src/icu_chain.c

   1 /* This file is part of the YAZ toolkit.
   2  * Copyright (C) Index Data
   3  * See the file LICENSE for details.
   4  */
   5
   6 /**
   7  * \file
   8  * \brief ICU chain
   9  */
  10
  11 #if HAVE_CONFIG_H
  12 #include "config.h"
  13 #endif
  14
  15 #if YAZ_HAVE_ICU
  16 #include <yaz/xmalloc.h>
  17
  18 #include <yaz/icu_I18N.h>
  19
  20 #include <yaz/stemmer.h>
  21
  22 #include <yaz/log.h>
  23 #include <yaz/nmem.h>
  24 #include <yaz/nmem_xml.h>
  25 #include <string.h>
  26 #include <stdlib.h>
  27 #include <stdio.h>
  28 #include <assert.h>
  29
  30 #include <unicode/ustring.h>  /* some more string fcns*/
  31 #include <unicode/uchar.h>    /* char names           */
  32
  33 enum icu_chain_step_type {
  34     ICU_chain_step_type_none,
  35     ICU_chain_step_type_display,        /* convert to utf8 display format */
  36     ICU_chain_step_type_casemap,        /* apply utf16 charmap */
  37     ICU_chain_step_type_transform,      /* apply utf16 transform */
  38     ICU_chain_step_type_tokenize,       /* apply utf16 tokenization */
  39     ICU_chain_step_type_transliterate,  /* apply utf16 tokenization */
  40     YAZ_chain_step_type_stemming,       /* apply utf16 stemming (YAZ) */
  41     ICU_chain_step_type_join
  42 };
  43
  44 struct icu_chain_step
  45 {
  46     /* type and action object */
  47     enum icu_chain_step_type type;
  48     union {
  49         struct icu_casemap   *casemap;
  50         struct icu_transform *transform;
  51         struct icu_tokenizer *tokenizer;
  52         yaz_stemmer_p         stemmer;
  53         struct icu_buf_utf16 *join;
  54     } u;
  55     struct icu_chain_step *previous;
  56 };
  57
  58 struct icu_chain
  59 {
  60     yaz_icu_iter_t iter;
  61     char *locale;
  62     int sort;
  63
  64     UCollator *coll;
  65
  66     /* linked list of chain steps */
  67     struct icu_chain_step *csteps;
  68 };
  69
  70 int icu_check_status(UErrorCode status)
  71 {
  72     if (U_FAILURE(status))
  73     {
  74         yaz_log(YLOG_WARN, "ICU: %d %s\n", status, u_errorName(status));
  75         return 0;
  76     }
  77     return 1;
  78 }
  79
  80 static struct icu_chain_step *icu_chain_insert_step(
  81     struct icu_chain *chain, enum icu_chain_step_type type,
  82     const char *rule, UErrorCode *status)
  83 {
  84     struct icu_chain_step *step = 0;
  85
  86     assert(chain);
  87     assert(type);
  88
  89     step = (struct icu_chain_step *) xmalloc(sizeof(*step));
  90     step->type = type;
  91
  92     switch (step->type)
  93     {
  94     case ICU_chain_step_type_display:
  95         break;
  96     case ICU_chain_step_type_casemap:
  97         assert(rule);
  98         step->u.casemap = icu_casemap_create(rule[0], status);
  99         break;
 100     case ICU_chain_step_type_transform:
 101         assert(rule);
 102         /* rule omitted. Only ID used */
 103         step->u.transform = icu_transform_create(rule, 'f', 0, status);
 104         break;
 105     case ICU_chain_step_type_tokenize:
 106         assert(rule);
 107         step->u.tokenizer = icu_tokenizer_create(chain->locale, rule[0], status);
 108         break;
 109     case ICU_chain_step_type_transliterate:
 110         assert(rule);
 111         /* we pass a dummy ID to utrans_openU.. */
 112         step->u.transform = icu_transform_create("custom", 'f', rule, status);
 113         break;
 114     case YAZ_chain_step_type_stemming:
 115         assert(rule);
 116         step->u.stemmer = yaz_stemmer_create(chain->locale, rule, status);
 117         break;
 118     case ICU_chain_step_type_join:
 119         assert(rule);
 120         step->u.join = icu_buf_utf16_create(0);
 121         icu_utf16_from_utf8_cstr(step->u.join, rule, status);
 122         break;
 123     default:
 124         break;
 125     }
 126     step->previous = chain->csteps;
 127     chain->csteps = step;
 128
 129     return step;
 130 }
 131
 132
 133 static void icu_chain_step_destroy(struct icu_chain_step *step)
 134 {
 135     if (!step)
 136         return;
 137
