Implemented sorting using ICU. Bug #1190. The Debian package now

[pazpar2-moved-to-github.git] / src / charsets.c

diff --git a/src/charsets.c b/src/charsets.c
index 2d5dad3..7d41283 100644 (file)
--- a/src/charsets.c
+++ b/src/charsets.c
@@ -1,4 +1,4 @@
-/* $Id: charsets.c,v 1.5 2007-05-25 10:32:55 marc Exp $
+/* $Id: charsets.c,v 1.6 2007-09-10 16:25:50 adam Exp $
    Copyright (c) 2006-2007, Index Data.
 
 This file is part of Pazpar2.
@@ -34,8 +34,7 @@ Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA
 #include <assert.h>
 
 #include "charsets.h"
-//#include "config.h"
-//#include "parameters.h"
+#include "normalize7bit.h"
 
 #ifdef HAVE_ICU
 #include "icu_I18N.h"
@@ -44,7 +43,7 @@ Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA
 /* charset handle */
 struct pp2_charset_s {
     const char *(*token_next_handler)(pp2_relevance_token_t prt);
-    /* other handlers will come as we see fit */
+    const char *(*get_sort_handler)(pp2_relevance_token_t prt, int skip);
 #ifdef HAVE_ICU
     struct icu_chain * icu_chn;
     UErrorCode icu_sts;
@@ -52,31 +51,72 @@ struct pp2_charset_s {
 };
 
 static const char *pp2_relevance_token_a_to_z(pp2_relevance_token_t prt);
+static const char *pp2_get_sort_ascii(pp2_relevance_token_t prt, int skip_article);
 
 #ifdef HAVE_ICU
 static const char *pp2_relevance_token_icu(pp2_relevance_token_t prt);
+static const char *pp2_get_sort_icu(pp2_relevance_token_t prt, int skip_article);
 #endif // HAVE_ICU
 
 /* tokenzier handle */
 struct pp2_relevance_token_s {
     const char *cp;     /* unnormalized buffer we're tokenizing */
+    const char *last_cp;  /* pointer to last token we're dealing with */
     pp2_charset_t pct;  /* our main charset handle (type+config) */
     WRBUF norm_str;     /* normized string we return (temporarily) */
+    WRBUF sort_str;     /* sort string we return (temporarily) */
 };
 
+
+pp2_charset_t pp2_charset_create_xml(xmlNode *xml_node)
+{
+#ifdef HAVE_ICU
+    UErrorCode status = U_ZERO_ERROR;
+    while (xml_node && xml_node->type != XML_ELEMENT_NODE)
+        xml_node = xml_node->next;
+    struct icu_chain *chain = icu_chain_xml_config(xml_node, &status);
+    if (!chain || U_FAILURE(status)){
+        //xmlDocPtr icu_doc = 0;
+        //xmlChar *xmlstr = 0;
+                //int size = 0;
+                //xmlDocDumpMemory(icu_doc, size);
+        
+        yaz_log(YLOG_FATAL, "Could not parse ICU chain config:\n"
+                "<%s>\n ... \n</%s>",
+                xml_node->name, xml_node->name);
+        return 0;
+    }
+    return pp2_charset_create(chain);
+#else // HAVE_ICU
+    yaz_log(YLOG_FATAL, "Error: ICU support requested with element:\n"
+            "<%s>\n ... \n</%s>",
+            n->name, n->name);
+    yaz_log(YLOG_FATAL, 
+            "But no ICU support compiled into pazpar2 server.");
+    yaz_log(YLOG_FATAL, 
+            "Please install libicu36-dev and icu-doc or similar, "
+            "re-configure and re-compile");            
+    return 0;
+#endif // HAVE_ICU
+}
+
+
 pp2_charset_t pp2_charset_create(struct icu_chain * icu_chn)
 {
     pp2_charset_t pct = xmalloc(sizeof(*pct));
 
     pct->token_next_handler = pp2_relevance_token_a_to_z;
+    pct->get_sort_handler  = pp2_get_sort_ascii;
 #ifdef HAVE_ICU
     pct->icu_chn = 0;
-    if (icu_chn){
+    if (icu_chn)
+    {
         pct->icu_chn = icu_chn;
         pct->icu_sts = U_ZERO_ERROR;
         pct->token_next_handler = pp2_relevance_token_icu;
+        pct->get_sort_handler = pp2_get_sort_icu;
     }
- #endif // HAVE_ICU
+#endif // HAVE_ICU
     return pct;
 }
 
