Merge branch 'master' of ssh://git.indexdata.com/home/git/pub/pazpar2
[pazpar2-moved-to-github.git] / src / session.c
index 36be061..64920da 100644 (file)
@@ -408,40 +408,43 @@ const char *session_setting_oneval(struct session_database *db, int offset)
 // setting. However, this is not a realistic use scenario.
 static int prepare_map(struct session *se, struct session_database *sdb)
 {
-    const char *s;
-
-    if (sdb->settings && sdb->settings[PZ_XSLT] && !sdb->map &&
-        (s = session_setting_oneval(sdb, PZ_XSLT)))        
+    if (sdb->settings && !sdb->map)
     {
-        char auto_stylesheet[256];
+        const char *s;
 
-        if (!strcmp(s, "auto"))
+        if (sdb->settings[PZ_XSLT] &&
+            (s = session_setting_oneval(sdb, PZ_XSLT)))        
         {
-            const char *request_syntax = session_setting_oneval(
-                sdb, PZ_REQUESTSYNTAX);
-            if (request_syntax)
+            char auto_stylesheet[256];
+            
+            if (!strcmp(s, "auto"))
             {
-                char *cp;
-                yaz_snprintf(auto_stylesheet, sizeof(auto_stylesheet),
-                             "%s.xsl", request_syntax);
-                for (cp = auto_stylesheet; *cp; cp++)
+                const char *request_syntax = session_setting_oneval(
+                    sdb, PZ_REQUESTSYNTAX);
+                if (request_syntax)
                 {
-                    /* deliberately only consider ASCII */
-                    if (*cp > 32 && *cp < 127)
-                        *cp = tolower(*cp);
+                    char *cp;
+                    yaz_snprintf(auto_stylesheet, sizeof(auto_stylesheet),
+                                 "%s.xsl", request_syntax);
+                    for (cp = auto_stylesheet; *cp; cp++)
+                    {
+                        /* deliberately only consider ASCII */
+                        if (*cp > 32 && *cp < 127)
+                            *cp = tolower(*cp);
+                    }
+                    s = auto_stylesheet;
+                }
+                else
+                {
+                    session_log(se, YLOG_WARN,
+                                "No pz:requestsyntax for auto stylesheet");
                 }
-                s = auto_stylesheet;
-            }
-            else
-            {
-                session_log(se, YLOG_WARN,
-                            "No pz:requestsyntax for auto stylesheet");
             }
+            sdb->map = normalize_cache_get(se->normalize_cache,
+                                           se->service, s);
+            if (!sdb->map)
+                return -1;
         }
-        sdb->map = normalize_cache_get(se->normalize_cache,
-                                       se->service, s);
-        if (!sdb->map)
-            return -1;
     }
     return 0;
 }
@@ -634,12 +637,12 @@ void session_sort(struct session *se, const char *field, int increasing)
     }
     if (sr)
     {
-        yaz_log(YLOG_LOG, "search_sort: field=%s increasing=%d already fetched",
+        session_log(se, YLOG_DEBUG, "search_sort: field=%s increasing=%d already fetched",
                 field, increasing);
         session_leave(se);
         return;
     }
-    yaz_log(YLOG_LOG, "search_sort: field=%s increasing=%d must fetch",
+    session_log(se, YLOG_DEBUG, "search_sort: field=%s increasing=%d must fetch",
             field, increasing);
     sr = nmem_malloc(se->nmem, sizeof(*sr));
     sr->field = nmem_strdup(se->nmem, field);
@@ -689,6 +692,8 @@ enum pazpar2_error_code session_search(struct session *se,
     se->reclist = 0;
     se->settings_modified = 0;
     relevance_destroy(&se->relevance);
+    if (nmem_total(se->nmem))
+        session_log(se, YLOG_DEBUG, "NMEN operation usage %zd", nmem_total(se->nmem));
     nmem_reset(se->nmem);
     se->total_records = se->total_merged = 0;
     se->num_termlists = 0;
@@ -746,12 +751,12 @@ enum pazpar2_error_code session_search(struct session *se,
                                        &tval);
             if (parse_ret == 1 && r == 2)
             {
-                session_log(se, YLOG_LOG, "client REUSE %s", client_get_id(cl));
+                session_log(se, YLOG_LOG, "client %s REUSE result", client_get_id(cl));
                 client_reingest(cl);
             }
             else if (r)
             {
-                session_log(se, YLOG_LOG, "client NEW %s", client_get_id(cl));
+                session_log(se, YLOG_LOG, "client %s NEW search", client_get_id(cl));
                 client_start_search(cl);
             }
             no_working++;
@@ -775,7 +780,7 @@ enum pazpar2_error_code session_search(struct session *se,
         else
             return PAZPAR2_NO_TARGETS;
     }
-    yaz_log(YLOG_LOG, "session_start_search done");
+    session_log(se, YLOG_LOG, "session_start_search done");
     return PAZPAR2_NO_ERROR;
 }
 
@@ -821,8 +826,7 @@ void session_init_databases(struct session *se)
 static struct session_database *load_session_database(struct session *se, 
                                                       char *id)
 {
-    struct database *db = new_database(id, se->session_nmem);
-
+    struct database *db = new_database_inherit_settings(id, se->session_nmem, se->service->settings);
     session_init_databases_fun((void*) se, db);
 
