/* This file is part of the YAZ toolkit.
- * Copyright (C) 1995-2010 Index Data
+ * Copyright (C) Index Data
* See the file LICENSE for details.
*/
#include <yaz/icu_I18N.h>
+#include <yaz/stemmer.h>
+
#include <yaz/log.h>
#include <yaz/nmem.h>
#include <yaz/nmem_xml.h>
enum icu_chain_step_type {
ICU_chain_step_type_none,
- ICU_chain_step_type_display, /* convert to utf8 display format */
- ICU_chain_step_type_casemap, /* apply utf16 charmap */
- ICU_chain_step_type_transform, /* apply utf16 transform */
- ICU_chain_step_type_tokenize, /* apply utf16 tokenization */
- ICU_chain_step_type_transliterate /* apply utf16 tokenization */
+ ICU_chain_step_type_display, /* convert to utf8 display format */
+ ICU_chain_step_type_casemap, /* apply utf16 charmap */
+ ICU_chain_step_type_transform, /* apply utf16 transform */
+ ICU_chain_step_type_tokenize, /* apply utf16 tokenization */
+ ICU_chain_step_type_transliterate, /* apply utf16 tokenization */
+ YAZ_chain_step_type_stemming, /* apply utf16 stemming (YAZ) */
+ ICU_chain_step_type_join
};
struct icu_chain_step
/* type and action object */
enum icu_chain_step_type type;
union {
- struct icu_casemap *casemap;
+ struct icu_casemap *casemap;
struct icu_transform *transform;
- struct icu_tokenizer *tokenizer;
+ struct icu_tokenizer *tokenizer;
+ yaz_stemmer_p stemmer;
+ struct icu_buf_utf16 *join;
} u;
struct icu_chain_step *previous;
};
int sort;
UCollator *coll;
-
+
/* linked list of chain steps */
struct icu_chain_step *csteps;
};
if (U_FAILURE(status))
{
yaz_log(YLOG_WARN, "ICU: %d %s\n", status, u_errorName(status));
- return 0;
+ return 0;
}
return 1;
}
-static struct icu_chain_step *icu_chain_step_create(
- struct icu_chain *chain, enum icu_chain_step_type type,
- const uint8_t *rule, UErrorCode *status)
+static struct icu_chain_step *icu_chain_insert_step(
+ struct icu_chain *chain, enum icu_chain_step_type type,
+ const char *rule, UErrorCode *status)
{
struct icu_chain_step *step = 0;
-
- if (!chain || !type || !rule)
- return 0;
- step = (struct icu_chain_step *) xmalloc(sizeof(*step));
+ assert(chain);
+ assert(type);
+ step = (struct icu_chain_step *) xmalloc(sizeof(*step));
step->type = type;
- /* create auxilary objects */
+
switch (step->type)
{
case ICU_chain_step_type_display:
break;
case ICU_chain_step_type_casemap:
+ assert(rule);
step->u.casemap = icu_casemap_create(rule[0], status);
break;
case ICU_chain_step_type_transform:
+ assert(rule);
/* rule omitted. Only ID used */
- step->u.transform = icu_transform_create((const char *) rule, 'f',
- 0, status);
+ step->u.transform = icu_transform_create(rule, 'f', 0, status);
break;
case ICU_chain_step_type_tokenize:
- step->u.tokenizer = icu_tokenizer_create((char *) chain->locale,
- (char) rule[0], status);
+ assert(rule);
+ step->u.tokenizer = icu_tokenizer_create(chain->locale, rule[0], status);
break;
case ICU_chain_step_type_transliterate:
+ assert(rule);
/* we pass a dummy ID to utrans_openU.. */
- step->u.transform = icu_transform_create("custom", 'f',
- (const char *) rule, status);
+ step->u.transform = icu_transform_create("custom", 'f', rule, status);
+ break;
+ case YAZ_chain_step_type_stemming:
+ assert(rule);
+ step->u.stemmer = yaz_stemmer_create(chain->locale, rule, status);
+ break;
+ case ICU_chain_step_type_join:
+ assert(rule);
+ step->u.join = icu_buf_utf16_create(0);
+ icu_utf16_from_utf8_cstr(step->u.join, rule, status);
