X-Git-Url: http://git.indexdata.com/?p=yaz-moved-to-github.git;a=blobdiff_plain;f=src%2Ficu_chain.c;h=b672975293c24ce880e6d18a0f7cbc15833e6a1e;hp=e7b9b4e9811f005d10acb732d43adbc32d814e08;hb=0b121738c22cd67b153792d645746447d8a81767;hpb=8e4840a9a7b6117d92d9d97fdb4048a7f8b9ce09

diff --git a/src/icu_chain.c b/src/icu_chain.c
index e7b9b4e..b672975 100644
--- a/src/icu_chain.c
+++ b/src/icu_chain.c
@@ -1,5 +1,5 @@
 /* This file is part of the YAZ toolkit.
- * Copyright (C) 1995-2013 Index Data
+ * Copyright (C) Index Data
  * See the file LICENSE for details.
  */
 
@@ -83,35 +83,40 @@ static struct icu_chain_step *icu_chain_insert_step(
 {
     struct icu_chain_step *step = 0;
 
-    if (!chain || !type || !rule)
-        return 0;
+    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(rule, 'f', 0, status);
         break;
     case ICU_chain_step_type_tokenize:
+        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', 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;
@@ -279,12 +284,18 @@ struct icu_chain *icu_chain_xml_config(const xmlNode *xml_node,
                 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 && strcmp((const char *) node->name, "display"))
+        {
+            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,
@@ -302,15 +313,13 @@ struct icu_chain *icu_chain_xml_config(const xmlNode *xml_node,
                                          rule, status);
         else if (!strcmp((const char *) node->name, "display"))
             step = icu_chain_insert_step(chain, ICU_chain_step_type_display,
-                                         "", status);
+                                         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. "
@@ -337,6 +346,9 @@ struct icu_chain *icu_chain_xml_config(const xmlNode *xml_node,
         }
         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;
         }
@@ -353,12 +365,16 @@ struct icu_chain *icu_chain_xml_config(const xmlNode *xml_node,
 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;
+    size_t utf8_base;
+    size_t utf16_base;
     struct icu_chain_step *steps;
 };
 
@@ -414,7 +430,8 @@ struct icu_buf_utf16 *icu_iter_invoke(yaz_icu_iter_t iter,
             }
             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;
@@ -474,22 +491,24 @@ yaz_icu_iter_t icu_iter_create(struct icu_chain *chain)
     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->utf8_base = iter->utf16_base = 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)
@@ -499,8 +518,7 @@ 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);
     }
@@ -508,20 +526,13 @@ void icu_iter_destroy(yaz_icu_iter_t 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. */
-        assert(iter->steps || !iter->chain->csteps);
-        iter->last = icu_iter_invoke(iter, iter->steps, iter->input);
-        iter->input = 0;
-
-        if (!iter->last)
-            return 0;
-
         iter->token_count++;
-
         if (iter->chain->sort)
         {
             icu_sortkey8_from_utf16(iter->chain->coll,
@@ -555,6 +566,37 @@ int icu_iter_get_token_number(yaz_icu_iter_t iter)
     return iter->token_count;
 }
 
+
+void icu_iter_get_org_info(yaz_icu_iter_t iter, size_t *start, size_t *len)
+{
+    int32_t len1 = 0, len2 = 0;
+    UErrorCode status = U_ZERO_ERROR;
+
+    if (iter->org_start < iter->utf16_base)
+    {
+        iter->utf8_base = 0;
+        iter->utf16_base = 0;
+    }
+    u_strToUTF8(0, 0, &len1,
+                iter->org->utf16 + iter->utf16_base,
+                iter->org_start - iter->utf16_base,
+                &status);
+
+    status = U_ZERO_ERROR;
+
+    *start = len1 + iter->utf8_base;
+
+    u_strToUTF8(0, 0, &len2,
+                iter->org->utf16 + iter->utf16_base,
+                iter->org_start - iter->utf16_base + iter->org_len,
+                &status);
+
+    *len = len2 - len1;
+
+    iter->utf8_base = *start;
+    iter->utf16_base = iter->org_start;
+}
+
 int icu_chain_assign_cstr(struct icu_chain *chain, const char *src8cstr,
                           UErrorCode *status)
 {
@@ -599,6 +641,13 @@ const char *icu_chain_token_sortkey(struct icu_chain *chain)
     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 */
 
 /*