     // New sdb is head of se->databases list
@@ -886,6 +890,10 @@ void session_destroy(struct session *se)
     normalize_cache_destroy(se->normalize_cache);
     relevance_destroy(&se->relevance);
     reclist_destroy(se->reclist);
+    if (nmem_total(se->nmem))
+        session_log(se, YLOG_DEBUG, "NMEN operation usage %zd", nmem_total(se->nmem));
+    if (nmem_total(se->session_nmem))
+        session_log(se, YLOG_DEBUG, "NMEN session usage %zd", nmem_total(se->session_nmem));
     nmem_destroy(se->nmem);
     service_destroy(se->service);
     yaz_mutex_destroy(&se->session_mutex);
@@ -1015,6 +1023,14 @@ static int cmp_ht(const void *p1, const void *p2)
     return h2->hits - h1->hits;
 }
 
+// Compares two hitsbytarget nodes by hitcount
+static int cmp_ht_approx(const void *p1, const void *p2)
+{
+    const struct hitsbytarget *h1 = p1;
+    const struct hitsbytarget *h2 = p2;
+    return h2->approximation - h1->approximation;
+}
+
 static int targets_termlist_nb(WRBUF wrbuf, struct session *se, int num,
                                NMEM nmem, int version)
 {
@@ -1022,7 +1038,10 @@ static int targets_termlist_nb(WRBUF wrbuf, struct session *se, int num,
     int count, i;
 
     ht = hitsbytarget_nb(se, &count, nmem);
-    qsort(ht, count, sizeof(struct hitsbytarget), cmp_ht);
+    if (version >= 2)
+        qsort(ht, count, sizeof(struct hitsbytarget), cmp_ht_approx);
+    else
+        qsort(ht, count, sizeof(struct hitsbytarget), cmp_ht);
     for (i = 0; i < count && i < num && ht[i].hits > 0; i++)
     {
 
@@ -1576,6 +1595,7 @@ int ingest_record(struct client *cl, const char *rec,
     return ret;
 }
 
+// Skip record on non-zero
 static int check_limit_local(struct client *cl,
                              struct record *record,
                              int record_no)
@@ -1594,9 +1614,7 @@ static int check_limit_local(struct client *cl,
         char **values = 0;
         int i, num_v = 0;
         
-        const char *name =
-            client_get_facet_limit_local(cl, sdb, &l, nmem_tmp, &num_v,
-                                         &values);
+        const char *name = client_get_facet_limit_local(cl, sdb, &l, nmem_tmp, &num_v, &values);
         if (!name)
             break;
         
@@ -1608,7 +1626,7 @@ static int check_limit_local(struct client *cl,
         }
         ser_md = &service->metadata[md_field_id];
         rec_md = record->metadata[md_field_id];
-        yaz_log(YLOG_LOG, "check limit local %s", name);
+        yaz_log(YLOG_DEBUG, "check limit local %s", name);
         for (i = 0; i < num_v; )
         {
             if (rec_md)
@@ -1623,10 +1641,10 @@ static int check_limit_local(struct client *cl,
                 }
                 else
                 {
-                    yaz_log(YLOG_LOG, "cmp: '%s' '%s'",
-                            rec_md->data.text.disp, values[i]);
+                    yaz_log(YLOG_DEBUG, "cmp: '%s' '%s'", rec_md->data.text.disp, values[i]);
                     if (!strcmp(rec_md->data.text.disp, values[i]))
                     {
+                        // Value equals, should not be filtered.
                         break;
                     }
                 }
@@ -1638,6 +1656,7 @@ static int check_limit_local(struct client *cl,
                 i++;
             }
         }
+        // At end , not match
         if (i == num_v)
         {
             skip_record = 1;
@@ -1776,6 +1795,8 @@ static int ingest_to_cluster(struct client *cl,
             struct record_metadata *rec_md = 0;
             int md_field_id = -1;
             int sk_field_id = -1;
+            int rank = 0;
+            xmlChar *rank_str = 0;
             
             type = xmlGetProp(n, (xmlChar *) "type");
             value = xmlNodeListGetString(xdoc, n->children, 1);
@@ -1789,6 +1810,15 @@ static int ingest_to_cluster(struct client *cl,
                 continue;
             
             ser_md = &service->metadata[md_field_id];
+
+            rank_str = xmlGetProp(n, (xmlChar *) "rank");
+            if (rank_str)
+            {
+                rank = atoi((const char *) rank_str);
+                xmlFree(rank_str);
+            }
+            else
+                rank = ser_md->rank;
             
             if (ser_md->sortkey_offset >= 0)
             {
@@ -1888,10 +1918,11 @@ static int ingest_to_cluster(struct client *cl,
 
 
             // ranking of _all_ fields enabled ... 
-            if (ser_md->rank)
+            if (rank)
+            {
                 relevance_countwords(se->relevance, cluster, 
-                                     (char *) value, ser_md->rank,
-                                     ser_md->name);
+                                     (char *) value, rank, ser_md->name);
+            }
 
             // construct facets ... unless the client already has reported them
             if (ser_md->termlist && !client_has_facet(cl, (char *) type))