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;
86 if (!chain || !type || !rule)
89 step = (struct icu_chain_step *) xmalloc(sizeof(*step));
92 /* create auxilary objects */
95 case ICU_chain_step_type_display:
97 case ICU_chain_step_type_casemap:
98 step->u.casemap = icu_casemap_create(rule[0], status);
100 case ICU_chain_step_type_transform:
101 /* rule omitted. Only ID used */
102 step->u.transform = icu_transform_create(rule, 'f', 0, status);
104 case ICU_chain_step_type_tokenize:
105 step->u.tokenizer = icu_tokenizer_create(chain->locale, rule[0], status);
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', rule, status);
111 case YAZ_chain_step_type_stemming:
112 step->u.stemmer = yaz_stemmer_create(chain->locale, rule, status);
114 case ICU_chain_step_type_join:
115 step->u.join = icu_buf_utf16_create(0);
116 icu_utf16_from_utf8_cstr(step->u.join, rule, status);
121 step->previous = chain->csteps;
122 chain->csteps = step;
128 static void icu_chain_step_destroy(struct icu_chain_step *step)
133 icu_chain_step_destroy(step->previous);
137 case ICU_chain_step_type_display:
139 case ICU_chain_step_type_casemap:
140 icu_casemap_destroy(step->u.casemap);
142 case ICU_chain_step_type_transform:
143 case ICU_chain_step_type_transliterate:
144 icu_transform_destroy(step->u.transform);
146 case ICU_chain_step_type_tokenize:
147 icu_tokenizer_destroy(step->u.tokenizer);
149 case YAZ_chain_step_type_stemming:
150 yaz_stemmer_destroy(step->u.stemmer);
152 case ICU_chain_step_type_join:
153 icu_buf_utf16_destroy(step->u.join);
161 struct icu_chain_step *icu_chain_step_clone(struct icu_chain_step *old)
163 struct icu_chain_step *step = 0;
164 struct icu_chain_step **sp = &step;
167 *sp = (struct icu_chain_step *) xmalloc(sizeof(**sp));
168 (*sp)->type = old->type;
172 case ICU_chain_step_type_display:
174 case ICU_chain_step_type_casemap:
175 (*sp)->u.casemap = icu_casemap_clone(old->u.casemap);
177 case ICU_chain_step_type_transform:
178 case ICU_chain_step_type_transliterate:
179 (*sp)->u.transform = icu_transform_clone(old->u.transform);
181 case ICU_chain_step_type_tokenize:
182 (*sp)->u.tokenizer = icu_tokenizer_clone(old->u.tokenizer);
184 case YAZ_chain_step_type_stemming:
185 (*sp)->u.stemmer = yaz_stemmer_clone(old->u.stemmer);
187 case ICU_chain_step_type_none:
189 case ICU_chain_step_type_join:
190 (*sp)->u.join = icu_buf_utf16_create(0);
191 (*sp)->u.join = icu_buf_utf16_copy((*sp)->u.join, old->u.join);
195 sp = &(*sp)->previous;
201 struct icu_chain *icu_chain_create(const char *locale, int sort,
204 struct icu_chain *chain;
205 UCollator *coll = ucol_open(locale, status);
207 if (U_FAILURE(*status))
210 chain = (struct icu_chain *) xmalloc(sizeof(*chain));
212 chain->locale = xstrdup(locale);
220 void icu_chain_destroy(struct icu_chain *chain)
225 ucol_close(chain->coll);
228 icu_iter_destroy(chain->iter);
229 icu_chain_step_destroy(chain->csteps);
230 xfree(chain->locale);
235 struct icu_chain *icu_chain_xml_config(const xmlNode *xml_node,
241 struct icu_chain *chain = 0;
244 *status = U_ZERO_ERROR;
246 if (xml_node && xml_node->type == XML_ELEMENT_NODE)
248 xmlChar *xml_locale = xmlGetProp((xmlNode *) xml_node,
249 (xmlChar *) "locale");
252 chain = icu_chain_create((const char *) xml_locale, sort, status);
260 nmem = nmem_create();
261 for (node = xml_node->children; node; node = node->next)
264 struct icu_chain_step *step = 0;