 138     icu_chain_step_destroy(step->previous);
 139
 140     switch (step->type)
 141     {
 142     case ICU_chain_step_type_display:
 143         break;
 144     case ICU_chain_step_type_casemap:
 145         icu_casemap_destroy(step->u.casemap);
 146         break;
 147     case ICU_chain_step_type_transform:
 148     case ICU_chain_step_type_transliterate:
 149         icu_transform_destroy(step->u.transform);
 150         break;
 151     case ICU_chain_step_type_tokenize:
 152         icu_tokenizer_destroy(step->u.tokenizer);
 153         break;
 154     case YAZ_chain_step_type_stemming:
 155         yaz_stemmer_destroy(step->u.stemmer);
 156         break;
 157     case ICU_chain_step_type_join:
 158         icu_buf_utf16_destroy(step->u.join);
 159         break;
 160     default:
 161         break;
 162     }
 163     xfree(step);
 164 }
 165
 166 struct icu_chain_step *icu_chain_step_clone(struct icu_chain_step *old)
 167 {
 168     struct icu_chain_step *step = 0;
 169     struct icu_chain_step **sp = &step;
 170     while (old)
 171     {
 172         *sp = (struct icu_chain_step *) xmalloc(sizeof(**sp));
 173         (*sp)->type = old->type;
 174
 175         switch ((*sp)->type)
 176         {
 177         case ICU_chain_step_type_display:
 178             break;
 179         case ICU_chain_step_type_casemap:
 180             (*sp)->u.casemap = icu_casemap_clone(old->u.casemap);
 181             break;
 182         case ICU_chain_step_type_transform:
 183         case ICU_chain_step_type_transliterate:
 184             (*sp)->u.transform = icu_transform_clone(old->u.transform);
 185             break;
 186         case ICU_chain_step_type_tokenize:
 187             (*sp)->u.tokenizer = icu_tokenizer_clone(old->u.tokenizer);
 188             break;
 189         case YAZ_chain_step_type_stemming:
 190             (*sp)->u.stemmer = yaz_stemmer_clone(old->u.stemmer);
 191             break;
 192         case ICU_chain_step_type_none:
 193             break;
 194         case ICU_chain_step_type_join:
 195             (*sp)->u.join = icu_buf_utf16_create(0);
 196             (*sp)->u.join = icu_buf_utf16_copy((*sp)->u.join, old->u.join);
 197             break;
 198         }
 199         old = old->previous;
 200         sp = &(*sp)->previous;
 201     }
 202     *sp = 0;
 203     return step;
 204 }
 205
 206 struct icu_chain *icu_chain_create(const char *locale, int sort,
 207                                    UErrorCode *status)
 208 {
 209     struct icu_chain *chain;
 210     UCollator *coll = ucol_open(locale, status);
 211
 212     if (U_FAILURE(*status))
 213         return 0;
 214
 215     chain = (struct icu_chain *) xmalloc(sizeof(*chain));
 216     chain->iter = 0;
 217     chain->locale = xstrdup(locale);
 218     chain->sort = sort;
 219     chain->coll = coll;
 220     chain->csteps = 0;
 221
 222     return chain;
 223 }
 224
 225 void icu_chain_destroy(struct icu_chain *chain)
 226 {
 227     if (chain)
 228     {
 229         if (chain->coll)
 230             ucol_close(chain->coll);
 231
 232         if (chain->iter)
 233             icu_iter_destroy(chain->iter);
 234         icu_chain_step_destroy(chain->csteps);
 235         xfree(chain->locale);
 236         xfree(chain);
 237     }
 238 }
 239
 240 struct icu_chain *icu_chain_xml_config(const xmlNode *xml_node,
 241                                        int sort,
 242                                        UErrorCode *status)
 243 {
 244     xmlNode *node = 0;
 245     int no_errors = 0;
 246     struct icu_chain *chain = 0;
 247     NMEM nmem = 0;
 248
 249     *status = U_ZERO_ERROR;
 250
 251     if (xml_node && xml_node->type == XML_ELEMENT_NODE)
 252     {
 253         xmlChar *xml_locale = xmlGetProp((xmlNode *) xml_node,
 254                                          (xmlChar *) "locale");
 255         if (xml_locale)
 256         {
 257             chain = icu_chain_create((const char *) xml_locale, sort, status);
 258             xmlFree(xml_locale);
 259         }
 260     }
 261
 262     if (!chain)
 263         return 0;
 264
