X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=src%2Freclists.c;h=8ef9ba4cc34e4337300f433bca02f8399594acdb;hb=d08893d173565ccde81b9f7cbf487a67cee5532a;hp=beee724bd5a64222a8ea575d9b28430c78e475c2;hpb=cc29eab9f928f6cd0f4231cb2e554e2ac7b0b1f3;p=pazpar2-moved-to-github.git

diff --git a/src/reclists.c b/src/reclists.c
index beee724..8ef9ba4 100644
--- a/src/reclists.c
+++ b/src/reclists.c
@@ -1,7 +1,5 @@
-/* $Id: reclists.c,v 1.8 2007-04-10 08:48:56 adam Exp $
-   Copyright (c) 2006-2007, Index Data.
-
-This file is part of Pazpar2.
+/* This file is part of Pazpar2.
+   Copyright (C) 2006-2013 Index Data
 
 Pazpar2 is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free
@@ -14,38 +12,49 @@ FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 for more details.
 
 You should have received a copy of the GNU General Public License
-along with Pazpar2; see the file LICENSE.  If not, write to the
-Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA
-02111-1307, USA.
- */
+along with this program; if not, write to the Free Software
+Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+
+*/
 
 #include <assert.h>
 
 #if HAVE_CONFIG_H
-#include <cconfig.h>
+#include <config.h>
 #endif
 
 #include <yaz/yaz-util.h>
 
-#include "pazpar2.h"
+#include "ppmutex.h"
+#include "session.h"
 #include "reclists.h"
+#include "jenkins_hash.h"
 
-extern struct parameters global_parameters;
+struct reclist
+{
+    struct reclist_bucket **hashtable;
+    unsigned hash_size;
 
-// Not threadsafe
-static struct reclist_sortparms *sortparms = 0;
+    int num_records;
+    struct reclist_bucket *sorted_list;
+    struct reclist_bucket *sorted_ptr;
+    NMEM nmem;
+    YAZ_MUTEX mutex;
+};
 
 struct reclist_bucket
 {
     struct record_cluster *record;
-    struct reclist_bucket *next;
+    struct reclist_bucket *hash_next;
+    struct reclist_bucket *sorted_next;
+    struct reclist_sortparms *sort_parms;
 };
 
-struct reclist_sortparms *reclist_parse_sortparms(NMEM nmem, const char *parms)
+struct reclist_sortparms *reclist_parse_sortparms(NMEM nmem, const char *parms,
+                                                  struct conf_service *service)
 {
     struct reclist_sortparms *res = 0;
     struct reclist_sortparms **rp = &res;
-    struct conf_service *service = config->servers->service;
 
     if (strlen(parms) > 256)
         return 0;
@@ -54,55 +63,76 @@ struct reclist_sortparms *reclist_parse_sortparms(NMEM nmem, const char *parms)
         char parm[256];
         char *pp;
         const char *cpp;
-        int increasing;
+        int increasing = 0;
         int i;
-        int offset;
-        enum conf_sortkey_type type;
+        int offset = 0;
+        enum conf_sortkey_type type = Metadata_sortkey_string;
         struct reclist_sortparms *new;
 
         if (!(cpp = strchr(parms, ',')))
             cpp = parms + strlen(parms);
-        strncpy(parm, parms, cpp - parms); 
+        strncpy(parm, parms, cpp - parms);
         parm[cpp-parms] = '\0';
 
         if ((pp = strchr(parm, ':')))
         {
-            increasing = pp[1] == '1' ? 1 : 0;
+            if (pp[1] == '1')
+                increasing = 1;
+            else if (pp[1] == '0')
+                increasing = 0;
+            else
+            {
+                yaz_log(YLOG_FATAL, "Bad sortkey modifier: %s", parm);
+                return 0;
+            }
+
+            if (pp[2])
+            {
+                if (pp[2] == 'p')
+                    type = Metadata_sortkey_position;
+                else
+                    yaz_log(YLOG_FATAL, "Bad sortkey modifier: %s", parm);
+            }
             *pp = '\0';
         }
-        else
-            increasing = 0;
-        if (!strcmp(parm, "relevance"))
+        if (type != Metadata_sortkey_position)
         {
-            type = Metadata_sortkey_relevance;
-            offset = -1;
-        }
-        else
-        {
-            for (i = 0; i < service->num_sortkeys; i++)
+            if (!strcmp(parm, "relevance"))
             {
-                struct conf_sortkey *sk = &service->sortkeys[i];
-                if (!strcmp(sk->name, parm))
-                {
-                    type = sk->type;
-                    if (type == Metadata_sortkey_skiparticle)
-                        type = Metadata_sortkey_string;
-                    break;
-                }
+                type = Metadata_sortkey_relevance;
             }
-            if (i >= service->num_sortkeys)
+            else if (!strcmp(parm, "position"))
             {
-                yaz_log(YLOG_FATAL, "Bad sortkey: %s", parm);
-                return 0;
+                type = Metadata_sortkey_position;
             }
             else
+            {
+                for (i = 0; i < service->num_sortkeys; i++)
+                {
+                    struct conf_sortkey *sk = &service->sortkeys[i];
+                    if (!strcmp(sk->name, parm))
+                    {
+                        type = sk->type;
+                        if (type == Metadata_sortkey_skiparticle)
+                            type = Metadata_sortkey_string;
+                        break;
+                    }
+                }
+                if (i >= service->num_sortkeys)
+                {
+                    yaz_log(YLOG_FATAL, "Sortkey not defined in service: %s",
+                            parm);
+                    return 0;
+                }
                 offset = i;
+            }
         }
         new = *rp = nmem_malloc(nmem, sizeof(struct reclist_sortparms));
         new->next = 0;
         new->offset = offset;
         new->type = type;
         new->increasing = increasing;
+        new->name = nmem_strdup(nmem, parm);
         rp = &new->next;
         if (*(parms = cpp))
             parms++;
@@ -112,153 +142,299 @@ struct reclist_sortparms *reclist_parse_sortparms(NMEM nmem, const char *parms)
 
