src/icu_chain.c

   1 /* This file is part of the YAZ toolkit.
   2  * Copyright (C) 1995-2013 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 };
  42
  43 struct icu_chain_step
  44 {
  45     /* type and action object */
  46     enum icu_chain_step_type type;
  47     union {
  48         struct icu_casemap   *casemap;
  49         struct icu_transform *transform;
  50         struct icu_tokenizer *tokenizer;
  51         yaz_stemmer_p         stemmer;
  52     } u;
  53     struct icu_chain_step *previous;
  54 };
  55
  56 struct icu_chain
  57 {
  58     yaz_icu_iter_t iter;
  59     char *locale;
  60     int sort;
  61
  62     UCollator *coll;
  63
  64     /* linked list of chain steps */
  65     struct icu_chain_step *csteps;
  66 };
  67
  68 int icu_check_status(UErrorCode status)
  69 {
  70     if (U_FAILURE(status))
  71     {
  72         yaz_log(YLOG_WARN, "ICU: %d %s\n", status, u_errorName(status));
  73         return 0;
  74     }
  75     return 1;
  76 }
  77
  78 static struct icu_chain_step *icu_chain_insert_step(
  79     struct icu_chain *chain, enum icu_chain_step_type type,
  80     const uint8_t *rule, UErrorCode *status)
  81 {
  82     struct icu_chain_step *step = 0;
  83
  84     if (!chain || !type || !rule)
  85         return 0;
  86
  87     step = (struct icu_chain_step *) xmalloc(sizeof(*step));
  88
  89     step->type = type;
  90     /* create auxilary objects */
  91     switch (step->type)
  92     {
  93     case ICU_chain_step_type_display:
  94         break;
  95     case ICU_chain_step_type_casemap:
  96         step->u.casemap = icu_casemap_create(rule[0], status);
  97         break;
  98     case ICU_chain_step_type_transform:
  99         /* rule omitted. Only ID used */
 100         step->u.transform = icu_transform_create((const char *) rule, 'f',
 101                                                  0, status);
 102         break;
 103     case ICU_chain_step_type_tokenize:
 104         step->u.tokenizer = icu_tokenizer_create(chain->locale,
 105                                                  (char) rule[0], status);
 106         break;
 107     case ICU_chain_step_type_transliterate:
 108         /* we pass a dummy ID to utrans_openU.. */
 109         step->u.transform = icu_transform_create("custom", 'f',
 110                                                  (const char *) rule, status);
 111         break;
 112     case YAZ_chain_step_type_stemming:
 113         step->u.stemmer = yaz_stemmer_create(chain->locale,
 114                                              (const char *) rule, status);
 115         break;
 116     default:
 117         break;
 118     }
 119     step->previous = chain->csteps;
 120     chain->csteps = step;
 121
 122     return step;
 123 }
 124
 125
 126 static void icu_chain_step_destroy(struct icu_chain_step *step)
 127 {
 128     if (!step)
 129         return;
 130
 131     icu_chain_step_destroy(step->previous);
 132
 133     switch (step->type)
 134     {
 135     case ICU_chain_step_type_display:
 136         break;
 137     case ICU_chain_step_type_casemap:
 138         icu_casemap_destroy(step->u.casemap);
 139         break;
 140     case ICU_chain_step_type_transform:
 141     case ICU_chain_step_type_transliterate:
 142         icu_transform_destroy(step->u.transform);
 143         break;
 144     case ICU_chain_step_type_tokenize:
 145         icu_tokenizer_destroy(step->u.tokenizer);
 146         break;
 147     case YAZ_chain_step_type_stemming:
 148         yaz_stemmer_destroy(step->u.stemmer);
 149         break;
 150     default:
 151         break;
 152     }
 153     xfree(step);
 154 }
 155
 156 struct icu_chain_step *icu_chain_step_clone(struct icu_chain_step *old)
 157 {
 158     struct icu_chain_step *step = 0;
 159     struct icu_chain_step **sp = &step;
 160     while (old)
 161     {
 162         *sp = (struct icu_chain_step *) xmalloc(sizeof(**sp));
 163         (*sp)->type = old->type;
 164
 165         switch ((*sp)->type)
 166         {
 167         case ICU_chain_step_type_display:
 168             break;
 169         case ICU_chain_step_type_casemap:
 170             (*sp)->u.casemap = icu_casemap_clone(old->u.casemap);
 171             break;
 172         case ICU_chain_step_type_transform:
 173         case ICU_chain_step_type_transliterate:
 174             (*sp)->u.transform = icu_transform_clone(old->u.transform);
 175             break;
 176         case ICU_chain_step_type_tokenize:
 177             (*sp)->u.tokenizer = icu_tokenizer_clone(old->u.tokenizer);
 178             break;
 179         case YAZ_chain_step_type_stemming:
 180             (*sp)->u.stemmer = yaz_stemmer_clone(old->u.stemmer);
 181             break;
 182         case ICU_chain_step_type_none:
 183             break;
 184         }
 185         old = old->previous;
 186         sp = &(*sp)->previous;
 187     }
 188     *sp = 0;
 189     return step;
 190 }
 191
 192 struct icu_chain *icu_chain_create(const char *locale, int sort,
 193                                    UErrorCode *status)
 194 {
 195     struct icu_chain *chain;
 196     UCollator *coll = ucol_open(locale, status);
 197
 198     if (U_FAILURE(*status))
 199         return 0;
 200
 201     chain = (struct icu_chain *) xmalloc(sizeof(*chain));
 202     chain->iter = 0;
 203     chain->locale = xstrdup(locale);
 204     chain->sort = sort;
 205     chain->coll = coll;
 206     chain->csteps = 0;
 207
 208     return chain;
 209 }
 210
 211 void icu_chain_destroy(struct icu_chain *chain)
 212 {
 213     if (chain)
 214     {
 215         if (chain->coll)
 216             ucol_close(chain->coll);
 217
 218         if (chain->iter)
 219             icu_iter_destroy(chain->iter);
 220         icu_chain_step_destroy(chain->csteps);
 221         xfree(chain->locale);
 222         xfree(chain);
 223     }
 224 }
 225
 226 struct icu_chain *icu_chain_xml_config(const xmlNode *xml_node,
 227                                        int sort,
 228                                        UErrorCode *status)
 229 {
 230     xmlNode *node = 0;
 231     int no_errors = 0;
 232     struct icu_chain *chain = 0;
 233     NMEM nmem = 0;
 234
 235     *status = U_ZERO_ERROR;
 236
 237     if (xml_node && xml_node->type == XML_ELEMENT_NODE)
 238     {
 239         xmlChar *xml_locale = xmlGetProp((xmlNode *) xml_node,
 240                                          (xmlChar *) "locale");
 241         if (xml_locale)
 242         {
 243             chain = icu_chain_create((const char *) xml_locale, sort, status);
 244             xmlFree(xml_locale);
 245         }
 246     }
 247
 248     if (!chain)
 249         return 0;
 250
 251     nmem = nmem_create();
 252     for (node = xml_node->children; node; node = node->next)
 253     {
 254         char *rule = 0;
 255         struct icu_chain_step *step = 0;
 256         struct _xmlAttr *attr;
 257
 258         nmem_reset(nmem);
 259         if (node->type != XML_ELEMENT_NODE)
 260             continue;
 261
 262         for (attr = node->properties; attr; attr = attr->next)
 263         {
 264             if (!strcmp((const char *) attr->name, "rule"))
 265             {
 266                 rule = nmem_text_node_cdata(attr->children, nmem);
 267             }
 268             else
 269             {
 270                 yaz_log(YLOG_WARN, "Unsupported attribute '%s' for "
 271                         "element '%s'", attr->name, node->name);
 272                 no_errors++;
 273                 continue;
 274             }
 275         }
 276         if (!rule && node->children)
 277             rule = nmem_text_node_cdata(node->children, nmem);
 278
 279         if (!strcmp((const char *) node->name, "casemap"))
 280             step = icu_chain_insert_step(chain, ICU_chain_step_type_casemap,
 281                                          (const uint8_t *) rule, status);
 282         else if (!strcmp((const char *) node->name, "transform"))
 283             step = icu_chain_insert_step(chain, ICU_chain_step_type_transform,
 284                                          (const uint8_t *) rule, status);
 285         else if (!strcmp((const char *) node->name, "transliterate"))
 286             step = icu_chain_insert_step(chain, ICU_chain_step_type_transliterate,
 287                                          (const uint8_t *) rule, status);
 288         else if (!strcmp((const char *) node->name, "tokenize"))
 289             step = icu_chain_insert_step(chain, ICU_chain_step_type_tokenize,
 290                                          (const uint8_t *) rule, status);
 291         else if (!strcmp((const char *) node->name, "display"))
 292             step = icu_chain_insert_step(chain, ICU_chain_step_type_display,
 293                                          (const uint8_t *) "", status);
 294         else if (!strcmp((const char *) node->name, "stemming"))
