1 /* This file is part of Pazpar2.
2 Copyright (C) Index Data
4 Pazpar2 is free software; you can redistribute it and/or modify it under
5 the terms of the GNU General Public License as published by the Free
6 Software Foundation; either version 2, or (at your option) any later
9 Pazpar2 is distributed in the hope that it will be useful, but WITHOUT ANY
10 WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
14 You should have received a copy of the GNU General Public License
15 along with this program; if not, write to the Free Software
16 Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
26 #include <yaz/yaz-util.h>
31 #include "jenkins_hash.h"
35 struct reclist_bucket **hashtable;
39 struct record_cluster *sorted_list;
40 struct record_cluster *sorted_ptr;
47 struct record_cluster *record;
48 struct reclist_bucket *hash_next;
51 static void append_merge_keys(struct record_metadata_attr **p,
52 const struct record_metadata_attr *a,
55 for (; a; a = a->next)
57 struct record_metadata_attr **pi = p;
58 for (; *pi; pi = &(*pi)->next)
59 if (!strcmp((*pi)->value, a->value))
63 *pi = (struct record_metadata_attr *) nmem_malloc(nmem, sizeof(**p));
64 (*pi)->name = nmem_strdup_null(nmem, a->name);
65 (*pi)->value = nmem_strdup_null(nmem, a->value);
71 struct reclist_sortparms *reclist_parse_sortparms(NMEM nmem, const char *parms,
72 struct conf_service *service)
74 struct reclist_sortparms *res = 0;
75 struct reclist_sortparms **rp = &res;
77 if (strlen(parms) > 256)
87 enum conf_metadata_type type = Metadata_type_generic;
88 struct reclist_sortparms *new;
90 if (!(cpp = strchr(parms, ',')))
91 cpp = parms + strlen(parms);
92 strncpy(parm, parms, cpp - parms);
93 parm[cpp-parms] = '\0';
95 if ((pp = strchr(parm, ':')))
99 else if (pp[1] == '0')
103 yaz_log(YLOG_FATAL, "Bad sortkey modifier: %s", parm);
110 type = Metadata_type_position;
112 yaz_log(YLOG_FATAL, "Bad sortkey modifier: %s", parm);
116 if (type != Metadata_type_position)
118 if (!strcmp(parm, "relevance"))
120 type = Metadata_type_relevance;
122 else if (!strcmp(parm, "position"))
124 type = Metadata_type_position;
126 else if (!strcmp(parm, "retrieval"))
128 type = Metadata_type_retrieval;
132 for (i = 0; i < service->num_sortkeys; i++)
134 struct conf_sortkey *sk = &service->sortkeys[i];
135 if (!strcmp(sk->name, parm))
141 if (i >= service->num_sortkeys)
143 yaz_log(YLOG_FATAL, "Sortkey not defined in service: %s",
150 new = *rp = nmem_malloc(nmem, sizeof(struct reclist_sortparms));
152 new->offset = offset;
154 new->increasing = increasing;
155 new->name = nmem_strdup(nmem, parm);
163 static int reclist_cmp(const void *p1, const void *p2)
165 struct record_cluster *r1 = (*(struct record_cluster**) p1);
166 struct record_cluster *r2 = (*(struct record_cluster**) p2);
167 struct reclist_sortparms *sortparms = r1->sort_parms;
168 struct reclist_sortparms *s;
171 for (s = sortparms; s && res == 0; s = s->next)
173 union data_types *ut1 = r1->sortkeys[s->offset];
174 union data_types *ut2 = r2->sortkeys[s->offset];
178 case Metadata_type_relevance:
179 res = r1->relevance_score - r2->relevance_score;
181 case Metadata_type_generic:
182 case Metadata_type_skiparticle:
183 s1 = ut1 ? ut1->text.sort : "";
184 s2 = ut2 ? ut2->text.sort : "";
185 res = strcmp(s1, s2);
187 case Metadata_type_year:
188 case Metadata_type_date:
192 res = ut1->number.min - ut2->number.min;
194 res = ut1->number.max - ut2->number.max;