 static int reclist_cmp(const void *p1, const void *p2)
 {
-    struct record_cluster *r1 = (*(struct record_cluster**) p1);
-    struct record_cluster *r2 = (*(struct record_cluster**) p2);
+    struct reclist_sortparms *sortparms =
+        (*(struct reclist_bucket **) p1)->sort_parms;
+    struct record_cluster *r1 = (*(struct reclist_bucket**) p1)->record;
+    struct record_cluster *r2 = (*(struct reclist_bucket**) p2)->record;
     struct reclist_sortparms *s;
+    int res = 0;
 
-    for (s = sortparms; s; s = s->next)
+    for (s = sortparms; s && res == 0; s = s->next)
     {
-        int res;
+        union data_types *ut1 = r1->sortkeys[s->offset];
+        union data_types *ut2 = r2->sortkeys[s->offset];
+        const char *s1, *s2;
         switch (s->type)
         {
-            case Metadata_sortkey_relevance:
-                res = r2->relevance - r1->relevance;
-                break;
-            case Metadata_sortkey_string:
-                res = strcmp(r2->sortkeys[s->offset]->text, r1->sortkeys[s->offset]->text);
-                break;
-            case Metadata_sortkey_numeric:
+        case Metadata_sortkey_relevance:
+            res = r2->relevance_score - r1->relevance_score;
+            break;
+        case Metadata_sortkey_string:
+            s1 = ut1 ? ut1->text.sort : "";
+            s2 = ut2 ? ut2->text.sort : "";
+            res = strcmp(s2, s1);
+            if (res)
+            {
+                if (s->increasing)
+                    res *= -1;
+            }
+            break;
+        case Metadata_sortkey_numeric:
+            if (ut1 && ut2)
+            {
+                if (s->increasing)
+                    res = ut1->number.min  - ut2->number.min;
+                else
+                    res = ut2->number.max  - ut1->number.max;
+            }
+            else if (ut1 && !ut2)
+                res = -1;
+            else if (!ut1 && ut2)
+                res = 1;
+            else
                 res = 0;
-                break;
-            default:
-                yaz_log(YLOG_FATAL, "Bad sort type: %d", s->type);
-                exit(1);
+            break;
+        case Metadata_sortkey_position:
+            if (r1->records && r2->records)
+            {
+                int pos1 = 0, pos2 = 0;
+                struct record *rec;
+                for (rec = r1->records; rec; rec = rec->next)
+                    if (pos1 == 0 || rec->position < pos1)
+                        pos1 = rec->position;
+                for (rec = r2->records; rec; rec = rec->next)
+                    if (pos2 == 0 || rec->position < pos2)
+                        pos2 = rec->position;
+                res = pos1 - pos2;
+            }
+            break;
+        default:
+            yaz_log(YLOG_WARN, "Bad sort type: %d", s->type);
+            res = 0;
+            break;
         }
-        if (res)
+    }
+    if (res == 0)
+        res = strcmp(r1->recid, r2->recid);
+    return res;
+}
+
+void reclist_limit(struct reclist *l, struct session *se, int lazy)
+{
+    unsigned i;
+    int num = 0;
+    struct reclist_bucket **pp = &l->sorted_list;
+
+    reclist_enter(l);
+
+    if (!lazy || !*pp)
+    {
+        for (i = 0; i < l->hash_size; i++)
         {
-            if (s->increasing)
-                res *= -1;
-            return res;
+            struct reclist_bucket *p;
+            for (p = l->hashtable[i]; p; p = p->hash_next)
+            {
+                if (session_check_cluster_limit(se, p->record))
+                {
+                    *pp = p;
+                    pp = &p->sorted_next;
+                    num++;
+                }
+            }
         }
+        *pp = 0;
     }
-    return 0;
+    l->num_records = num;
+    reclist_leave(l);
 }
 
