1 /* This file is part of the YAZ toolkit.
2 * Copyright (C) 1995-2013 Index Data
3 * See the file LICENSE for details.
16 #include <yaz/xmalloc.h>
18 #include <yaz/icu_I18N.h>
20 #include <yaz/stemmer.h>
24 #include <yaz/nmem_xml.h>
30 #include <unicode/ustring.h> /* some more string fcns*/
31 #include <unicode/uchar.h> /* char names */
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
46 /* type and action object */
47 enum icu_chain_step_type type;
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;
55 struct icu_chain_step *previous;
66 /* linked list of chain steps */
67 struct icu_chain_step *csteps;
70 int icu_check_status(UErrorCode status)
72 if (U_FAILURE(status))
74 yaz_log(YLOG_WARN, "ICU: %d %s\n", status, u_errorName(status));
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)
84 struct icu_chain_step *step = 0;
89 step = (struct icu_chain_step *) xmalloc(sizeof(*step));
94 case ICU_chain_step_type_display:
96 case ICU_chain_step_type_casemap:
98 step->u.casemap = icu_casemap_create(rule[0], status);
100 case ICU_chain_step_type_transform:
102 /* rule omitted. Only ID used */
103 step->u.transform = icu_transform_create(rule, 'f', 0, status);
105 case ICU_chain_step_type_tokenize:
107 step->u.tokenizer = icu_tokenizer_create(chain->locale, rule[0], status);
109 case ICU_chain_step_type_transliterate:
111 /* we pass a dummy ID to utrans_openU.. */
112 step->u.transform = icu_transform_create("custom", 'f', rule, status);
114 case YAZ_chain_step_type_stemming:
116 step->u.stemmer = yaz_stemmer_create(chain->locale, rule, status);
118 case ICU_chain_step_type_join:
120 step->u.join = icu_buf_utf16_create(0);
121 icu_utf16_from_utf8_cstr(step->u.join, rule, status);
126 step->previous = chain->csteps;
127 chain->csteps = step;
133 static void icu_chain_step_destroy(struct icu_chain_step *step)
138 icu_chain_step_destroy(step->previous);
142 case ICU_chain_step_type_display:
144 case ICU_chain_step_type_casemap:
145 icu_casemap_destroy(step->u.casemap);
147 case ICU_chain_step_type_transform:
148 case ICU_chain_step_type_transliterate:
149 icu_transform_destroy(step->u.transform);
151 case ICU_chain_step_type_tokenize:
152 icu_tokenizer_destroy(step->u.tokenizer);
154 case YAZ_chain_step_type_stemming:
155 yaz_stemmer_destroy(step->u.stemmer);
157 case ICU_chain_step_type_join:
158 icu_buf_utf16_destroy(step->u.join);
166 struct icu_chain_step *icu_chain_step_clone(struct icu_chain_step *old)
168 struct icu_chain_step *step = 0;
169 struct icu_chain_step **sp = &step;
172 *sp = (struct icu_chain_step *) xmalloc(sizeof(**sp));
173 (*sp)->type = old->type;
177 case ICU_chain_step_type_display:
179 case ICU_chain_step_type_casemap:
180 (*sp)->u.casemap = icu_casemap_clone(old->u.casemap);
182 case ICU_chain_step_type_transform:
183 case ICU_chain_step_type_transliterate:
184 (*sp)->u.transform = icu_transform_clone(old->u.transform);
186 case ICU_chain_step_type_tokenize:
187 (*sp)->u.tokenizer = icu_tokenizer_clone(old->u.tokenizer);
189 case YAZ_chain_step_type_stemming:
190 (*sp)->u.stemmer = yaz_stemmer_clone(old->u.stemmer);
192 case ICU_chain_step_type_none:
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);
200 sp = &(*sp)->previous;
206 struct icu_chain *icu_chain_create(const char *locale, int sort,
209 struct icu_chain *chain;
210 UCollator *coll = ucol_open(locale, status);
212 if (U_FAILURE(*status))
215 chain = (struct icu_chain *) xmalloc(sizeof(*chain));
217 chain->locale = xstrdup(locale);
225 void icu_chain_destroy(struct icu_chain *chain)
230 ucol_close(chain->coll);
233 icu_iter_destroy(chain->iter);
234 icu_chain_step_destroy(chain->csteps);
235 xfree(chain->locale);
240 struct icu_chain *icu_chain_xml_config(const xmlNode *xml_node,
246 struct icu_chain *chain = 0;
249 *status = U_ZERO_ERROR;
251 if (xml_node && xml_node->type == XML_ELEMENT_NODE)
253 xmlChar *xml_locale = xmlGetProp((xmlNode *) xml_node,
254 (xmlChar *) "locale");