265 struct _xmlAttr *attr;
268 if (node->type != XML_ELEMENT_NODE)
271 for (attr = node->properties; attr; attr = attr->next)
273 if (!strcmp((const char *) attr->name, "rule"))
275 rule = nmem_text_node_cdata(attr->children, nmem);
279 yaz_log(YLOG_WARN, "Unsupported attribute '%s' for "
280 "element '%s'", attr->name, node->name);
285 if (!rule && node->children)
286 rule = nmem_text_node_cdata(node->children, nmem);
288 if (!strcmp((const char *) node->name, "casemap"))
289 step = icu_chain_insert_step(chain,
290 ICU_chain_step_type_casemap,
292 else if (!strcmp((const char *) node->name, "transform"))
293 step = icu_chain_insert_step(chain,
294 ICU_chain_step_type_transform,
296 else if (!strcmp((const char *) node->name, "transliterate"))
297 step = icu_chain_insert_step(chain,
298 ICU_chain_step_type_transliterate,
300 else if (!strcmp((const char *) node->name, "tokenize"))
301 step = icu_chain_insert_step(chain, ICU_chain_step_type_tokenize,
303 else if (!strcmp((const char *) node->name, "display"))
304 step = icu_chain_insert_step(chain, ICU_chain_step_type_display,
306 else if (!strcmp((const char *) node->name, "stemming"))
307 step = icu_chain_insert_step(chain, YAZ_chain_step_type_stemming,
309 else if (!strcmp((const char *) node->name, "join"))
311 step = icu_chain_insert_step(chain, ICU_chain_step_type_join,
314 else if (!strcmp((const char *) node->name, "normalize"))
316 yaz_log(YLOG_WARN, "Element %s is deprecated. "
317 "Use transform instead", node->name);
318 step = icu_chain_insert_step(chain, ICU_chain_step_type_transform,
321 else if (!strcmp((const char *) node->name, "index")
322 || !strcmp((const char *) node->name, "sortkey"))
324 yaz_log(YLOG_WARN, "Element %s is no longer needed. "
325 "Remove it from the configuration", node->name);
329 yaz_log(YLOG_WARN, "Unknown element %s", node->name);
335 yaz_log(YLOG_WARN, "Step not created for %s", node->name);
338 if (step && U_FAILURE(*status))
347 icu_chain_destroy(chain);
354 struct icu_chain *chain;
355 struct icu_buf_utf16 *last;
357 struct icu_buf_utf8 *display;
358 struct icu_buf_utf8 *sort8;
359 struct icu_buf_utf8 *result;
360 struct icu_buf_utf16 *input;
362 struct icu_chain_step *steps;
365 void icu_utf16_print(struct icu_buf_utf16 *src16)
367 UErrorCode status = U_ZERO_ERROR;
369 struct icu_buf_utf8 *dst8 = icu_buf_utf8_create(0);
370 icu_utf16_to_utf8(dst8, src16, &status);
372 if (U_FAILURE(status))
378 p = icu_buf_utf8_to_cstr(dst8);
381 icu_buf_utf8_destroy(dst8);
384 struct icu_buf_utf16 *icu_iter_invoke(yaz_icu_iter_t iter,
385 struct icu_chain_step *step,
386 struct icu_buf_utf16 *src)
392 struct icu_buf_utf16 *dst = icu_iter_invoke(iter, step->previous, src);
396 case ICU_chain_step_type_casemap:
399 struct icu_buf_utf16 *src = dst;
401 dst = icu_buf_utf16_create(0);
402 icu_casemap_casemap(step->u.casemap, dst, src, &iter->status,
403 iter->chain->locale);
404 icu_buf_utf16_destroy(src);
407 case ICU_chain_step_type_tokenize:
410 struct icu_buf_utf16 *src = dst;
412 icu_tokenizer_attach(step->u.tokenizer, src, &iter->status);
413 icu_buf_utf16_destroy(src);
415 dst = icu_buf_utf16_create(0);
416 iter->status = U_ZERO_ERROR;
417 if (!icu_tokenizer_next_token(step->u.tokenizer, dst, &iter->status))
419 icu_buf_utf16_destroy(dst);
423 case ICU_chain_step_type_transform:
424 case ICU_chain_step_type_transliterate:
427 struct icu_buf_utf16 *src = dst;