 265     nmem = nmem_create();
 266     for (node = xml_node->children; node; node = node->next)
 267     {
 268         char *rule = 0;
 269         struct icu_chain_step *step = 0;
 270         struct _xmlAttr *attr;
 271
 272         nmem_reset(nmem);
 273         if (node->type != XML_ELEMENT_NODE)
 274             continue;
 275
 276         for (attr = node->properties; attr; attr = attr->next)
 277         {
 278             if (!strcmp((const char *) attr->name, "rule"))
 279             {
 280                 rule = nmem_text_node_cdata(attr->children, nmem);
 281             }
 282             else
 283             {
 284                 yaz_log(YLOG_WARN, "Unsupported attribute '%s' for "
 285                         "element '%s'", attr->name, node->name);
 286                 no_errors++;
 287             }
 288         }
 289         if (!rule && node->children)
 290             rule = nmem_text_node_cdata(node->children, nmem);
 291
 292         if (!rule && strcmp((const char *) node->name, "display"))
 293         {
 294             yaz_log(YLOG_WARN, "Missing attribute 'rule' for element %s",
 295                     (const char *) node->name);
 296             no_errors++;
 297             continue;
 298         }
 299         if (!strcmp((const char *) node->name, "casemap"))
 300             step = icu_chain_insert_step(chain,
 301                                          ICU_chain_step_type_casemap,
 302                                          rule, status);
 303         else if (!strcmp((const char *) node->name, "transform"))
 304             step = icu_chain_insert_step(chain,
 305                                          ICU_chain_step_type_transform,
 306                                          rule, status);
 307         else if (!strcmp((const char *) node->name, "transliterate"))
 308             step = icu_chain_insert_step(chain,
 309                                          ICU_chain_step_type_transliterate,
 310                                          rule, status);
 311         else if (!strcmp((const char *) node->name, "tokenize"))
 312             step = icu_chain_insert_step(chain, ICU_chain_step_type_tokenize,
 313                                          rule, status);
 314         else if (!strcmp((const char *) node->name, "display"))
 315             step = icu_chain_insert_step(chain, ICU_chain_step_type_display,
 316                                          rule, status);
 317         else if (!strcmp((const char *) node->name, "stemming"))
 318             step = icu_chain_insert_step(chain, YAZ_chain_step_type_stemming,
 319                                          rule, status);
 320         else if (!strcmp((const char *) node->name, "join"))
 321             step = icu_chain_insert_step(chain, ICU_chain_step_type_join,
 322                                          rule, status);
 323         else if (!strcmp((const char *) node->name, "normalize"))
 324         {
 325             yaz_log(YLOG_WARN, "Element %s is deprecated. "
 326                     "Use transform instead", node->name);
 327             step = icu_chain_insert_step(chain, ICU_chain_step_type_transform,
 328                                          rule, status);
 329         }
 330         else if (!strcmp((const char *) node->name, "index")
 331                  || !strcmp((const char *) node->name, "sortkey"))
 332         {
 333             yaz_log(YLOG_WARN, "Element %s is no longer needed. "
 334                     "Remove it from the configuration", node->name);
 335         }
 336         else
 337         {
 338             yaz_log(YLOG_WARN, "Unknown element %s", node->name);
 339             no_errors++;
 340             continue;
 341         }
 342         if (!step)
 343         {
 344             yaz_log(YLOG_WARN, "Step not created for %s", node->name);
 345             no_errors++;
 346         }
 347         if (step && U_FAILURE(*status))
 348         {
 349             yaz_log(YLOG_WARN, "ICU Error %d %s for element %s, rule %s",
 350                     *status, u_errorName(*status), node->name, rule ?