 void reclist_sort(struct reclist *l, struct reclist_sortparms *parms)
 {
-    sortparms = parms;
-    qsort(l->flatlist, l->num_records, sizeof(struct record_cluster*), reclist_cmp);
-    reclist_rewind(l);
+    struct reclist_bucket **flatlist = xmalloc(sizeof(*flatlist) * l->num_records);
+    struct reclist_bucket *ptr;
+    struct reclist_bucket **prev;
+    int i = 0;
+
+    reclist_enter(l);
+
+    ptr = l->sorted_list;
+    prev = &l->sorted_list;
+    while (ptr)
+    {
+        ptr->sort_parms = parms;
+        flatlist[i] = ptr;
+        ptr = ptr->sorted_next;
+        i++;
+    }
+    assert(i == l->num_records);
+
+    qsort(flatlist, l->num_records, sizeof(*flatlist), reclist_cmp);
+    for (i = 0; i < l->num_records; i++)
+    {
+        *prev = flatlist[i];
+        prev = &flatlist[i]->sorted_next;
+    }
+    *prev = 0;
+
+    xfree(flatlist);
+
+    reclist_leave(l);
 }
 
 struct record_cluster *reclist_read_record(struct reclist *l)
 {
-    if (l->pointer < l->num_records)
-        return l->flatlist[l->pointer++];
+    if (l && l->sorted_ptr)
+    {
+        struct record_cluster *t = l->sorted_ptr->record;
+        l->sorted_ptr = l->sorted_ptr->sorted_next;
+        return t;
+    }
     else
         return 0;
 }
 
-void reclist_rewind(struct reclist *l)
+void reclist_enter(struct reclist *l)
 {
-    l->pointer = 0;
+    yaz_mutex_enter(l->mutex);
+    if (l)
+        l->sorted_ptr = l->sorted_list;
 }
 
-// Jenkins one-at-a-time hash (from wikipedia)
-static unsigned int hash(const unsigned char *key)
-{
-    unsigned int hash = 0;
 
-    while (*key)
-    {
-        hash += *(key++);
-        hash += (hash << 10);
-        hash ^= (hash >> 6);
-    }
-    hash += (hash << 3);
-    hash ^= (hash >> 11);
-    hash += (hash << 15);
-    return hash;
+void reclist_leave(struct reclist *l)
+{
+    yaz_mutex_leave(l->mutex);
+    if (l)
+        l->sorted_ptr = l->sorted_list;
 }
 
-struct reclist *reclist_create(NMEM nmem, int numrecs)
+
+struct reclist *reclist_create(NMEM nmem)
 {
-    int hashsize = 1;
-    struct reclist *res;
-
-    assert(numrecs);
-    while (hashsize < numrecs)
-        hashsize <<= 1;
-    res = nmem_malloc(nmem, sizeof(struct reclist));
-    res->hashtable = nmem_malloc(nmem, hashsize * sizeof(struct reclist_bucket*));
-    memset(res->hashtable, 0, hashsize * sizeof(struct reclist_bucket*));
-    res->hashtable_size = hashsize;
+    struct reclist *res = nmem_malloc(nmem, sizeof(struct reclist));
+    res->hash_size = 399;
+    res->hashtable
+        = nmem_malloc(nmem, res->hash_size * sizeof(struct reclist_bucket*));
+    memset(res->hashtable, 0, res->hash_size * sizeof(struct reclist_bucket*));
     res->nmem = nmem;
-    res->hashmask = hashsize - 1; // Creates a bitmask
 
-    res->num_records = 0;
-    res->flatlist = nmem_malloc(nmem, numrecs * sizeof(struct record_cluster*));
-    res->flatlist_size = numrecs;
+    res->sorted_ptr = 0;
+    res->sorted_list = 0;
 
+    res->num_records = 0;
+    res->mutex = 0;
+    pazpar2_mutex_create(&res->mutex, "reclist");
     return res;
 }
 
+void reclist_destroy(struct reclist *l)
+{
+    if (l)
+    {
+        unsigned i;
+        for (i = 0; i < l->hash_size; i++)
+        {
+            struct reclist_bucket *p;
+            for (p = l->hashtable[i]; p; p = p->hash_next)
+            {
+                wrbuf_destroy(p->record->relevance_explain1);
+                wrbuf_destroy(p->record->relevance_explain2);
+            }
+        }
+        yaz_mutex_destroy(&l->mutex);
+    }
+}
+
+int reclist_get_num_records(struct reclist *l)
+{
+    if (l)
+        return l->num_records;
+    return 0;
+}
+
 // Insert a record. Return record cluster (newly formed or pre-existing)
-struct record_cluster *reclist_insert(struct reclist *l, struct record  *record,
-        char *merge_key, int *total)
+struct record_cluster *reclist_insert(struct reclist *l,
+                                      struct conf_service *service,
+                                      struct record *record,
+                                      const char *merge_key, int *total)
 {
     unsigned int bucket;
     struct reclist_bucket **p;
     struct record_cluster *cluster = 0;
-    struct conf_service *service = global_parameters.server->service;
 