196 else if (ut1 && !ut2)
198 res = -1; /* without date/year: last! */
201 else if (!ut1 && ut2)
203 res = 1; /* without date/year: last! */
209 case Metadata_type_position:
210 if (r1->records && r2->records)
212 int pos1 = 0, pos2 = 0;
214 for (rec = r1->records; rec; rec = rec->next)
215 if (pos1 == 0 || rec->position < pos1)
216 pos1 = rec->position;
217 for (rec = r2->records; rec; rec = rec->next)
218 if (pos2 == 0 || rec->position < pos2)
219 pos2 = rec->position;
224 case Metadata_type_float:
227 if (ut1->fnumber == ut2->fnumber)
229 else if (ut1->fnumber > ut2->fnumber)
241 case Metadata_type_retrieval:
242 res = r1->retrieval_position - r2->retrieval_position;
245 if (res && !s->increasing)
249 res = strcmp(r1->recid, r2->recid);
253 void reclist_limit(struct reclist *l, struct session *se, int lazy)
257 struct record_cluster **pp = &l->sorted_list;
263 for (i = 0; i < l->hash_size; i++)
265 struct reclist_bucket *p;
266 for (p = l->hashtable[i]; p; p = p->hash_next)
267 p->record->sorted_next = 0;
269 for (i = 0; i < l->hash_size; i++)
271 struct reclist_bucket *p;
272 for (p = l->hashtable[i]; p; p = p->hash_next)
274 if (session_check_cluster_limit(se, p->record))
276 if (!p->record->sorted_next)
279 pp = &p->record->sorted_next;
280 *pp = p->record; /* signal already in use */
288 l->num_records = num;
292 void reclist_sort(struct reclist *l, struct reclist_sortparms *parms)
294 struct record_cluster **flatlist = xmalloc(sizeof(*flatlist) * l->num_records);
295 struct record_cluster *ptr;
296 struct record_cluster **prev;
301 ptr = l->sorted_list;
302 prev = &l->sorted_list;
305 ptr->sort_parms = parms;
307 ptr = ptr->sorted_next;
310 assert(i == l->num_records);
312 qsort(flatlist, l->num_records, sizeof(*flatlist), reclist_cmp);
313 for (i = 0; i < l->num_records; i++)
316 prev = &flatlist[i]->sorted_next;
325 struct record_cluster *reclist_read_record(struct reclist *l)
327 if (l && l->sorted_ptr)
329 struct record_cluster *t = l->sorted_ptr;
330 l->sorted_ptr = l->sorted_ptr->sorted_next;
337 void reclist_enter(struct reclist *l)
339 yaz_mutex_enter(l->mutex);
340 l->sorted_ptr = l->sorted_list;
344 void reclist_leave(struct reclist *l)
346 yaz_mutex_leave(l->mutex);
347 l->sorted_ptr = l->sorted_list;
350 void reclist_rewind(struct reclist *l)
352 l->sorted_ptr = l->sorted_list;
356 struct reclist *reclist_create(NMEM nmem)
358 struct reclist *res = nmem_malloc(nmem, sizeof(struct reclist));
359 res->hash_size = 399;
361 = nmem_malloc(nmem, res->hash_size * sizeof(struct reclist_bucket*));
362 memset(res->hashtable, 0, res->hash_size * sizeof(struct reclist_bucket*));
366 res->sorted_list = 0;
368 res->num_records = 0;
370 pazpar2_mutex_create(&res->mutex, "reclist");
374 void reclist_destroy(struct reclist *l)
379 for (i = 0; i < l->hash_size; i++)
381 struct reclist_bucket *p;
382 for (p = l->hashtable[i]; p; p = p->hash_next)
384 wrbuf_destroy(p->record->relevance_explain1);
385 wrbuf_destroy(p->record->relevance_explain2);
386 p->record->relevance_explain1 = 0;
387 p->record->relevance_explain2 = 0;
390 yaz_mutex_destroy(&l->mutex);
394 int reclist_get_num_records(struct reclist *l)
397 return l->num_records;
401 static void merge_cluster(struct reclist *l,
403 struct record_cluster *dst,
404 struct record_cluster *src)
406 struct record_metadata_attr *mkl;
407 struct record **rp = &dst->records;
408 for (; *rp; rp = &(*rp)->next)
