fixed wrong token count when tokens disappear with ICU normalization
[yaz-moved-to-github.git] / test / tst_icu_I18N.c
index 655c459..717e760 100644 (file)
@@ -1,4 +1,4 @@
-/* $Id: tst_icu_I18N.c,v 1.2 2007-10-22 17:32:07 adam Exp $
+/* $Id: tst_icu_I18N.c,v 1.5 2007-10-24 13:23:34 marc Exp $
    Copyright (c) 2006-2007, Index Data.
 
    This file is part of Pazpar2.
@@ -508,7 +508,7 @@ void test_icu_I18N_chain(int argc, char **argv)
     xmlNode *xml_node = xmlDocGetRootElement(doc);
     YAZ_CHECK(xml_node);
 
-    printf("ICU chain:\ninput: '%s'\n", en_str);
+    // printf("ICU chain:\ninput: '%s'\n", en_str);
 
 
     chain = icu_chain_xml_config(xml_node, &status);
@@ -552,10 +552,11 @@ void test_icu_I18N_chain(int argc, char **argv)
     YAZ_CHECK(icu_chain_assign_cstr(chain, en_str, &status));
 
     while (icu_chain_next_token(chain, &status)){
-        printf("%d '%s' '%s'\n",
-               icu_chain_get_token_count(chain),
-               icu_chain_get_norm(chain),
-               icu_chain_get_display(chain));
+        ;
+        // printf("%d '%s' '%s'\n",
+        //       icu_chain_get_token_count(chain),
+        //       icu_chain_get_norm(chain),
+        //       icu_chain_get_display(chain));
     }
 
     YAZ_CHECK_EQ(icu_chain_get_token_count(chain), 7);
@@ -564,10 +565,11 @@ void test_icu_I18N_chain(int argc, char **argv)
     YAZ_CHECK(icu_chain_assign_cstr(chain, "what is this?", &status));
 
     while (icu_chain_next_token(chain, &status)){
-        printf("%d '%s' '%s'\n",
-               icu_chain_get_token_count(chain),
-               icu_chain_get_norm(chain),
-               icu_chain_get_display(chain));
+        ;
+        //printf("%d '%s' '%s'\n",
+        //       icu_chain_get_token_count(chain),
+        //       icu_chain_get_norm(chain),
+        //       icu_chain_get_display(chain));
     }
 
 
@@ -610,18 +612,26 @@ void test_bug_1140(void)
                   chain,  "O Romeo, Romeo! wherefore art thou\t Romeo?",
                   &status));
 
-    while (icu_chain_next_token(chain, &status))
+    while (icu_chain_next_token(chain, &status)){    
         ;
+        //printf("%d '%s' '%s'\n",
+        //       icu_chain_get_token_count(chain),
+        //       icu_chain_get_norm(chain),
+        //       icu_chain_get_display(chain));
+
+    }
+    
 
     YAZ_CHECK_EQ(icu_chain_get_token_count(chain), 7);
 
     YAZ_CHECK(icu_chain_assign_cstr(chain, "what is this?", &status));
 
     while (icu_chain_next_token(chain, &status)){
-        printf("%d '%s' '%s'\n",
-               icu_chain_get_token_count(chain),
-               icu_chain_get_norm(chain),
-               icu_chain_get_display(chain));
+       ;
+       //printf("%d '%s' '%s'\n",
+       //        icu_chain_get_token_count(chain),
+       //        icu_chain_get_norm(chain),
+       //        icu_chain_get_display(chain));
     }
 
     /* we expect 'what' 'is' 'this', i.e. 3 tokens */
@@ -630,6 +640,45 @@ void test_bug_1140(void)
     icu_chain_destroy(chain);
 }
 
+
+
+void test_chain_empty_token(void)
+{
+    UErrorCode status = U_ZERO_ERROR;
+    struct icu_chain * chain = 0;
+
+    const char * xml_str = "<icu_chain id=\"en:word\" locale=\"en\">"
+        "<tokenize rule=\"w\"/>"
+        "<normalize rule=\"[[:WhiteSpace:][:Punctuation:]] Remove\"/>"
+        "<index/>"
+        "</icu_chain>";
+    
+    xmlDoc *doc = xmlParseMemory(xml_str, strlen(xml_str));
+    xmlNode *xml_node = xmlDocGetRootElement(doc);
+    YAZ_CHECK(xml_node);
+
+    chain = icu_chain_xml_config(xml_node, &status);
+
+    xmlFreeDoc(doc);
+    YAZ_CHECK(chain);
+    
+    YAZ_CHECK(icu_chain_assign_cstr(
+                  chain,  "a string with 15 wordbreaks and 8 tokens",
+                  &status));
+
+    while (icu_chain_next_token(chain, &status)){
+        ;
+        //printf("%d '%s' '%s'\n",
+        //       icu_chain_get_token_count(chain),
+        //       icu_chain_get_norm(chain),
+        //       icu_chain_get_display(chain));
+    }
+
+    YAZ_CHECK_EQ(icu_chain_get_token_count(chain), 8);
+
+    icu_chain_destroy(chain);
+}
+
 #endif // HAVE_ICU
 
 /* DO NOT EDIT THIS FILE IF YOUR EDITOR DOES NOT SUPPORT UTF-8 */
@@ -644,9 +693,10 @@ int main(int argc, char **argv)
 
     test_icu_I18N_casemap(argc, argv);
     test_icu_I18N_sortmap(argc, argv);
-    test_icu_I18N_normalizer(argc, argv);
+    test_icu_I18N_normalizer(argc, argv); 
     test_icu_I18N_tokenizer(argc, argv);
     test_icu_I18N_chain(argc, argv);
+    test_chain_empty_token();
     test_bug_1140();
 
 #else /* HAVE_ICU */