/* This file is part of the YAZ toolkit.
- * Copyright (C) 1995-2010 Index Data
+ * Copyright (C) 1995-2011 Index Data
* See the file LICENSE for details.
*/
#include <yaz/icu_I18N.h>
-#include <yaz/log.h>
+#include <yaz/stemmer.h>
+#include <yaz/log.h>
+#include <yaz/nmem.h>
+#include <yaz/nmem_xml.h>
#include <string.h>
#include <stdlib.h>
#include <stdio.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) */
};
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;
+ yaz_stemmer_p stemmer;
} u;
struct icu_chain_step *previous;
};
return 1;
}
-static struct icu_chain_step *icu_chain_step_create(
- struct icu_chain *chain, enum icu_chain_step_type type,
+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;
step->u.transform = icu_transform_create("custom", 'f',
(const char *) rule, status);
break;
+ case YAZ_chain_step_type_stemming:
+ step->u.stemmer = yaz_stemmer_create((char *) chain->locale, (const char *) 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;
default:
break;
}
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;
}
UErrorCode *status)
{
xmlNode *node = 0;
+ int no_errors = 0;
struct icu_chain *chain = 0;
+ NMEM nmem = 0;
*status = U_ZERO_ERROR;
if (!chain)
return 0;
+ nmem = nmem_create();
for (node = xml_node->children; node; node = node->next)
{
- xmlChar *xml_rule;
+ char *rule = 0;
struct icu_chain_step *step = 0;
+ struct _xmlAttr *attr;
+ nmem_reset(nmem);
if (node->type != XML_ELEMENT_NODE)
continue;
- xml_rule = xmlGetProp(node, (xmlChar *) "rule");
-
- yaz_log(YLOG_LOG, "rule=%s", xml_rule);
-
+ for (attr = node->properties; attr; attr = attr->next)
+ {
+ if (!strcmp((const char *) attr->name, "rule"))
+ {
+ rule = nmem_text_node_cdata(attr->children, nmem);
+ }
+ else
+ {
+ 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 (!strcmp((const char *) node->name, "casemap"))
step = icu_chain_insert_step(chain, ICU_chain_step_type_casemap,
- (const uint8_t *) xml_rule, status);
+ (const uint8_t *) rule, status);
else if (!strcmp((const char *) node->name, "transform"))
step = icu_chain_insert_step(chain, ICU_chain_step_type_transform,
- (const uint8_t *) xml_rule, status);
+ (const uint8_t *) rule, status);
else if (!strcmp((const char *) node->name, "transliterate"))
step = icu_chain_insert_step(chain, ICU_chain_step_type_transliterate,
- (const uint8_t *) xml_rule, status);
+ (const uint8_t *) rule, status);
else if (!strcmp((const char *) node->name, "tokenize"))
step = icu_chain_insert_step(chain, ICU_chain_step_type_tokenize,
- (const uint8_t *) xml_rule, status);
+ (const uint8_t *) 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);
+ else if (!strcmp((const char *) node->name, "stemming"))
+ step = icu_chain_insert_step(chain, YAZ_chain_step_type_stemming,
+ (const uint8_t *) 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 *) xml_rule, status);
+ (const uint8_t *) rule, status);
}
else if (!strcmp((const char *) node->name, "index")
|| !strcmp((const char *) node->name, "sortkey"))
else
{
yaz_log(YLOG_WARN, "Unknown element %s", node->name);
- icu_chain_destroy(chain);
- return 0;
+ no_errors++;
+ continue;
}
- xmlFree(xml_rule);
if (step && U_FAILURE(*status))
{
- icu_chain_destroy(chain);
- return 0;
+ 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)
+ nmem_destroy(nmem);
+ if (no_errors)
+ {
+ icu_chain_destroy(chain);
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;
+ }
+ return chain;
}
struct icu_iter {
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;
default:
assert(0);
}