
 920     switch(type)
 921     {
 922     case ICU_chain_step_type_display:
 923         buf16 = src16;
 924         break;
 925     case ICU_chain_step_type_casemap:
 926         buf16 = icu_buf_utf16_create(0);
 927         break;
 928     case ICU_chain_step_type_transform:
 929     case ICU_chain_step_type_transliterate:
 930         buf16 = icu_buf_utf16_create(0);
 931         break;
 932     case ICU_chain_step_type_tokenize:
 933         buf16 = icu_buf_utf16_create(0);
 934         break;
 935         break;
 936     default:
 937         break;
 938     }
 939     /* create actual chain step with this buffer */
 940     step = icu_chain_step_create(chain, type, rule, buf16, status);
 941
 942     step->previous = chain->steps;
 943     chain->steps = step;
 944
 945     return step;
 946 }
 947
 948 static int icu_chain_step_next_token(struct icu_chain * chain,
 949                                      struct icu_chain_step * step,
 950                                      UErrorCode *status)
 951 {
 952     struct icu_buf_utf16 * src16 = 0;
 953     int got_new_token = 0;
 954
 955     if (!chain || !chain->src16 || !step || !step->more_tokens)
 956         return 0;
 957
 958     /* assign utf16 src buffers as needed, advance in previous steps
 959        tokens until non-zero token met, and setting stop condition */
 960
 961     if (step->previous)
 962     {
 963         src16 = step->previous->buf16;
 964         /* tokens might be killed in previous steps, therefore looping */
 965
 966         while (step->need_new_token
 967                && step->previous->more_tokens
 968                && !got_new_token)
 969             got_new_token
 970                 = icu_chain_step_next_token(chain, step->previous, status);
 971     }
 972     else
 973     { /* first step can only work once on chain->src16 input buffer */
 974         src16 = chain->src16;
 975         step->more_tokens = 0;
 976         got_new_token = 1;
 977     }
 978
 979     if (!src16)
 980         return 0;
 981
 982     /* stop if nothing to process */
 983     if (step->need_new_token && !got_new_token)
 984     {
 985         step->more_tokens = 0;
 986         return 0;
 987     }
 988
 989     /* either an old token not finished yet, or a new token, thus
 990        perform the work, eventually put this steps output in
 991        step->buf16 or the chains UTF8 output buffers  */
 992
 993     switch(step->type)
 994     {
 995     case ICU_chain_step_type_display:
 996         icu_utf16_to_utf8(chain->display8, src16, status);
 997         break;
 998     case ICU_chain_step_type_casemap:
 999         icu_casemap_casemap(step->u.casemap,
1000                             step->buf16, src16, status,
1001                             chain->locale);
1002         break;
1003     case ICU_chain_step_type_transform:
1004     case ICU_chain_step_type_transliterate:
1005         icu_transform_trans(step->u.transform,
1006                             step->buf16, src16, status);
1007         break;
1008     case ICU_chain_step_type_tokenize:
1009         /* attach to new src16 token only first time during splitting */
1010         if (step->need_new_token)
1011         {
1012             icu_tokenizer_attach(step->u.tokenizer, src16, status);
1013             step->need_new_token = 0;
1014         }
1015
1016         /* splitting one src16 token into multiple buf16 tokens */
1017         step->more_tokens
1018             = icu_tokenizer_next_token(step->u.tokenizer,
1019                                        step->buf16, status);
1020
1021         /* make sure to get new previous token if this one had been used up
1022            by recursive call to _same_ step */
1023
1024         if (!step->more_tokens)
1025         {
1026             step->more_tokens = icu_chain_step_next_token(chain, step, status);
1027             return step->more_tokens;  /* avoid one token count too much! */