-    bucket = hash((unsigned char*) merge_key) & l->hashmask;
-    for (p = &l->hashtable[bucket]; *p; p = &(*p)->next)
+    assert(service);
+    assert(l);
+    assert(record);
+    assert(merge_key);
+    assert(total);
+
+    bucket = jenkins_hash((unsigned char*) merge_key) % l->hash_size;
+
+    yaz_mutex_enter(l->mutex);
+    for (p = &l->hashtable[bucket]; *p; p = &(*p)->hash_next)
     {
         // We found a matching record. Merge them
         if (!strcmp(merge_key, (*p)->record->merge_key))
         {
-            struct record_cluster *existing = (*p)->record;
-            record->next = existing->records;
-            existing->records = record;
-            cluster = existing;
+            struct record **re;
+
+            cluster = (*p)->record;
+            for (re = &cluster->records; *re; re = &(*re)->next)
+            {
+                if ((*re)->client == record->client &&
+                    record_compare(record, *re, service))
+                {
+                    yaz_mutex_leave(l->mutex);
+                    return 0;
+                }
+            }
+            *re = record;
+            record->next = 0;
             break;
         }
     }
-    if (!cluster && l->num_records < l->flatlist_size)
+    if (!cluster)
     {
         struct reclist_bucket *new =
-            nmem_malloc(l->nmem, sizeof(struct reclist_bucket));
-        struct record_cluster *newc =
-            nmem_malloc(l->nmem, sizeof(struct record_cluster));
-        
+            nmem_malloc(l->nmem, sizeof(*new));
+
+        cluster = nmem_malloc(l->nmem, sizeof(*cluster));
+
         record->next = 0;
-        new->record = newc;
-        new->next = 0;
-        newc->records = record;
-        newc->merge_key = merge_key;
-        newc->relevance = 0;
-        newc->term_frequency_vec = 0;
-        newc->recid = (*total)++;
-        newc->metadata = nmem_malloc(l->nmem,
-                sizeof(struct record_metadata*) * service->num_metadata);
-        memset(newc->metadata, 0, sizeof(struct record_metadata*) * service->num_metadata);
-        newc->sortkeys = nmem_malloc(l->nmem,
-                sizeof(struct record_metadata*) * service->num_sortkeys);
-        memset(newc->sortkeys, 0, sizeof(union data_types*) * service->num_sortkeys);
+        new->record = cluster;
+        new->hash_next = 0;
+        cluster->records = record;
+        cluster->merge_key = nmem_strdup(l->nmem, merge_key);
+        cluster->relevance_score = 0;
+        cluster->term_frequency_vec = 0;
+        cluster->recid = nmem_strdup(l->nmem, merge_key);
+        (*total)++;
+        cluster->metadata =
+            nmem_malloc(l->nmem,
+                        sizeof(struct record_metadata*) * service->num_metadata);
+        memset(cluster->metadata, 0,
+               sizeof(struct record_metadata*) * service->num_metadata);
+        cluster->sortkeys =
+            nmem_malloc(l->nmem, sizeof(struct record_metadata*) * service->num_sortkeys);
+        memset(cluster->sortkeys, 0,
+               sizeof(union data_types*) * service->num_sortkeys);
 
+        cluster->relevance_explain1 = wrbuf_alloc();
+        cluster->relevance_explain2 = wrbuf_alloc();
+        /* attach to hash list */
         *p = new;
-        l->flatlist[l->num_records++] = newc;
-        cluster = newc;
+        l->num_records++;
     }
+    yaz_mutex_leave(l->mutex);
     return cluster;
 }
 
-
+int reclist_sortparms_cmp(struct reclist_sortparms *sort1, struct reclist_sortparms *sort2)
+{
+    int rc;
+    if (sort1 == sort2)
+        return 0;
+    if (sort1 == 0 || sort2 == 0)
+        return 1;
+    rc = strcmp(sort1->name, sort2->name) || sort1->increasing != sort2->increasing || sort1->type != sort2->type;
+    return rc;
+}
 /*
  * Local variables:
  * c-basic-offset: 4
+ * c-file-style: "Stroustrup"
  * indent-tabs-mode: nil
  * End:
  * vim: shiftwidth=4 tabstop=8 expandtab
  */
+