257 chain = icu_chain_create((const char *) xml_locale, sort, status);
265 nmem = nmem_create();
266 for (node = xml_node->children; node; node = node->next)
269 struct icu_chain_step *step = 0;
270 struct _xmlAttr *attr;
273 if (node->type != XML_ELEMENT_NODE)
276 for (attr = node->properties; attr; attr = attr->next)
278 if (!strcmp((const char *) attr->name, "rule"))
280 rule = nmem_text_node_cdata(attr->children, nmem);
284 yaz_log(YLOG_WARN, "Unsupported attribute '%s' for "
285 "element '%s'", attr->name, node->name);
289 if (!rule && node->children)
290 rule = nmem_text_node_cdata(node->children, nmem);
292 if (!rule && strcmp((const char *) node->name, "display"))
294 yaz_log(YLOG_WARN, "Missing attribute rule for element %s",
295 (const char *) node->name);
299 if (!strcmp((const char *) node->name, "casemap"))
300 step = icu_chain_insert_step(chain,
301 ICU_chain_step_type_casemap,
303 else if (!strcmp((const char *) node->name, "transform"))
304 step = icu_chain_insert_step(chain,
305 ICU_chain_step_type_transform,
307 else if (!strcmp((const char *) node->name, "transliterate"))
308 step = icu_chain_insert_step(chain,
309 ICU_chain_step_type_transliterate,
311 else if (!strcmp((const char *) node->name, "tokenize"))
312 step = icu_chain_insert_step(chain, ICU_chain_step_type_tokenize,
314 else if (!strcmp((const char *) node->name, "display"))
315 step = icu_chain_insert_step(chain, ICU_chain_step_type_display,
317 else if (!strcmp((const char *) node->name, "stemming"))
318 step = icu_chain_insert_step(chain, YAZ_chain_step_type_stemming,
320 else if (!strcmp((const char *) node->name, "join"))
321 step = icu_chain_insert_step(chain, ICU_chain_step_type_join,
323 else if (!strcmp((const char *) node->name, "normalize"))
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,
330 else if (!strcmp((const char *) node->name, "index")
331 || !strcmp((const char *) node->name, "sortkey"))
333 yaz_log(YLOG_WARN, "Element %s is no longer needed. "
334 "Remove it from the configuration", node->name);
338 yaz_log(YLOG_WARN, "Unknown element %s", node->name);
344 yaz_log(YLOG_WARN, "Step not created for %s", node->name);
347 if (step && U_FAILURE(*status))
356 icu_chain_destroy(chain);
363 struct icu_chain *chain;
364 struct icu_buf_utf16 *last;
365 struct icu_buf_utf16 *org;
367 struct icu_buf_utf8 *display;
368 struct icu_buf_utf8 *sort8;
369 struct icu_buf_utf8 *result;
373 struct icu_chain_step *steps;
376 void icu_utf16_print(struct icu_buf_utf16 *src16)
378 UErrorCode status = U_ZERO_ERROR;
380 struct icu_buf_utf8 *dst8 = icu_buf_utf8_create(0);
381 icu_utf16_to_utf8(dst8, src16, &status);
383 if (U_FAILURE(status))
389 p = icu_buf_utf8_to_cstr(dst8);
392 icu_buf_utf8_destroy(dst8);
395 struct icu_buf_utf16 *icu_iter_invoke(yaz_icu_iter_t iter,
396 struct icu_chain_step *step,
397 struct icu_buf_utf16 *src)
403 struct icu_buf_utf16 *dst = icu_iter_invoke(iter, step->previous, src);
407 case ICU_chain_step_type_casemap:
410 struct icu_buf_utf16 *src = dst;
412 dst = icu_buf_utf16_create(0);
413 icu_casemap_casemap(step->u.casemap, dst, src, &iter->status,
414 iter->chain->locale);
415 icu_buf_utf16_destroy(src);
418 case ICU_chain_step_type_tokenize:
421 struct icu_buf_utf16 *src = dst;
423 icu_tokenizer_attach(step->u.tokenizer, src, &iter->status);
424 icu_buf_utf16_destroy(src);
426 dst = icu_buf_utf16_create(0);
427 iter->status = U_ZERO_ERROR;
428 if (!icu_tokenizer_next_token(step->u.tokenizer, dst, &iter->status,
429 &iter->org_start, &iter->org_len))
431 icu_buf_utf16_destroy(dst);
435 case ICU_chain_step_type_transform:
436 case ICU_chain_step_type_transliterate:
439 struct icu_buf_utf16 *src = dst;
440 dst = icu_buf_utf16_create(0);
441 icu_transform_trans(step->u.transform, dst, src, &iter->status);
442 icu_buf_utf16_destroy(src);
445 case ICU_chain_step_type_display:
447 icu_utf16_to_utf8(iter->display, dst, &iter->status);