412 for (mkl = src->merge_keys; mkl; mkl = mkl->next)
414 const char *merge_key = mkl->value;
415 unsigned int bucket =
416 jenkins_hash((unsigned char*) merge_key) % l->hash_size;
417 struct reclist_bucket *p;
419 for (p = l->hashtable[bucket]; p; p = p->hash_next)
420 if (p->record == src)
424 /* not merging metadata and sortkeys yet */
426 relevance_mergerec(r, dst, src);
428 wrbuf_puts(dst->relevance_explain1, wrbuf_cstr(src->relevance_explain1));
429 wrbuf_puts(dst->relevance_explain2, wrbuf_cstr(src->relevance_explain2));
431 wrbuf_destroy(src->relevance_explain1);
432 src->relevance_explain1 = 0;
433 wrbuf_destroy(src->relevance_explain2);
434 src->relevance_explain2 = 0;
436 append_merge_keys(&dst->merge_keys, src->merge_keys, l->nmem);
439 static struct record_cluster *new_cluster(
442 struct conf_service *service,
443 struct record *record,
444 struct record_metadata_attr *merge_keys
447 struct record_cluster *cluster;
448 cluster = nmem_malloc(l->nmem, sizeof(*cluster));
451 cluster->retrieval_position = l->num_records;
452 cluster->records = record;
453 cluster->merge_keys = 0;
454 append_merge_keys(&cluster->merge_keys, merge_keys, l->nmem);
455 cluster->relevance_score = 0;
456 cluster->recid = cluster->merge_keys->value;
459 sizeof(struct record_metadata*) * service->num_metadata);
460 memset(cluster->metadata, 0,
461 sizeof(struct record_metadata*) * service->num_metadata);
463 nmem_malloc(l->nmem, sizeof(struct record_metadata*) * service->num_sortkeys);
464 memset(cluster->sortkeys, 0,
465 sizeof(union data_types*) * service->num_sortkeys);
466 relevance_newrec(r, cluster);
467 cluster->relevance_explain1 = wrbuf_alloc();
468 cluster->relevance_explain2 = wrbuf_alloc();
469 /* attach to hash list */
471 l->sorted_list = l->sorted_ptr = 0;
475 // Insert a record. Return record cluster (newly formed or pre-existing)
476 struct record_cluster *reclist_insert(struct reclist *l,
478 struct conf_service *service,
479 struct record *record,
480 struct record_metadata_attr *merge_keys,
483 struct record_cluster *cluster = 0;
484 struct record_metadata_attr *mkl = merge_keys;
492 yaz_mutex_enter(l->mutex);
494 for (; mkl; mkl = mkl->next)
496 const char *merge_key = mkl->value;
497 unsigned int bucket =
498 jenkins_hash((unsigned char*) merge_key) % l->hash_size;
499 struct reclist_bucket **p;
500 struct reclist_bucket *rb = 0;
502 for (p = &l->hashtable[bucket]; *p; p = &(*p)->hash_next)
504 struct record_metadata_attr *mkr = (*p)->record->merge_keys;
505 for (; mkr; mkr = mkr->next)
507 // We found a matching record. Merge them
508 if (!strcmp(merge_key, mkr->value))
514 for (re = &rb->record->records; *re; re = &(*re)->next)
516 if ((*re)->client == record->client &&
517 record_compare(record, *re, service))
519 yaz_mutex_leave(l->mutex);
523 cluster = rb->record;
529 if (cluster != rb->record)
531 assert(rb->record->relevance_explain1);
532 merge_cluster(l, r, cluster, rb->record);
542 cluster = new_cluster(l, r, service, record, merge_keys);
547 rb = nmem_malloc(l->nmem, sizeof(*rb));
548 rb->record = cluster;
554 yaz_mutex_leave(l->mutex);
558 int reclist_sortparms_cmp(struct reclist_sortparms *sort1, struct reclist_sortparms *sort2)
563 if (sort1 == 0 || sort2 == 0)
565 rc = strcmp(sort1->name, sort2->name) || sort1->increasing != sort2->increasing || sort1->type != sort2->type;
571 * c-file-style: "Stroustrup"
572 * indent-tabs-mode: nil
574 * vim: shiftwidth=4 tabstop=8 expandtab