 351                     rule : "");
 352             no_errors++;
 353             break;
 354         }
 355     }
 356     nmem_destroy(nmem);
 357     if (no_errors)
 358     {
 359         icu_chain_destroy(chain);
 360         return 0;
 361     }
 362     return chain;
 363 }
 364
 365 struct icu_iter {
 366     struct icu_chain *chain;
 367     struct icu_buf_utf16 *last;
 368     struct icu_buf_utf16 *org;
 369     UErrorCode status;
 370     struct icu_buf_utf8 *display;
 371     struct icu_buf_utf8 *sort8;
 372     struct icu_buf_utf8 *result;
 373     int token_count;
 374     size_t org_start;
 375     size_t org_len;
 376     size_t utf8_base;
 377     size_t utf16_base;
 378     struct icu_chain_step *steps;
 379 };
 380
 381 void icu_utf16_print(struct icu_buf_utf16 *src16)
 382 {
 383     UErrorCode status = U_ZERO_ERROR;
 384     const char *p;
 385     struct icu_buf_utf8 *dst8 = icu_buf_utf8_create(0);
 386     icu_utf16_to_utf8(dst8, src16, &status);
 387
 388     if (U_FAILURE(status))
 389     {
 390         printf("failure");
 391     }
 392     else
 393     {
 394         p = icu_buf_utf8_to_cstr(dst8);
 395         printf("%s", p);
 396     }
 397     icu_buf_utf8_destroy(dst8);
 398 }
 399
 400 struct icu_buf_utf16 *icu_iter_invoke(yaz_icu_iter_t iter,
 401                                       struct icu_chain_step *step,
 402                                       struct icu_buf_utf16 *src)
 403 {
 404     if (!step)
 405         return src;
 406     else
 407     {
 408         struct icu_buf_utf16 *dst = icu_iter_invoke(iter, step->previous, src);
 409
 410         switch (step->type)
 411         {
 412         case ICU_chain_step_type_casemap:
 413             if (dst)
 414             {
 415                 struct icu_buf_utf16 *src = dst;
 416
 417                 dst = icu_buf_utf16_create(0);
 418                 icu_casemap_casemap(step->u.casemap, dst, src, &iter->status,
 419                                     iter->chain->locale);
 420                 icu_buf_utf16_destroy(src);
 421             }
 422             break;
 423         case ICU_chain_step_type_tokenize:
 424             if (dst)
 425             {
 426                 struct icu_buf_utf16 *src = dst;
 427
 428                 icu_tokenizer_attach(step->u.tokenizer, src, &iter->status);
 429                 iter->utf8_base = iter->utf16_base = 0;
 430                 icu_buf_utf16_copy(iter->org, src);
 431                 icu_buf_utf16_destroy(src);
 432             }
 433             dst = icu_buf_utf16_create(0);
 434             iter->status = U_ZERO_ERROR;
 435             if (!icu_tokenizer_next_token(step->u.tokenizer, dst, &iter->status,
 436                                           &iter->org_start, &iter->org_len))
 437             {
 438                 icu_buf_utf16_destroy(dst);
 439                 dst = 0;
 440             }
 441             break;
 442         case ICU_chain_step_type_transform:
 443         case ICU_chain_step_type_transliterate:
 444             if (dst)
 445             {
 446                 struct icu_buf_utf16 *src = dst;
 447                 dst = icu_buf_utf16_create(0);
 448                 icu_transform_trans(step->u.transform, dst, src, &iter->status);