@@ -93,7 +133,9 @@ pp2_relevance_token_t pp2_relevance_tokenize(pp2_charset_t pct,
     assert(pct);
 
     prt->norm_str = wrbuf_alloc();
+    prt->sort_str = wrbuf_alloc();
     prt->cp = buf;
+    prt->last_cp = 0;
     prt->pct = pct;
 
 #ifdef HAVE_ICU
@@ -104,7 +146,6 @@ pp2_relevance_token_t pp2_relevance_tokenize(pp2_charset_t pct,
         ok = icu_chain_assign_cstr(pct->icu_chn, buf, &pct->icu_sts);
         //printf("\nfield ok: %d '%s'\n", ok, buf);
         prt->pct = pct;
-        prt->norm_str = 0;
     }
 #endif // HAVE_ICU
     return prt;
@@ -116,6 +157,8 @@ void pp2_relevance_token_destroy(pp2_relevance_token_t prt)
     assert(prt);
     if(prt->norm_str) 
         wrbuf_destroy(prt->norm_str);
+    if(prt->sort_str) 
+        wrbuf_destroy(prt->sort_str);
     xfree(prt);
 }
 
@@ -125,7 +168,12 @@ const char *pp2_relevance_token_next(pp2_relevance_token_t prt)
     return (prt->pct->token_next_handler)(prt);
 }
 
-#define raw_char(c) (((c) >= 'a' && (c) <= 'z') ? (c) - 'a' + 1 : -1)
+const char *pp2_get_sort(pp2_relevance_token_t prt, int skip)
+{
+    return prt->pct->get_sort_handler(prt, skip);
+}
+
+#define raw_char(c) (((c) >= 'a' && (c) <= 'z') ? (c) : -1)
 /* original tokenizer with our tokenize interface, but we
    add +1 to ensure no '\0' are in our string (except for EOF)
 */
@@ -140,9 +188,12 @@ static const char *pp2_relevance_token_a_to_z(pp2_relevance_token_t prt)
     if (*cp == '\0')
     {
         prt->cp = cp;
+        prt->last_cp = 0;
         return 0;
     }
     /* now read the term itself */
+
+    prt->last_cp = cp;
     wrbuf_rewind(prt->norm_str);
     while (*cp && (c = raw_char(tolower(*cp))) >= 0)
     {
@@ -153,24 +204,45 @@ static const char *pp2_relevance_token_a_to_z(pp2_relevance_token_t prt)
     return wrbuf_cstr(prt->norm_str);
 }
 
+static const char *pp2_get_sort_ascii(pp2_relevance_token_t prt,
+                                    int skip_article)
+{
+    if (prt->last_cp == 0)
+        return 0;
+    else
+    {
+        char *tmp = xstrdup(prt->last_cp);
+        char *result = 0;
+        result = normalize7bit_mergekey(tmp, skip_article);
+        
+        wrbuf_rewind(prt->sort_str);
+        wrbuf_puts(prt->sort_str, result);
+        xfree(tmp);
+        return wrbuf_cstr(prt->sort_str);
+    }
+}
+
 
 #ifdef HAVE_ICU
 static const char *pp2_relevance_token_icu(pp2_relevance_token_t prt)
 {
-    //&& U_SUCCESS(pct->icu_sts))
-    if (icu_chain_next_token(prt->pct->icu_chn, &prt->pct->icu_sts)){
-        //printf("'%s' ",  icu_chain_get_norm(prt->pct->icu_chn)); 
+    if (icu_chain_next_token(prt->pct->icu_chn, &prt->pct->icu_sts))
+    {
         if (U_FAILURE(prt->pct->icu_sts))
         {
-            //printf("ICU status failure\n "); 
             return 0;
         }
-            
         return icu_chain_get_norm(prt->pct->icu_chn);
     }
-    //printf ("EOF\n");
     return 0;
-};
+}
+
+static const char *pp2_get_sort_icu(pp2_relevance_token_t prt,
+                                    int skip_article)
+{
+    return icu_chain_get_sort(prt->pct->icu_chn);
+}
+
 #endif // HAVE_ICU