428 dst = icu_buf_utf16_create(0);
429 icu_transform_trans(step->u.transform, dst, src, &iter->status);
430 icu_buf_utf16_destroy(src);
433 case ICU_chain_step_type_display:
435 icu_utf16_to_utf8(iter->display, dst, &iter->status);
437 case YAZ_chain_step_type_stemming:
440 struct icu_buf_utf16 *src = dst;
441 dst = icu_buf_utf16_create(0);
442 yaz_stemmer_stem(step->u.stemmer, dst, src, &iter->status);
443 icu_buf_utf16_destroy(src);
446 case ICU_chain_step_type_join:
451 struct icu_buf_utf16 *dst1 =
452 icu_iter_invoke(iter, step->previous, 0);
456 dst = icu_buf_utf16_append(dst, step->u.join);
457 dst = icu_buf_utf16_append(dst, dst1);
458 icu_buf_utf16_destroy(dst1);
469 yaz_icu_iter_t icu_iter_create(struct icu_chain *chain)
471 yaz_icu_iter_t iter = xmalloc(sizeof(*iter));
473 iter->status = U_ZERO_ERROR;
474 iter->display = icu_buf_utf8_create(0);
475 iter->sort8 = icu_buf_utf8_create(0);
476 iter->result = icu_buf_utf8_create(0);
477 iter->last = 0; /* no last returned string (yet) */
478 iter->steps = icu_chain_step_clone(chain->csteps);
484 void icu_iter_first(yaz_icu_iter_t iter, const char *src8cstr)
487 icu_buf_utf16_destroy(iter->input);
488 iter->input = icu_buf_utf16_create(0);
489 iter->token_count = 0;
490 /* fill and assign input string.. It will be 0 after
492 icu_utf16_from_utf8_cstr(iter->input, src8cstr, &iter->status);
495 void icu_iter_destroy(yaz_icu_iter_t iter)
499 icu_buf_utf8_destroy(iter->display);
500 icu_buf_utf8_destroy(iter->sort8);
501 icu_buf_utf8_destroy(iter->result);
503 icu_buf_utf16_destroy(iter->input);
504 icu_chain_step_destroy(iter->steps);
509 int icu_iter_next(yaz_icu_iter_t iter)
511 if (!iter->input && iter->last == 0)
515 /* on first call, iter->input is the input string. Thereafter: 0. */
516 assert(iter->steps || !iter->chain->csteps);
517 iter->last = icu_iter_invoke(iter, iter->steps, iter->input);
525 if (iter->chain->sort)
527 icu_sortkey8_from_utf16(iter->chain->coll,
528 iter->sort8, iter->last,
531 icu_utf16_to_utf8(iter->result, iter->last, &iter->status);
532 icu_buf_utf16_destroy(iter->last);
538 const char *icu_iter_get_norm(yaz_icu_iter_t iter)
540 return icu_buf_utf8_to_cstr(iter->result);
543 const char *icu_iter_get_sortkey(yaz_icu_iter_t iter)
545 return icu_buf_utf8_to_cstr(iter->sort8);
548 const char *icu_iter_get_display(yaz_icu_iter_t iter)
550 return icu_buf_utf8_to_cstr(iter->display);
553 int icu_iter_get_token_number(yaz_icu_iter_t iter)
555 return iter->token_count;
558 int icu_chain_assign_cstr(struct icu_chain *chain, const char *src8cstr,
562 icu_iter_destroy(chain->iter);
563 chain->iter = icu_iter_create(chain);
564 icu_iter_first(chain->iter, src8cstr);
568 int icu_chain_next_token(struct icu_chain *chain, UErrorCode *status)
570 *status = U_ZERO_ERROR;
571 return icu_iter_next(chain->iter);
574 int icu_chain_token_number(struct icu_chain *chain)
576 if (chain && chain->iter)
577 return chain->iter->token_count;
581 const char *icu_chain_token_display(struct icu_chain *chain)
584 return icu_iter_get_display(chain->iter);
588 const char *icu_chain_token_norm(struct icu_chain *chain)
591 return icu_iter_get_norm(chain->iter);
595 const char *icu_chain_token_sortkey(struct icu_chain *chain)
598 return icu_iter_get_sortkey(chain->iter);
602 #endif /* YAZ_HAVE_ICU */
607 * c-file-style: "Stroustrup"
608 * indent-tabs-mode: nil
610 * vim: shiftwidth=4 tabstop=8 expandtab