X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=src%2Freclists.c;h=1b431698021e5e6b31b4c47ca8257fb07c8295eb;hb=fb19bc317370b4f4b508088b034b4c2d19257042;hp=2594430fe411617a331fc11c536e71487e429cf4;hpb=ec99fbbff55468137dd755ad3245eb5e62bcc885;p=pazpar2-moved-to-github.git

diff --git a/src/reclists.c b/src/reclists.c
index 2594430..1b43169 100644
--- a/src/reclists.c
+++ b/src/reclists.c
@@ -1,5 +1,5 @@
 /* This file is part of Pazpar2.
-   Copyright (C) 2006-2012 Index Data
+   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
@@ -36,8 +36,8 @@ struct reclist
     unsigned hash_size;
 
     int num_records;
-    struct reclist_bucket *sorted_list;
-    struct reclist_bucket *sorted_ptr;
+    struct record_cluster *sorted_list;
+    struct record_cluster *sorted_ptr;
     NMEM nmem;
     YAZ_MUTEX mutex;
 };
@@ -46,10 +46,28 @@ struct reclist_bucket
 {
     struct record_cluster *record;
     struct reclist_bucket *hash_next;
-    struct reclist_bucket *sorted_next;
-    struct reclist_sortparms *sort_parms;
 };
 
+static void append_merge_keys(struct record_metadata_attr **p,
+                              const struct record_metadata_attr *a,
+                              NMEM nmem)
+{
+    for (; a; a = a->next)
+    {
+        struct record_metadata_attr **pi = p;
+        for (; *pi; pi = &(*pi)->next)
+            if (!strcmp((*pi)->value, a->value))
+                break;
+        if (!*pi)
+        {
+            *pi = (struct record_metadata_attr *) nmem_malloc(nmem, sizeof(**p));
+            (*pi)->name = nmem_strdup_null(nmem, a->name);
+            (*pi)->value = nmem_strdup_null(nmem, a->value);
+            (*pi)->next = 0;
+        }
+    }
+}
+
 struct reclist_sortparms *reclist_parse_sortparms(NMEM nmem, const char *parms,
                                                   struct conf_service *service)
 {
@@ -66,7 +84,7 @@ struct reclist_sortparms *reclist_parse_sortparms(NMEM nmem, const char *parms,
         int increasing = 0;
         int i;
         int offset = 0;
-        enum conf_sortkey_type type = Metadata_sortkey_string;
+        enum conf_metadata_type type = Metadata_type_generic;
         struct reclist_sortparms *new;
 
         if (!(cpp = strchr(parms, ',')))
@@ -89,21 +107,21 @@ struct reclist_sortparms *reclist_parse_sortparms(NMEM nmem, const char *parms,
             if (pp[2])
             {
                 if (pp[2] == 'p')
-                    type = Metadata_sortkey_position;
+                    type = Metadata_type_position;
                 else
                     yaz_log(YLOG_FATAL, "Bad sortkey modifier: %s", parm);
             }
             *pp = '\0';
         }
-        if (type != Metadata_sortkey_position)
+        if (type != Metadata_type_position)
         {
             if (!strcmp(parm, "relevance"))
             {
-                type = Metadata_sortkey_relevance;
+                type = Metadata_type_relevance;
             }
             else if (!strcmp(parm, "position"))
             {
-                type = Metadata_sortkey_position;
+                type = Metadata_type_position;
             }
             else
             {
@@ -113,8 +131,6 @@ struct reclist_sortparms *reclist_parse_sortparms(NMEM nmem, const char *parms,
                     if (!strcmp(sk->name, parm))
                     {
                         type = sk->type;
-                        if (type == Metadata_sortkey_skiparticle)
-                            type = Metadata_sortkey_string;
                         break;
                     }
                 }
@@ -142,10 +158,9 @@ struct reclist_sortparms *reclist_parse_sortparms(NMEM nmem, const char *parms,
 
 static int reclist_cmp(const void *p1, const void *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 record_cluster *r1 = (*(struct record_cluster**) p1);
+    struct record_cluster *r2 = (*(struct record_cluster**) p2);
+    struct reclist_sortparms *sortparms = r1->sort_parms;
     struct reclist_sortparms *s;
     int res = 0;
 