449 case YAZ_chain_step_type_stemming:
452 struct icu_buf_utf16 *src = dst;
453 dst = icu_buf_utf16_create(0);
454 yaz_stemmer_stem(step->u.stemmer, dst, src, &iter->status);
455 icu_buf_utf16_destroy(src);
458 case ICU_chain_step_type_join:
463 struct icu_buf_utf16 *dst1 =
464 icu_iter_invoke(iter, step->previous, 0);
468 dst = icu_buf_utf16_append(dst, step->u.join);
469 dst = icu_buf_utf16_append(dst, dst1);
470 icu_buf_utf16_destroy(dst1);
481 yaz_icu_iter_t icu_iter_create(struct icu_chain *chain)
483 yaz_icu_iter_t iter = xmalloc(sizeof(*iter));
485 iter->status = U_ZERO_ERROR;
486 iter->display = icu_buf_utf8_create(0);
487 iter->sort8 = icu_buf_utf8_create(0);
488 iter->result = icu_buf_utf8_create(0);
489 iter->org = icu_buf_utf16_create(0);
490 iter->last = 0; /* no last returned string (yet) */
491 iter->steps = icu_chain_step_clone(chain->csteps);
492 iter->token_count = 0;
497 void icu_iter_first(yaz_icu_iter_t iter, const char *src8cstr)
499 struct icu_buf_utf16 *src = icu_buf_utf16_create(0);
500 icu_utf16_from_utf8_cstr(src, src8cstr, &iter->status);
501 icu_buf_utf16_copy(iter->org, src);
502 iter->token_count = 0;
504 iter->org_len = src->utf16_len;
505 iter->last = icu_iter_invoke(iter, iter->steps, src);
508 void icu_iter_destroy(yaz_icu_iter_t iter)
512 icu_buf_utf8_destroy(iter->display);
513 icu_buf_utf8_destroy(iter->sort8);
514 icu_buf_utf8_destroy(iter->result);
515 icu_buf_utf16_destroy(iter->org);
516 icu_chain_step_destroy(iter->steps);
521 int icu_iter_next(yaz_icu_iter_t iter)
523 if (iter->token_count && iter->last)
524 iter->last = icu_iter_invoke(iter, iter->steps, 0);
530 if (iter->chain->sort)
532 icu_sortkey8_from_utf16(iter->chain->coll,
533 iter->sort8, iter->last,
536 icu_utf16_to_utf8(iter->result, iter->last, &iter->status);
537 icu_buf_utf16_destroy(iter->last);
543 const char *icu_iter_get_norm(yaz_icu_iter_t iter)
545 return icu_buf_utf8_to_cstr(iter->result);
548 const char *icu_iter_get_sortkey(yaz_icu_iter_t iter)
550 return icu_buf_utf8_to_cstr(iter->sort8);
553 const char *icu_iter_get_display(yaz_icu_iter_t iter)
555 return icu_buf_utf8_to_cstr(iter->display);
558 int icu_iter_get_token_number(yaz_icu_iter_t iter)
560 return iter->token_count;
564 void icu_iter_get_org_info(yaz_icu_iter_t iter, size_t *start, size_t *len)
566 /* save full length of org since we're gonna cut it */
567 int32_t save_len = iter->org->utf16_len;
569 struct icu_buf_utf8 *tmp = icu_buf_utf8_create(0);
570 UErrorCode status = U_ZERO_ERROR;
572 iter->org->utf16_len = iter->org_start;
573 icu_utf16_to_utf8(tmp, iter->org, &status);
574 if (U_SUCCESS(status))
575 *start = tmp->utf8_len;
578 iter->org->utf16_len = iter->org_start + iter->org_len;
579 icu_utf16_to_utf8(tmp, iter->org, &status);
580 if (U_SUCCESS(status))
581 *len = tmp->utf8_len - *start;
584 icu_buf_utf8_destroy(tmp);
585 iter->org->utf16_len = save_len;
588 int icu_chain_assign_cstr(struct icu_chain *chain, const char *src8cstr,
592 icu_iter_destroy(chain->iter);
593 chain->iter = icu_iter_create(chain);
594 icu_iter_first(chain->iter, src8cstr);
598 int icu_chain_next_token(struct icu_chain *chain, UErrorCode *status)
600 *status = U_ZERO_ERROR;
601 return icu_iter_next(chain->iter);
604 int icu_chain_token_number(struct icu_chain *chain)
606 if (chain && chain->iter)
607 return chain->iter->token_count;
611 const char *icu_chain_token_display(struct icu_chain *chain)
614 return icu_iter_get_display(chain->iter);
618 const char *icu_chain_token_norm(struct icu_chain *chain)
621 return icu_iter_get_norm(chain->iter);
625 const char *icu_chain_token_sortkey(struct icu_chain *chain)
628 return icu_iter_get_sortkey(chain->iter);
632 void icu_chain_get_org_info(struct icu_chain *chain, size_t *start, size_t *len)
635 icu_iter_get_org_info(chain->iter, start, len);
639 #endif /* YAZ_HAVE_ICU */
644 * c-file-style: "Stroustrup"
645 * indent-tabs-mode: nil
647 * vim: shiftwidth=4 tabstop=8 expandtab