#include <yaz/icu_I18N.h>
+#include <yaz/stemmer.h>
+
#include <yaz/log.h>
#include <string.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_transform * transform;
- struct icu_tokenizer * tokenizer;
+ struct icu_casemap * casemap;
+ struct icu_transform * transform;
+ struct icu_tokenizer * tokenizer;
+ yaz_stemmer_p stemmer;
} u;
struct icu_chain_step * previous;
};
struct icu_chain
{
- struct icu_iter *iter;
+ yaz_icu_iter_t iter;
char *locale;
int sort;
UCollator * coll;
- /* utf8 output buffers */
- struct icu_buf_utf8 * norm8;
-
/* linked list of chain steps */
struct icu_chain_step * csteps;
};
if (!chain || !type || !rule)
return 0;
- step = (struct icu_chain_step *) xmalloc(sizeof(struct icu_chain_step));
+ step = (struct icu_chain_step *) xmalloc(sizeof(*step));
step->type = type;
/* create auxilary objects */
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;
}
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:
+ yaz_stemmer_clone(step->u.stemmer);
+ break;
case ICU_chain_step_type_none:
break;
}
UErrorCode * status)
{
struct icu_chain * chain
- = (struct icu_chain *) xmalloc(sizeof(struct icu_chain));
+ = (struct icu_chain *) xmalloc(sizeof(*chain));
*status = U_ZERO_ERROR;
if (U_FAILURE(*status))
return 0;
- chain->norm8 = icu_buf_utf8_create(0);
chain->csteps = 0;
return chain;
if (chain->coll)
ucol_close(chain->coll);
- icu_buf_utf8_destroy(chain->norm8);
if (chain->iter)
icu_iter_destroy(chain->iter);
icu_chain_step_destroy(chain->csteps);
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 = yaz_chain_insert_step(chain, YAZ_chain_step_type_stemming,
+ (const uint8_t *) xml_rule, status);
else if (!strcmp((const char *) node->name, "normalize"))
{
yaz_log(YLOG_WARN, "Element %s is deprecated. "
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;
struct icu_chain_step *steps;
icu_buf_utf8_destroy(dst8);
}
-struct icu_buf_utf16 *icu_iter_invoke(struct icu_iter *iter,
+struct icu_buf_utf16 *icu_iter_invoke(yaz_icu_iter_t iter,
struct icu_chain_step *step,
struct icu_buf_utf16 *src)
{
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);
}
}
}
-struct icu_iter *icu_iter_create(struct icu_chain *chain,
- const char *src8cstr)
+yaz_icu_iter_t icu_iter_create(struct icu_chain *chain)
{
- if (!src8cstr)
- return 0;
- else
- {
- struct icu_iter *iter = xmalloc(sizeof(*iter));
- iter->chain = chain;
- iter->status = U_ZERO_ERROR;
- iter->display = icu_buf_utf8_create(0);
- iter->sort8 = icu_buf_utf8_create(0);
- iter->token_count = 0;
- iter->last = 0; /* no last returned string (yet) */
-#if 0
- iter->steps = icu_chain_step_clone(chain->csteps);
-#else
- iter->steps = 0;
-#endif
+ yaz_icu_iter_t iter = xmalloc(sizeof(*iter));
+ iter->chain = chain;
+ iter->status = U_ZERO_ERROR;
+ iter->display = icu_buf_utf8_create(0);
+ iter->sort8 = icu_buf_utf8_create(0);
+ iter->result = icu_buf_utf8_create(0);
+ iter->last = 0; /* no last returned string (yet) */
+ iter->steps = icu_chain_step_clone(chain->csteps);
+ iter->input = 0;
+
+ return iter;
+}
- /* fill and assign input string.. It will be 0 after
- first iteration */
- iter->input = icu_buf_utf16_create(0);
- icu_utf16_from_utf8_cstr(iter->input, src8cstr, &iter->status);
- 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);
+ 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);
}
-void icu_iter_destroy(struct icu_iter *iter)
+void icu_iter_destroy(yaz_icu_iter_t iter)
{
if (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_chain_step_destroy(iter->steps);
}
}
-int icu_iter_next(struct icu_iter *iter, struct icu_buf_utf8 *result)
+int icu_iter_next(yaz_icu_iter_t iter)
{
if (!iter->input && iter->last == 0)
return 0;
iter->sort8, iter->last,
&iter->status);
}
- icu_utf16_to_utf8(result, iter->last, &iter->status);
+ icu_utf16_to_utf8(iter->result, iter->last, &iter->status);
icu_buf_utf16_destroy(iter->last);
return 1;
}
}
-const char *icu_iter_get_sortkey(struct icu_iter *iter)
+const char *icu_iter_get_norm(yaz_icu_iter_t iter)
+{
+ return icu_buf_utf8_to_cstr(iter->result);
+}
+
+const char *icu_iter_get_sortkey(yaz_icu_iter_t iter)
{
return icu_buf_utf8_to_cstr(iter->sort8);
}
-const char *icu_iter_get_display(struct icu_iter *iter)
+const char *icu_iter_get_display(yaz_icu_iter_t iter)
{
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,
UErrorCode *status)
{
if (chain->iter)
icu_iter_destroy(chain->iter);
- chain->iter = icu_iter_create(chain, src8cstr);
+ chain->iter = icu_iter_create(chain);
+ icu_iter_first(chain->iter, src8cstr);
return 1;
}
int icu_chain_next_token(struct icu_chain * chain, UErrorCode *status)
{
*status = U_ZERO_ERROR;
- return icu_iter_next(chain->iter, chain->norm8);
+ return icu_iter_next(chain->iter);
}
int icu_chain_token_number(struct icu_chain * chain)
const char * icu_chain_token_norm(struct icu_chain * chain)
{
- if (chain->norm8)
- return icu_buf_utf8_to_cstr(chain->norm8);
+ if (chain->iter)
+ return icu_iter_get_norm(chain->iter);
return 0;
}
projects / yaz-moved-to-github.git / blobdiff