@@ -156,35 +171,38 @@ static int reclist_cmp(const void *p1, const void *p2)
         const char *s1, *s2;
         switch (s->type)
         {
-        case Metadata_sortkey_relevance:
-            res = r2->relevance_score - r1->relevance_score;
+        case Metadata_type_relevance:
+            res = r1->relevance_score - r2->relevance_score;
             break;
-        case Metadata_sortkey_string:
+        case Metadata_type_generic:
+        case Metadata_type_skiparticle:
             s1 = ut1 ? ut1->text.sort : "";
             s2 = ut2 ? ut2->text.sort : "";
-            res = strcmp(s2, s1);
-            if (res)
-            {
-                if (s->increasing)
-                    res *= -1;
-            }
+            res = strcmp(s1, s2);
             break;
-        case Metadata_sortkey_numeric:
+        case Metadata_type_year:
+        case Metadata_type_date:
             if (ut1 && ut2)
             {
                 if (s->increasing)
                     res = ut1->number.min  - ut2->number.min;
                 else
-                    res = ut2->number.max  - ut1->number.max;
+                    res = ut1->number.max  - ut2->number.max;
             }
             else if (ut1 && !ut2)
-                res = -1;
+            {
+                res = -1; /* without date/year: last! */
+                continue;
+            }
             else if (!ut1 && ut2)
-                res = 1;
+            {
+                res = 1; /* without date/year: last! */
+                continue;
+            }
             else
                 res = 0;
             break;
-        case Metadata_sortkey_position:
+        case Metadata_type_position:
             if (r1->records && r2->records)
             {
                 int pos1 = 0, pos2 = 0;
@@ -196,49 +214,79 @@ static int reclist_cmp(const void *p1, const void *p2)
                     if (pos2 == 0 || rec->position < pos2)
                         pos2 = rec->position;
                 res = pos1 - pos2;
+                continue;
             }
             break;
-        default:
-            yaz_log(YLOG_WARN, "Bad sort type: %d", s->type);
-            res = 0;
+        case Metadata_type_float:
+            if (ut1 && ut2)
+            {
+                if (ut1->fnumber == ut2->fnumber)
+                    res = 0;
+                else if (ut1->fnumber > ut2->fnumber)
+                    res = 1;
+                else
+                    res = -1;
+            }
+            else if (ut1)
+                res = 1;
+            else if (ut2)
+                res = -1;
+            else
+                res = 0;
             break;
         }
+        if (res && !s->increasing)
+            res *= -1;
     }
     if (res == 0)
         res = strcmp(r1->recid, r2->recid);
     return res;
 }
 
-void reclist_limit(struct reclist *l, struct session *se)
+void reclist_limit(struct reclist *l, struct session *se, int lazy)
 {
     unsigned i;
     int num = 0;
-    struct reclist_bucket **pp = &l->sorted_list;
+    struct record_cluster **pp = &l->sorted_list;
 
     reclist_enter(l);
-    for (i = 0; i < l->hash_size; i++)
+
+    if (!lazy || !*pp)
     {
-        struct reclist_bucket *p;
-        for (p = l->hashtable[i]; p; p = p->hash_next)
+        for (i = 0; i < l->hash_size; i++)
         {
-            if (session_check_cluster_limit(se, p->record))
+            struct reclist_bucket *p;
+            for (p = l->hashtable[i]; p; p = p->hash_next)
+                p->record->sorted_next = 0;
+        }
+        for (i = 0; i < l->hash_size; i++)
+        {
+            struct reclist_bucket *p;
+            for (p = l->hashtable[i]; p; p = p->hash_next)
             {
-                *pp = p;
-                pp = &p->sorted_next;
-                num++;
+                if (session_check_cluster_limit(se, p->record))
+                {
+                    if (!p->record->sorted_next)
+                    {
+                        *pp = p->record;
+                        pp = &p->record->sorted_next;
+                        *pp = p->record; /* signal already in use */
+                        num++;
+                    }
+                }
             }
         }
+        *pp = 0;
     }
-    *pp = 0;
     l->num_records = num;
     reclist_leave(l);
 }
 