 449                 icu_buf_utf16_destroy(src);
 450             }
 451             break;
 452         case ICU_chain_step_type_display:
 453             if (dst)
 454                 icu_utf16_to_utf8(iter->display, dst, &iter->status);
 455             break;
 456         case YAZ_chain_step_type_stemming:
 457             if (dst)
 458             {
 459                 struct icu_buf_utf16 *src = dst;
 460                 dst = icu_buf_utf16_create(0);
 461                 yaz_stemmer_stem(step->u.stemmer, dst, src, &iter->status);
 462                 icu_buf_utf16_destroy(src);
 463             }
 464             break;
 465         case ICU_chain_step_type_join:
 466             if (dst)
 467             {
 468                 while (1)
 469                 {
 470                     struct icu_buf_utf16 *dst1 =
 471                         icu_iter_invoke(iter, step->previous, 0);
 472
 473                     if (!dst1)
 474                         break;
 475                     dst = icu_buf_utf16_append(dst, step->u.join);
 476                     dst = icu_buf_utf16_append(dst, dst1);
 477                     icu_buf_utf16_destroy(dst1);
 478                 }
 479             }
 480             break;
 481         default:
 482             assert(0);
 483         }
 484         return dst;
 485     }
 486 }
 487
 488 yaz_icu_iter_t icu_iter_create(struct icu_chain *chain)
 489 {
 490     yaz_icu_iter_t iter = xmalloc(sizeof(*iter));
 491     iter->chain = chain;
 492     iter->status = U_ZERO_ERROR;
 493     iter->display = icu_buf_utf8_create(0);
 494     iter->sort8 = icu_buf_utf8_create(0);
 495     iter->result = icu_buf_utf8_create(0);
 496     iter->org = icu_buf_utf16_create(0);
 497     iter->last = 0; /* no last returned string (yet) */
 498     iter->steps = icu_chain_step_clone(chain->csteps);
 499     iter->token_count = 0;
 500
 501     return iter;
 502 }
 503
 504 void icu_iter_first(yaz_icu_iter_t iter, const char *src8cstr)
 505 {
 506     struct icu_buf_utf16 *src = icu_buf_utf16_create(0);
 507     icu_utf16_from_utf8_cstr(src, src8cstr, &iter->status);
 508     icu_buf_utf16_copy(iter->org, src);
 509     iter->token_count = 0;
 510     iter->org_start = 0;
 511     iter->utf8_base = iter->utf16_base = 0;
 512     iter->org_len = src->utf16_len;
 513     iter->last = icu_iter_invoke(iter, iter->steps, src);
 514 }
 515
 516 void icu_iter_destroy(yaz_icu_iter_t iter)
 517 {
 518     if (iter)
 519     {
 520         icu_buf_utf8_destroy(iter->display);
 521         icu_buf_utf8_destroy(iter->sort8);
 522         icu_buf_utf8_destroy(iter->result);
 523         icu_buf_utf16_destroy(iter->org);
 524         icu_chain_step_destroy(iter->steps);
 525         xfree(iter);
 526     }
 527 }
 528
 529 int icu_iter_next(yaz_icu_iter_t iter)
 530 {
 531     if (iter->token_count && iter->last)
 532         iter->last = icu_iter_invoke(iter, iter->steps, 0);
 533     if (!iter->last)
 534         return 0;
 535     else
 536     {
 537         iter->token_count++;
 538         if (iter->chain->sort)
 539         {
 540             icu_sortkey8_from_utf16(iter->chain->coll,
 541                                     iter->sort8, iter->last,
 542                                     &iter->status);