break;
default:
break;
}
+ step->previous = chain->csteps;
+ chain->csteps = step;
+
return step;
}
case ICU_chain_step_type_tokenize:
icu_tokenizer_destroy(step->u.tokenizer);
break;
+ case YAZ_chain_step_type_stemming:
+ yaz_stemmer_destroy(step->u.stemmer);
+ break;
+ case ICU_chain_step_type_join:
+ icu_buf_utf16_destroy(step->u.join);
+ break;
default:
break;
}
{
*sp = (struct icu_chain_step *) xmalloc(sizeof(**sp));
(*sp)->type = old->type;
-
+
switch ((*sp)->type)
{
case ICU_chain_step_type_display:
case ICU_chain_step_type_tokenize:
(*sp)->u.tokenizer = icu_tokenizer_clone(old->u.tokenizer);
break;
+ case YAZ_chain_step_type_stemming:
+ (*sp)->u.stemmer = yaz_stemmer_clone(old->u.stemmer);
+ break;
case ICU_chain_step_type_none:
break;
+ case ICU_chain_step_type_join:
+ (*sp)->u.join = icu_buf_utf16_create(0);
+ (*sp)->u.join = icu_buf_utf16_copy((*sp)->u.join, old->u.join);
+ break;
}
old = old->previous;
sp = &(*sp)->previous;
struct icu_chain *icu_chain_create(const char *locale, int sort,
UErrorCode *status)
{
- struct icu_chain *chain
- = (struct icu_chain *) xmalloc(sizeof(*chain));
+ struct icu_chain *chain;
+ UCollator *coll = ucol_open(locale, status);
- *status = U_ZERO_ERROR;
+ if (U_FAILURE(*status))
+ return 0;
+ chain = (struct icu_chain *) xmalloc(sizeof(*chain));
chain->iter = 0;
chain->locale = xstrdup(locale);
-
chain->sort = sort;
-
- chain->coll = ucol_open((const char *) chain->locale, status);
-
- if (U_FAILURE(*status))
- return 0;
-
+ chain->coll = coll;
chain->csteps = 0;
return chain;
}
}
-static struct icu_chain_step *icu_chain_insert_step(
- struct icu_chain *chain, enum icu_chain_step_type type,
- const uint8_t *rule, UErrorCode *status);
-
-struct icu_chain *icu_chain_xml_config(const xmlNode *xml_node,
+struct icu_chain *icu_chain_xml_config(const xmlNode *xml_node,
int sort,
UErrorCode *status)
{
int no_errors = 0;
struct icu_chain *chain = 0;
NMEM nmem = 0;
-
+
*status = U_ZERO_ERROR;
- if (!xml_node ||xml_node->type != XML_ELEMENT_NODE)
- return 0;
-
+ if (xml_node && xml_node->type == XML_ELEMENT_NODE)
{
- xmlChar *xml_locale = xmlGetProp((xmlNode *) xml_node,
+ xmlChar *xml_locale = xmlGetProp((xmlNode *) xml_node,
(xmlChar *) "locale");
-
if (xml_locale)
{
chain = icu_chain_create((const char *) xml_locale, sort, status);
xmlFree(xml_locale);
}
-
}
+
if (!chain)
return 0;
yaz_log(YLOG_WARN, "Unsupported attribute '%s' for "
"element '%s'", attr->name, node->name);
no_errors++;
- continue;
}
}
if (!rule && node->children)
rule = nmem_text_node_cdata(node->children, nmem);
-
- if (!rule || !*rule)
+
+ if (!rule && strcmp((const char *) node->name, "display"))
{
- yaz_log(YLOG_WARN, "Missing rule for element '%s'", node->name);
+ yaz_log(YLOG_WARN, "Missing attribute 'rule' for element %s",
+ (const char *) node->name);
no_errors++;
continue;
}
-
if (!strcmp((const char *) node->name, "casemap"))
- step = icu_chain_insert_step(chain, ICU_chain_step_type_casemap,
- (const uint8_t *) rule, status);
+ step = icu_chain_insert_step(chain,
+ ICU_chain_step_type_casemap,
+ rule, status);
else if (!strcmp((const char *) node->name, "transform"))
- step = icu_chain_insert_step(chain, ICU_chain_step_type_transform,
- (const uint8_t *) rule, status);
+ step = icu_chain_insert_step(chain,
+ ICU_chain_step_type_transform,
+ rule, status);
else if (!strcmp((const char *) node->name, "transliterate"))