 void reclist_sort(struct reclist *l, struct reclist_sortparms *parms)
 {
-    struct reclist_bucket **flatlist = xmalloc(sizeof(*flatlist) * l->num_records);
-    struct reclist_bucket *ptr;
-    struct reclist_bucket **prev;
+    struct record_cluster **flatlist = xmalloc(sizeof(*flatlist) * l->num_records);
+    struct record_cluster *ptr;
+    struct record_cluster **prev;
     int i = 0;
 
     reclist_enter(l);
@@ -271,7 +319,7 @@ struct record_cluster *reclist_read_record(struct reclist *l)
 {
     if (l && l->sorted_ptr)
     {
-        struct record_cluster *t = l->sorted_ptr->record;
+        struct record_cluster *t = l->sorted_ptr;
         l->sorted_ptr = l->sorted_ptr->sorted_next;
         return t;
     }
@@ -325,6 +373,8 @@ void reclist_destroy(struct reclist *l)
             {
                 wrbuf_destroy(p->record->relevance_explain1);
                 wrbuf_destroy(p->record->relevance_explain2);
+                p->record->relevance_explain1 = 0;
+                p->record->relevance_explain2 = 0;
             }
         }
         yaz_mutex_destroy(&l->mutex);
@@ -338,79 +388,162 @@ int reclist_get_num_records(struct reclist *l)
     return 0;
 }
 
+static void merge_cluster(struct reclist *l,
+                          struct relevance *r,
+                          struct record_cluster *dst,
+                          struct record_cluster *src)
+{
+    struct record_metadata_attr *mkl;
+    struct record **rp = &dst->records;
+    for (; *rp; rp = &(*rp)->next)
+        ;
+    *rp = src->records;
+
+    for (mkl = src->merge_keys; mkl; mkl = mkl->next)
+    {
+        const char *merge_key = mkl->value;
+        unsigned int bucket =
+            jenkins_hash((unsigned char*) merge_key) % l->hash_size;
+        struct reclist_bucket *p;
+
+        for (p = l->hashtable[bucket]; p; p = p->hash_next)
+            if (p->record == src)
+                p->record = dst;
+    }
+
+    /* not merging metadata and sortkeys yet */
+
+    relevance_mergerec(r, dst, src);
+
+    wrbuf_puts(dst->relevance_explain1, wrbuf_cstr(src->relevance_explain1));
+    wrbuf_puts(dst->relevance_explain2, wrbuf_cstr(src->relevance_explain2));
+
+    wrbuf_destroy(src->relevance_explain1);
+    src->relevance_explain1 = 0;
+    wrbuf_destroy(src->relevance_explain2);
+    src->relevance_explain2 = 0;
+
+    append_merge_keys(&dst->merge_keys, src->merge_keys, l->nmem);
+}
+
+static struct record_cluster *new_cluster(
+    struct reclist *l,
+    struct relevance *r,
+    struct conf_service *service,
+    struct record *record,
+    struct record_metadata_attr *merge_keys
+    )
+{
+    struct record_cluster *cluster;
+    cluster = nmem_malloc(l->nmem, sizeof(*cluster));
+
+    record->next = 0;
+    cluster->records = record;
+    cluster->merge_keys = 0;
+    append_merge_keys(&cluster->merge_keys, merge_keys, l->nmem);
+    cluster->relevance_score = 0;
+    cluster->recid = cluster->merge_keys->value;
+    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);
+    relevance_newrec(r, cluster);
+    cluster->relevance_explain1 = wrbuf_alloc();
+    cluster->relevance_explain2 = wrbuf_alloc();
+    /* attach to hash list */
+    l->num_records++;
+    l->sorted_list = l->sorted_ptr = 0;
+    return cluster;
+}
+
 // Insert a record. Return record cluster (newly formed or pre-existing)
 struct record_cluster *reclist_insert(struct reclist *l,
+                                      struct relevance *r,
                                       struct conf_service *service,
                                       struct record *record,
-                                      const char *merge_key, int *total)
+                                      struct record_metadata_attr *merge_keys,
+                                      int *total)
 {
-    unsigned int bucket;
-    struct reclist_bucket **p;
     struct record_cluster *cluster = 0;
+    struct record_metadata_attr *mkl = merge_keys;
 
