-/* $Id: tst_icu_I18N.c,v 1.2 2007-10-22 17:32:07 adam Exp $
+/* $Id: tst_icu_I18N.c,v 1.10 2007-10-29 10:22:23 marc Exp $
Copyright (c) 2006-2007, Index Data.
This file is part of Pazpar2.
struct icu_chain * chain = 0;
- const char * xml_str = "<icu_chain id=\"en:word\" locale=\"en\">"
+ const char * xml_str = "<icu>"
"<normalize rule=\"[:Control:] Any-Remove\"/>"
"<tokenize rule=\"l\"/>"
"<normalize rule=\"[[:WhiteSpace:][:Punctuation:]] Remove\"/>"
"<display/>"
"<casemap rule=\"l\"/>"
- "<index/>"
- "<sortkey/>"
- "</icu_chain>";
+ "</icu>";
xmlDoc *doc = xmlParseMemory(xml_str, strlen(xml_str));
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);
-
-#if 0
- chain = icu_chain_create((uint8_t *) "en:word", (uint8_t *) "en");
- step = icu_chain_insert_step(chain, ICU_chain_step_type_normalize,
- (const uint8_t *) "[:Control:] Any-Remove",
- &status);
- step = icu_chain_insert_step(chain, ICU_chain_step_type_tokenize,
- (const uint8_t *) "s",
- &status);
- step = icu_chain_insert_step(chain, ICU_chain_step_type_tokenize,
- (const uint8_t *) "l",
- &status);
- step = icu_chain_insert_step(chain, ICU_chain_step_type_normalize,
- (const uint8_t *)
- "[[:WhiteSpace:][:Punctuation:]] Any-Remove",
- &status);
- 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_normalize, */
-/* (const uint8_t *) "Lower", */
-/* &status); */
- step = icu_chain_insert_step(chain, ICU_chain_step_type_casemap,
- (const uint8_t *) "l",
- &status);
- step = icu_chain_insert_step(chain, ICU_chain_step_type_index,
- (const uint8_t *)"",
- &status);
-/* step = icu_chain_insert_step(chain, ICU_chain_step_type_sortkey, */
-/* (const uint8_t *)"", */
-/* &status); */
-
-#endif
+ chain = icu_chain_xml_config(xml_node, (uint8_t *) "en", 0, &status);
xmlFreeDoc(doc);
YAZ_CHECK(chain);
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_token_number(chain),
+ // icu_chain_token_norm(chain),
+ // icu_chain_token_display(chain));
}
- YAZ_CHECK_EQ(icu_chain_get_token_count(chain), 7);
+ YAZ_CHECK_EQ(icu_chain_token_number(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_token_number(chain),
+ // icu_chain_token_norm(chain),
+ // icu_chain_token_display(chain));
}
- YAZ_CHECK_EQ(icu_chain_get_token_count(chain), 3);
+ YAZ_CHECK_EQ(icu_chain_token_number(chain), 3);
icu_chain_destroy(chain);
}
UErrorCode status = U_ZERO_ERROR;
struct icu_chain * chain = 0;
- const char * xml_str = "<icu_chain id=\"en:word\" locale=\"en\">"
+ const char * xml_str = "<icu>"
/* if the first rule is normalize instead. Then it works */
#if 0
"<normalize rule=\"[[:WhiteSpace:][:Punctuation:]] Remove\"/>"
"<display/>"
"<casemap rule=\"l\"/>"
- "<index/>"
- "<sortkey/>"
- "</icu_chain>";
+ "</icu>";
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);
+ chain = icu_chain_xml_config(xml_node, (uint8_t *) "en", 0, &status);
xmlFreeDoc(doc);
YAZ_CHECK(chain);
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_token_number(chain),
+ // icu_chain_token_norm(chain),
+ // icu_chain_token_display(chain));
+
+ }
+
- YAZ_CHECK_EQ(icu_chain_get_token_count(chain), 7);
+ YAZ_CHECK_EQ(icu_chain_token_number(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_token_number(chain),
+ // icu_chain_token_norm(chain),
+ // icu_chain_token_display(chain));
}
/* we expect 'what' 'is' 'this', i.e. 3 tokens */
- YAZ_CHECK_EQ(icu_chain_get_token_count(chain), 3);
+ YAZ_CHECK_EQ(icu_chain_token_number(chain), 3);
+
+ 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>"
+ "<tokenize rule=\"w\"/>"
+ "<normalize rule=\"[[:WhiteSpace:][:Punctuation:]] Remove\"/>"
+ "</icu>";
+
+ xmlDoc *doc = xmlParseMemory(xml_str, strlen(xml_str));
+ xmlNode *xml_node = xmlDocGetRootElement(doc);
+ YAZ_CHECK(xml_node);
+
+ chain = icu_chain_xml_config(xml_node, (uint8_t *) "en", 0, &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_token_number(chain),
+ // icu_chain_token_norm(chain),
+ // icu_chain_token_display(chain));
+ }
+
+ YAZ_CHECK_EQ(icu_chain_token_number(chain), 8);
+
+ icu_chain_destroy(chain);
+}
+
+void test_chain_empty_chain(void)
+{
+ UErrorCode status = U_ZERO_ERROR;
+ struct icu_chain * chain = 0;
+
+ const char * xml_str = "<icu>"
+ "</icu>";
+
+ const char * src8 = "some 5487 weired !ยค%&(/& sTuFf";
+ char * dest8 = 0;
+
+ xmlDoc *doc = xmlParseMemory(xml_str, strlen(xml_str));
+ xmlNode *xml_node = xmlDocGetRootElement(doc);
+ YAZ_CHECK(xml_node);
+
+ chain = icu_chain_xml_config(xml_node, (uint8_t *) "en", 0, &status);
+
+ xmlFreeDoc(doc);
+ YAZ_CHECK(chain);
+
+ YAZ_CHECK(icu_chain_assign_cstr(
+ chain, src8,
+ &status));
+
+ while (icu_chain_next_token(chain, &status)){
+ ;
+ //printf("%d '%s' '%s'\n",
+ // icu_chain_token_number(chain),
+ // icu_chain_token_norm(chain),
+ // icu_chain_token_display(chain));
+ }
+
+ YAZ_CHECK_EQ(icu_chain_token_number(chain), 1);
+
+ dest8 = icu_chain_token_norm(chain);
+ YAZ_CHECK_EQ(strcmp(src8, dest8), 0);
+
icu_chain_destroy(chain);
}
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_chain_empty_chain();
test_bug_1140();
#else /* HAVE_ICU */