 295             step = icu_chain_insert_step(chain, YAZ_chain_step_type_stemming,
 296                                          (const uint8_t *) rule, status);
 297         else if (!strcmp((const char *) node->name, "normalize"))
 298         {
 299             yaz_log(YLOG_WARN, "Element %s is deprecated. "
 300                     "Use transform instead", node->name);
 301             step = icu_chain_insert_step(chain, ICU_chain_step_type_transform,
 302                                          (const uint8_t *) rule, status);
 303         }
 304         else if (!strcmp((const char *) node->name, "index")
 305                  || !strcmp((const char *) node->name, "sortkey"))
 306         {
 307             yaz_log(YLOG_WARN, "Element %s is no longer needed. "
 308                     "Remove it from the configuration", node->name);
 309         }
 310         else
 311         {
 312             yaz_log(YLOG_WARN, "Unknown element %s", node->name);
 313             no_errors++;
 314             continue;
 315         }
 316         if (step && U_FAILURE(*status))
 317         {
 318             no_errors++;
 319             break;
 320         }
 321     }
 322     nmem_destroy(nmem);
 323     if (no_errors)
 324     {
 325         icu_chain_destroy(chain);
 326         return 0;
 327     }
 328     return chain;
 329 }
 330
 331 struct icu_iter {
 332     struct icu_chain *chain;
 333     struct icu_buf_utf16 *last;
 334     UErrorCode status;
 335     struct icu_buf_utf8 *display;
 336     struct icu_buf_utf8 *sort8;
 337     struct icu_buf_utf8 *result;
 338     struct icu_buf_utf16 *input;
 339     int token_count;
 340     struct icu_chain_step *steps;
 341 };
 342
 343 void icu_utf16_print(struct icu_buf_utf16 *src16)
 344 {
 345     UErrorCode status = U_ZERO_ERROR;
 346     const char *p;
 347     struct icu_buf_utf8 *dst8 = icu_buf_utf8_create(0);
 348     icu_utf16_to_utf8(dst8, src16, &status);
 349
 350     if (U_FAILURE(status))
 351     {
 352         printf("failure");
 353     }
 354     else
 355     {
 356         p = icu_buf_utf8_to_cstr(dst8);
 357         printf("%s", p);
 358     }
 359     icu_buf_utf8_destroy(dst8);
 360 }
 361
 362 struct icu_buf_utf16 *icu_iter_invoke(yaz_icu_iter_t iter,
 363                                       struct icu_chain_step *step,
 364                                       struct icu_buf_utf16 *src)
 365 {
 366     if (!step)
 367         return src;
 368     else
 369     {
 370         struct icu_buf_utf16 *dst = icu_iter_invoke(iter, step->previous, src);
 371
 372         switch (step->type)
 373         {
 374         case ICU_chain_step_type_casemap:
 375             if (dst)
 376             {
 377                 struct icu_buf_utf16 *src = dst;
 378
 379                 dst = icu_buf_utf16_create(0);
 380                 icu_casemap_casemap(step->u.casemap, dst, src, &iter->status,
 381                                     iter->chain->locale);
 382                 icu_buf_utf16_destroy(src);
 383             }
 384             break;
 385         case ICU_chain_step_type_tokenize:
 386             if (dst)
 387             {
 388                 struct icu_buf_utf16 *src = dst;
 389
 390                 icu_tokenizer_attach(step->u.tokenizer, src, &iter->status);
 391                 icu_buf_utf16_destroy(src);
 392             }
 393             dst = icu_buf_utf16_create(0);
 394             iter->status = U_ZERO_ERROR;
 395             if (!icu_tokenizer_next_token(step->u.tokenizer, dst, &iter->status))
 396             {
 397                 icu_buf_utf16_destroy(dst);
 398                 dst = 0;
 399             }
 400             break;
 401         case ICU_chain_step_type_transform:
 402         case ICU_chain_step_type_transliterate:
 403             if (dst)
 404             {
 405                 struct icu_buf_utf16 *src = dst;
 406                 dst = icu_buf_utf16_create(0);
 407                 icu_transform_trans(step->u.transform, dst, src, &iter->status);
 408                 icu_buf_utf16_destroy(src);
 409             }
 410             break;
 411         case ICU_chain_step_type_display:
 412             if (dst)
 413                 icu_utf16_to_utf8(iter->display, dst, &iter->status);
 414             break;
 415         case YAZ_chain_step_type_stemming:
 416             if (dst)
 417             {
 418                 struct icu_buf_utf16 *src = dst;
 419                 dst = icu_buf_utf16_create(0);