- step = icu_chain_insert_step(chain, ICU_chain_step_type_transliterate,
- (const uint8_t *) rule, status);
+ step = icu_chain_insert_step(chain,
+ ICU_chain_step_type_transliterate,
+ rule, status);
else if (!strcmp((const char *) node->name, "tokenize"))
- step = icu_chain_insert_step(chain, ICU_chain_step_type_tokenize,
- (const uint8_t *) rule, status);
+ step = icu_chain_insert_step(chain, ICU_chain_step_type_tokenize,
+ rule, status);
else if (!strcmp((const char *) node->name, "display"))
- step = icu_chain_insert_step(chain, ICU_chain_step_type_display,
- (const uint8_t *) "", status);
+ step = icu_chain_insert_step(chain, ICU_chain_step_type_display,
+ rule, status);
+ else if (!strcmp((const char *) node->name, "stemming"))
+ step = icu_chain_insert_step(chain, YAZ_chain_step_type_stemming,
+ rule, status);
+ else if (!strcmp((const char *) node->name, "join"))
+ step = icu_chain_insert_step(chain, ICU_chain_step_type_join,
+ rule, status);
else if (!strcmp((const char *) node->name, "normalize"))
{
yaz_log(YLOG_WARN, "Element %s is deprecated. "
"Use transform instead", node->name);
- step = icu_chain_insert_step(chain, ICU_chain_step_type_transform,
- (const uint8_t *) rule, status);
+ step = icu_chain_insert_step(chain, ICU_chain_step_type_transform,
+ rule, status);
}
else if (!strcmp((const char *) node->name, "index")
|| !strcmp((const char *) node->name, "sortkey"))
no_errors++;
continue;
}
+ if (!step)
+ {
+ yaz_log(YLOG_WARN, "Step not created for %s", node->name);
+ no_errors++;
+ }
if (step && U_FAILURE(*status))
{
+ yaz_log(YLOG_WARN, "ICU Error %d %s for element %s, rule %s",
+ *status, u_errorName(*status), node->name, rule ?
+ rule : "");
no_errors++;
break;
}
return chain;
}
-
-static struct icu_chain_step *icu_chain_insert_step(
- struct icu_chain *chain, enum icu_chain_step_type type,
- const uint8_t *rule, UErrorCode *status)
-{
- struct icu_chain_step *step = 0;
- if (!chain || !type || !rule)
- return 0;
-
- /* create actual chain step with this buffer */
- step = icu_chain_step_create(chain, type, rule,
- status);
-
- step->previous = chain->csteps;
- chain->csteps = step;
-
- return step;
-}
-
struct icu_iter {
struct icu_chain *chain;
struct icu_buf_utf16 *last;
+ struct icu_buf_utf16 *org;
UErrorCode status;
struct icu_buf_utf8 *display;
struct icu_buf_utf8 *sort8;
struct icu_buf_utf8 *result;
- struct icu_buf_utf16 *input;
int token_count;
+ size_t org_start;
+ size_t org_len;
struct icu_chain_step *steps;
};
struct icu_buf_utf8 *dst8 = icu_buf_utf8_create(0);
icu_utf16_to_utf8(dst8, src16, &status);
- assert(status != 1234);
if (U_FAILURE(status))
{
printf("failure");
else
{
struct icu_buf_utf16 *dst = icu_iter_invoke(iter, step->previous, src);
-
+
switch (step->type)
{
case ICU_chain_step_type_casemap:
}
dst = icu_buf_utf16_create(0);
iter->status = U_ZERO_ERROR;
- if (!icu_tokenizer_next_token(step->u.tokenizer, dst, &iter->status))
+ if (!icu_tokenizer_next_token(step->u.tokenizer, dst, &iter->status,
+ &iter->org_start, &iter->org_len))
{
icu_buf_utf16_destroy(dst);
dst = 0;
if (dst)
icu_utf16_to_utf8(iter->display, dst, &iter->status);
break;
+ case YAZ_chain_step_type_stemming:
+ if (dst)
+ {
+ struct icu_buf_utf16 *src = dst;
+ dst = icu_buf_utf16_create(0);
+ yaz_stemmer_stem(step->u.stemmer, dst, src, &iter->status);
+ icu_buf_utf16_destroy(src);
+ }
+ break;
+ case ICU_chain_step_type_join:
+ if (dst)
+ {
+ while (1)
+ {
+ struct icu_buf_utf16 *dst1 =
+ icu_iter_invoke(iter, step->previous, 0);
+
+ if (!dst1)
+ break;
+ dst = icu_buf_utf16_append(dst, step->u.join);
+ dst = icu_buf_utf16_append(dst, dst1);
+ icu_buf_utf16_destroy(dst1);
+ }
+ }
+ break;
default:
assert(0);
}
iter->display = icu_buf_utf8_create(0);
iter->sort8 = icu_buf_utf8_create(0);
iter->result = icu_buf_utf8_create(0);
+ iter->org = icu_buf_utf16_create(0);
iter->last = 0; /* no last returned string (yet) */
iter->steps = icu_chain_step_clone(chain->csteps);
- iter->input = 0;
+ iter->token_count = 0;
return iter;
}
void icu_iter_first(yaz_icu_iter_t iter, const char *src8cstr)
{
- if (iter->input)
- icu_buf_utf16_destroy(iter->input);
- iter->input = icu_buf_utf16_create(0);
+ struct icu_buf_utf16 *src = icu_buf_utf16_create(0);
+ icu_utf16_from_utf8_cstr(src, src8cstr, &iter->status);
+ icu_buf_utf16_copy(iter->org, src);
iter->token_count = 0;
- /* fill and assign input string.. It will be 0 after
- first iteration */
- icu_utf16_from_utf8_cstr(iter->input, src8cstr, &iter->status);
+ iter->org_start = 0;
+ iter->org_len = src->utf16_len;
+ iter->last = icu_iter_invoke(iter, iter->steps, src);
}
void icu_iter_destroy(yaz_icu_iter_t iter)
icu_buf_utf8_destroy(iter->display);
icu_buf_utf8_destroy(iter->sort8);
icu_buf_utf8_destroy(iter->result);
- if (iter->input)
- icu_buf_utf16_destroy(iter->input);
+ icu_buf_utf16_destroy(iter->org);
icu_chain_step_destroy(iter->steps);
xfree(iter);
}
int icu_iter_next(yaz_icu_iter_t iter)
{
- if (!iter->input && iter->last == 0)
+ if (iter->token_count && iter->last)
+ iter->last = icu_iter_invoke(iter, iter->steps, 0);
+ if (!iter->last)
return 0;
else
{
- /* on first call, iter->input is the input string. Thereafter: 0. */
- iter->last = icu_iter_invoke(iter, iter->steps ?
- iter->steps : iter->chain->csteps,
- iter->input);
- iter->input = 0;
-
- if (!iter->last)
- return 0;
-
iter->token_count++;
-
if (iter->chain->sort)
- {
+ {
icu_sortkey8_from_utf16(iter->chain->coll,
iter->sort8, iter->last,
&iter->status);
}
const char *icu_iter_get_display(yaz_icu_iter_t iter)
-{
- return icu_buf_utf8_to_cstr(iter->display);
+{
+ return icu_buf_utf8_to_cstr(iter->display);
}
int icu_iter_get_token_number(yaz_icu_iter_t iter)
-{
+{
return iter->token_count;
}
-int icu_chain_assign_cstr(struct icu_chain *chain, const char *src8cstr,
+
+void icu_iter_get_org_info(yaz_icu_iter_t iter, size_t *start, size_t *len)
+{
+ /* save full length of org since we're gonna cut it */
+ int32_t save_len = iter->org->utf16_len;
+
+ struct icu_buf_utf8 *tmp = icu_buf_utf8_create(0);
+ UErrorCode status = U_ZERO_ERROR;
+
+ iter->org->utf16_len = iter->org_start;
+ icu_utf16_to_utf8(tmp, iter->org, &status);
+ if (U_SUCCESS(status))
+ *start = tmp->utf8_len;
+ else
+ *start = 0;
+ iter->org->utf16_len = iter->org_start + iter->org_len;
+ icu_utf16_to_utf8(tmp, iter->org, &status);
+ if (U_SUCCESS(status))
+ *len = tmp->utf8_len - *start;
+ else
+ *len = 0;
+ icu_buf_utf8_destroy(tmp);
+ iter->org->utf16_len = save_len;
+}
+
+int icu_chain_assign_cstr(struct icu_chain *chain, const char *src8cstr,
UErrorCode *status)
{
if (chain->iter)
return 0;
}
+void icu_chain_get_org_info(struct icu_chain *chain, size_t *start, size_t *len)
+{
+ if (chain->iter)
+ icu_iter_get_org_info(chain->iter, start, len);
+}
+
+
#endif /* YAZ_HAVE_ICU */
/*