 543         }
 544         icu_utf16_to_utf8(iter->result, iter->last, &iter->status);
 545         icu_buf_utf16_destroy(iter->last);
 546
 547         return 1;
 548     }
 549 }
 550
 551 const char *icu_iter_get_norm(yaz_icu_iter_t iter)
 552 {
 553     return icu_buf_utf8_to_cstr(iter->result);
 554 }
 555
 556 const char *icu_iter_get_sortkey(yaz_icu_iter_t iter)
 557 {
 558     return icu_buf_utf8_to_cstr(iter->sort8);
 559 }
 560
 561 const char *icu_iter_get_display(yaz_icu_iter_t iter)
 562 {
 563     return icu_buf_utf8_to_cstr(iter->display);
 564 }
 565
 566 int icu_iter_get_token_number(yaz_icu_iter_t iter)
 567 {
 568     return iter->token_count;
 569 }
 570
 571
 572 void icu_iter_get_org_info(yaz_icu_iter_t iter, size_t *start, size_t *len)
 573 {
 574     int32_t len1 = 0, len2 = 0;
 575     UErrorCode status = U_ZERO_ERROR;
 576
 577     if (iter->org_start < iter->utf16_base)
 578     {
 579         iter->utf8_base = 0;
 580         iter->utf16_base = 0;
 581     }
 582     u_strToUTF8(0, 0, &len1,
 583                 iter->org->utf16 + iter->utf16_base,
 584                 iter->org_start - iter->utf16_base,
 585                 &status);
 586
 587     status = U_ZERO_ERROR;
 588
 589     *start = len1 + iter->utf8_base;
 590
 591     u_strToUTF8(0, 0, &len2,
 592                 iter->org->utf16 + iter->utf16_base,
 593                 iter->org_start - iter->utf16_base + iter->org_len,
 594                 &status);
 595
 596     *len = len2 - len1;
 597
 598     iter->utf8_base = *start;
 599     iter->utf16_base = iter->org_start;
 600 }
 601
 602 int icu_chain_assign_cstr(struct icu_chain *chain, const char *src8cstr,
 603                           UErrorCode *status)
 604 {
 605     if (chain->iter)
 606         icu_iter_destroy(chain->iter);
 607     chain->iter = icu_iter_create(chain);
 608     icu_iter_first(chain->iter, src8cstr);
 609     return 1;
 610 }
 611
 612 int icu_chain_next_token(struct icu_chain *chain, UErrorCode *status)
 613 {
 614     *status = U_ZERO_ERROR;
 615     return icu_iter_next(chain->iter);
 616 }
 617
 618 int icu_chain_token_number(struct icu_chain *chain)
 619 {
 620     if (chain && chain->iter)
 621         return chain->iter->token_count;
 622     return 0;
 623 }
 624
 625 const char *icu_chain_token_display(struct icu_chain *chain)
 626 {
 627     if (chain->iter)
 628         return icu_iter_get_display(chain->iter);
 629     return 0;
 630 }
 631
 632 const char *icu_chain_token_norm(struct icu_chain *chain)
 633 {
 634     if (chain->iter)
 635         return icu_iter_get_norm(chain->iter);
 636     return 0;
 637 }
 638
 639 const char *icu_chain_token_sortkey(struct icu_chain *chain)
 640 {
 641     if (chain->iter)
 642         return icu_iter_get_sortkey(chain->iter);
 643     return 0;
 644 }
 645
 646 void icu_chain_get_org_info(struct icu_chain *chain, size_t *start, size_t *len)
 647 {
 648     if (chain->iter)
 649         icu_iter_get_org_info(chain->iter, start, len);
 650 }
 651
 652
 653 #endif /* YAZ_HAVE_ICU */
 654
 655 /*
 656  * Local variables:
 657  * c-basic-offset: 4
 658  * c-file-style: "Stroustrup"
 659  * indent-tabs-mode: nil
 660  * End:
 661  * vim: shiftwidth=4 tabstop=8 expandtab
 662  */
 663