 420                 yaz_stemmer_stem(step->u.stemmer, dst, src, &iter->status);
 421                 icu_buf_utf16_destroy(src);
 422             }
 423             break;
 424         default:
 425             assert(0);
 426         }
 427         return dst;
 428     }
 429 }
 430
 431 yaz_icu_iter_t icu_iter_create(struct icu_chain *chain)
 432 {
 433     yaz_icu_iter_t iter = xmalloc(sizeof(*iter));
 434     iter->chain = chain;
 435     iter->status = U_ZERO_ERROR;
 436     iter->display = icu_buf_utf8_create(0);
 437     iter->sort8 = icu_buf_utf8_create(0);
 438     iter->result = icu_buf_utf8_create(0);
 439     iter->last = 0; /* no last returned string (yet) */
 440     iter->steps = icu_chain_step_clone(chain->csteps);
 441     iter->input = 0;
 442
 443     return iter;
 444 }
 445
 446 void icu_iter_first(yaz_icu_iter_t iter, const char *src8cstr)
 447 {
 448     if (iter->input)
 449         icu_buf_utf16_destroy(iter->input);
 450     iter->input = icu_buf_utf16_create(0);
 451     iter->token_count = 0;
 452     /* fill and assign input string.. It will be 0 after
 453        first iteration */
 454     icu_utf16_from_utf8_cstr(iter->input, src8cstr, &iter->status);
 455 }
 456
 457 void icu_iter_destroy(yaz_icu_iter_t iter)
 458 {
 459     if (iter)
 460     {
 461         icu_buf_utf8_destroy(iter->display);
 462         icu_buf_utf8_destroy(iter->sort8);
 463         icu_buf_utf8_destroy(iter->result);
 464         if (iter->input)
 465             icu_buf_utf16_destroy(iter->input);
 466         icu_chain_step_destroy(iter->steps);
 467         xfree(iter);
 468     }
 469 }
 470
 471 int icu_iter_next(yaz_icu_iter_t iter)
 472 {
 473     if (!iter->input && iter->last == 0)
 474         return 0;
 475     else
 476     {
 477         /* on first call, iter->input is the input string. Thereafter: 0. */
 478         assert(iter->steps || !iter->chain->csteps);
 479         iter->last = icu_iter_invoke(iter, iter->steps, iter->input);
 480         iter->input = 0;
 481
 482         if (!iter->last)
 483             return 0;
 484
 485         iter->token_count++;
 486
 487         if (iter->chain->sort)
 488         {
 489             icu_sortkey8_from_utf16(iter->chain->coll,
 490                                     iter->sort8, iter->last,
 491                                     &iter->status);
 492         }
 493         icu_utf16_to_utf8(iter->result, iter->last, &iter->status);
 494         icu_buf_utf16_destroy(iter->last);
 495
 496         return 1;
 497     }
 498 }
 499
 500 const char *icu_iter_get_norm(yaz_icu_iter_t iter)
 501 {
 502     return icu_buf_utf8_to_cstr(iter->result);
 503 }
 504
 505 const char *icu_iter_get_sortkey(yaz_icu_iter_t iter)
 506 {
 507     return icu_buf_utf8_to_cstr(iter->sort8);
 508 }
 509
 510 const char *icu_iter_get_display(yaz_icu_iter_t iter)
 511 {
 512     return icu_buf_utf8_to_cstr(iter->display);
 513 }
 514
 515 int icu_iter_get_token_number(yaz_icu_iter_t iter)
 516 {
 517     return iter->token_count;
 518 }
 519
 520 int icu_chain_assign_cstr(struct icu_chain *chain, const char *src8cstr,
 521                           UErrorCode *status)
 522 {
 523     if (chain->iter)
 524         icu_iter_destroy(chain->iter);
 525     chain->iter = icu_iter_create(chain);
 526     icu_iter_first(chain->iter, src8cstr);
 527     return 1;
 528 }
 529
 530 int icu_chain_next_token(struct icu_chain *chain, UErrorCode *status)
 531 {
 532     *status = U_ZERO_ERROR;
 533     return icu_iter_next(chain->iter);
 534 }
 535
 536 int icu_chain_token_number(struct icu_chain *chain)
 537 {
 538     if (chain && chain->iter)
 539         return chain->iter->token_count;
 540     return 0;
 541 }
 542
 543 const char *icu_chain_token_display(struct icu_chain *chain)
 544 {
 545     if (chain->iter)
 546         return icu_iter_get_display(chain->iter);
 547     return 0;
 548 }
 549
 550 const char *icu_chain_token_norm(struct icu_chain *chain)
 551 {
 552     if (chain->iter)
 553         return icu_iter_get_norm(chain->iter);
 554     return 0;
 555 }
 556
 557 const char *icu_chain_token_sortkey(struct icu_chain *chain)
 558 {
 559     if (chain->iter)
 560         return icu_iter_get_sortkey(chain->iter);
 561     return 0;
 562 }
 563
 564 #endif /* YAZ_HAVE_ICU */
 565
 566 /*
 567  * Local variables:
 568  * c-basic-offset: 4
 569  * c-file-style: "Stroustrup"
 570  * indent-tabs-mode: nil
 571  * End:
 572  * vim: shiftwidth=4 tabstop=8 expandtab
 573  */
 574