     assert(service);
     assert(l);
     assert(record);
-    assert(merge_key);
+    assert(merge_keys);
     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)
+
+    for (; mkl; mkl = mkl->next)
     {
-        // We found a matching record. Merge them
-        if (!strcmp(merge_key, (*p)->record->merge_key))
-        {
-            struct record_cluster *existing = (*p)->record;
-            struct record *re = existing->records;
+        const char *merge_key = mkl->value;
+        unsigned int bucket =
+            jenkins_hash((unsigned char*) merge_key) % l->hash_size;
+        struct reclist_bucket **p;
+        struct reclist_bucket *rb = 0;
 
-            for (; re; re = re->next)
+        for (p = &l->hashtable[bucket]; *p; p = &(*p)->hash_next)
+        {
+            struct record_metadata_attr *mkr = (*p)->record->merge_keys;
+            for (; mkr; mkr = mkr->next)
             {
-                if (re->client == record->client &&
-                    record_compare(record, re, service))
+                // We found a matching record. Merge them
+                if (!strcmp(merge_key, mkr->value))
                 {
-                    yaz_mutex_leave(l->mutex);
-                    return 0;
+                    struct record **re;
+
+                    rb = *p;
+                    for (re = &rb->record->records; *re; re = &(*re)->next)
+                    {
+                        if ((*re)->client == record->client &&
+                            record_compare(record, *re, service))
+                        {
+                            yaz_mutex_leave(l->mutex);
+                            return 0;
+                        }
+                    }
+
+                    if (!cluster)
+                    {
+                        cluster = rb->record;
+                        *re = record;
+                        record->next = 0;
+                        yaz_log(YLOG_LOG, "reclist: record insert %p", cluster);
+                    }
+                    else
+                    {
+                        if (cluster != rb->record)
+                        {
+                            assert(rb->record->relevance_explain1);
+                            yaz_log(YLOG_LOG, "reclist: cluster merge %p %p", cluster, rb->record);
+                            merge_cluster(l, r, cluster, rb->record);
+                            (*total)--;
+                        }
+                    }
                 }
             }
-            record->next = existing->records;
-            existing->records = record;
-            cluster = existing;
-            break;
         }
-    }
-    if (!cluster)
-    {
-        struct reclist_bucket *new =
-            nmem_malloc(l->nmem, sizeof(*new));
-
-        cluster = nmem_malloc(l->nmem, sizeof(*cluster));
-
-        record->next = 0;
-        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->num_records++;
+        if (!cluster)
+        {
+            (*total)++;
+            cluster = new_cluster(l, r, service, record, merge_keys);
+            yaz_log(YLOG_LOG, "reclist: new cluster p=%p", cluster);
+        }
+
+        if (!rb)
+        {
+            rb = nmem_malloc(l->nmem, sizeof(*rb));
+            rb->record = cluster;
+            rb->hash_next = 0;
+
+            *p = rb;
+        }
     }
     yaz_mutex_leave(l->mutex);
     return cluster;