1028         }
1029         break;
1030     default:
1031         return 0;
1032         break;
1033     }
1034
1035     if (U_FAILURE(*status))
1036         return 0;
1037
1038     /* if token disappered into thin air, tell caller */
1039     /* if (!step->buf16->utf16_len && !step->more_tokens) */
1040     /*    return 0; */
1041
1042     return 1;
1043 }
1044
1045 int icu_chain_assign_cstr(struct icu_chain * chain, const char * src8cstr,
1046                           UErrorCode *status)
1047 {
1048     struct icu_chain_step * stp = 0;
1049
1050     if (!chain || !src8cstr)
1051         return 0;
1052
1053     chain->src8cstr = src8cstr;
1054
1055     stp = chain->steps;
1056
1057     /* clear token count */
1058     chain->token_count = 0;
1059
1060     /* clear all steps stop states */
1061     while (stp)
1062     {
1063         stp->more_tokens = 1;
1064         stp->need_new_token = 1;
1065         stp = stp->previous;
1066     }
1067
1068     /* finally convert UTF8 to UTF16 string if needed */
1069     if (chain->steps || chain->sort)
1070         icu_utf16_from_utf8_cstr(chain->src16, chain->src8cstr, status);
1071
1072     if (U_FAILURE(*status))
1073         return 0;
1074
1075     return 1;
1076 }
1077
1078 int icu_chain_next_token(struct icu_chain * chain, UErrorCode *status)
1079 {
1080     int got_token = 0;
1081
1082     *status = U_ZERO_ERROR;
1083
1084     if (!chain)
1085         return 0;
1086
1087     /* special case with no steps - same as index type binary */
1088     if (!chain->steps)
1089     {
1090         if (chain->token_count)
1091             return 0;
1092         else
1093         {
1094             chain->token_count++;
1095
1096             if (chain->sort)
1097                 icu_sortkey8_from_utf16(chain->coll,
1098                                         chain->sort8, chain->steps->buf16,
1099                                         status);
1100             return chain->token_count;
1101         }
1102     }
1103     /* usual case, one or more icu chain steps existing */
1104     else
1105     {
1106         while (!got_token && chain->steps && chain->steps->more_tokens)
1107             got_token = icu_chain_step_next_token(chain, chain->steps, status);
1108
1109         if (got_token)
1110         {
1111             chain->token_count++;
1112
1113             icu_utf16_to_utf8(chain->norm8, chain->steps->buf16, status);
1114
1115             if (chain->sort)
1116                 icu_sortkey8_from_utf16(chain->coll,
1117                                         chain->sort8, chain->steps->buf16,
1118                                         status);
1119             return chain->token_count;
1120         }
1121     }
1122
1123     return 0;
1124 }
1125
1126 int icu_chain_token_number(struct icu_chain * chain)
1127 {
1128     if (!chain)
1129         return 0;
1130
1131     return chain->token_count;
1132 }
1133
1134 const char * icu_chain_token_display(struct icu_chain * chain)
1135 {
1136     if (chain->display8)
1137         return icu_buf_utf8_to_cstr(chain->display8);
1138
1139     return 0;
1140 }
1141
1142 const char * icu_chain_token_norm(struct icu_chain * chain)
1143 {
1144     if (!chain->steps)
1145         return chain->src8cstr;
1146
1147     if (chain->norm8)
1148         return icu_buf_utf8_to_cstr(chain->norm8);
1149
1150     return 0;
1151 }
1152
1153 const char * icu_chain_token_sortkey(struct icu_chain * chain)
1154 {
1155     if (chain->sort8)
1156         return icu_buf_utf8_to_cstr(chain->sort8);
1157
1158     return 0;
1159 }
1160
1161 const UCollator * icu_chain_get_coll(struct icu_chain * chain)
1162 {
1163     return chain->coll;
1164 }
1165
1166 #endif /* YAZ_HAVE_ICU */
1167
1168 /*
1169  * Local variables:
1170  * c-basic-offset: 4
1171  * c-file-style: "Stroustrup"
1172  * indent-tabs-mode: nil
1173  * End:
1174  * vim: shiftwidth=4 tabstop=8 expandtab
1175  */
1176