1 /* $Id: logic.c,v 1.38 2007-06-07 12:27:03 adam Exp $
2 Copyright (c) 2006-2007, Index Data.
4 This file is part of Pazpar2.
6 Pazpar2 is free software; you can redistribute it and/or modify it under
7 the terms of the GNU General Public License as published by the Free
8 Software Foundation; either version 2, or (at your option) any later
11 Pazpar2 is distributed in the hope that it will be useful, but WITHOUT ANY
12 WARRANTY; without even the implied warranty of MERCHANTABILITY or
13 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
16 You should have received a copy of the GNU General Public License
17 along with Pazpar2; see the file LICENSE. If not, write to the
18 Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA
23 \brief high-level logic; mostly user sessions and settings
31 #include <sys/socket.h>
37 #include <yaz/marcdisp.h>
38 #include <yaz/comstack.h>
39 #include <yaz/tcpip.h>
40 #include <yaz/proto.h>
41 #include <yaz/readconf.h>
42 #include <yaz/pquery.h>
43 #include <yaz/otherinfo.h>
44 #include <yaz/yaz-util.h>
46 #include <yaz/query-charset.h>
47 #include <yaz/querytowrbuf.h>
48 #if YAZ_VERSIONL >= 0x020163
49 #include <yaz/oid_db.h>
58 #include <yaz/timing.h>
61 #include <netinet/in.h>
66 #include "termlists.h"
68 #include "relevance.h"
73 #include "normalize7bit.h"
77 // Note: Some things in this structure will eventually move to configuration
78 struct parameters global_parameters =
99 // Recursively traverse query structure to extract terms.
100 void pull_terms(NMEM nmem, struct ccl_rpn_node *n, char **termlist, int *num)
112 pull_terms(nmem, n->u.p[0], termlist, num);
113 pull_terms(nmem, n->u.p[1], termlist, num);
116 nmem_strsplit(nmem, " ", n->u.t.term, &words, &numwords);
117 for (i = 0; i < numwords; i++)
118 termlist[(*num)++] = words[i];
127 static void add_facet(struct session *s, const char *type, const char *value)
133 for (i = 0; i < s->num_termlists; i++)
134 if (!strcmp(s->termlists[i].name, type))
136 if (i == s->num_termlists)
138 if (i == SESSION_MAX_TERMLISTS)
140 yaz_log(YLOG_FATAL, "Too many termlists");
144 s->termlists[i].name = nmem_strdup(s->nmem, type);
145 s->termlists[i].termlist
146 = termlist_create(s->nmem, s->expected_maxrecs, 15);
147 s->num_termlists = i + 1;
149 termlist_insert(s->termlists[i].termlist, value);
152 xmlDoc *normalize_record(struct session_database *sdb, Z_External *rec)
154 struct database_retrievalmap *m;
155 struct database *db = sdb->database;
157 const Odr_oid *oid = rec->direct_reference;
159 /* convert response record to XML somehow */
160 if (rec->which == Z_External_octet && oid
161 && !oid_oidcmp(oid, yaz_oid_recsyn_xml))
164 rdoc = xmlParseMemory((char*) rec->u.octet_aligned->buf,
165 rec->u.octet_aligned->len);
168 yaz_log(YLOG_FATAL, "Non-wellformed XML received from %s",
173 else if (oid && yaz_oid_is_iso2709(oid))
175 /* ISO2709 gets converted to MARCXML */
178 yaz_log(YLOG_FATAL, "Unable to handle ISO2709 record");
187 if (rec->which != Z_External_octet)
189 yaz_log(YLOG_WARN, "Unexpected external branch, probably BER %s",
193 buf = (char*) rec->u.octet_aligned->buf;
194 len = rec->u.octet_aligned->len;
195 if (yaz_marc_read_iso2709(sdb->yaz_marc, buf, len) < 0)
197 yaz_log(YLOG_WARN, "Failed to decode MARC %s", db->url);
201 yaz_marc_write_using_libxml2(sdb->yaz_marc, 1);
202 if (yaz_marc_write_xml(sdb->yaz_marc, &res,
203 "http://www.loc.gov/MARC21/slim", 0, 0) < 0)
205 yaz_log(YLOG_WARN, "Failed to encode as XML %s",
209 rdoc = xmlNewDoc((xmlChar *) "1.0");
210 xmlDocSetRootElement(rdoc, res);
215 char oid_name_buf[OID_STR_MAX];
216 const char *oid_name = yaz_oid_to_string_buf(oid, 0, oid_name_buf);
218 "Unable to handle record of type %s from %s",
223 if (global_parameters.dump_records){
225 "Input Record (normalized) from %s\n----------------\n",
227 #if LIBXML_VERSION >= 20600
228 xmlDocFormatDump(stderr, rdoc, 1);
230 xmlDocDump(stderr, rdoc);
234 for (m = sdb->map; m; m = m->next){
239 new = xsltApplyStylesheet(m->stylesheet, rdoc, 0);
240 root= xmlDocGetRootElement(new);
241 if (!new || !root || !(root->children))
243 yaz_log(YLOG_WARN, "XSLT transformation failed from %s",
254 if (global_parameters.dump_records)
256 fprintf(stderr, "Record from %s\n----------------\n",
258 #if LIBXML_VERSION >= 20600
259 xmlDocFormatDump(stderr, rdoc, 1);
261 xmlDocDump(stderr, rdoc);
267 // Retrieve first defined value for 'name' for given database.
268 // Will be extended to take into account user associated with session
269 char *session_setting_oneval(struct session_database *db, int offset)
271 if (!db->settings[offset])
273 return db->settings[offset]->value;
278 // Initialize YAZ Map structures for MARC-based targets
279 static int prepare_yazmarc(struct session_database *sdb)
285 yaz_log(YLOG_WARN, "No settings for %s", sdb->database->url);
288 if ((s = session_setting_oneval(sdb, PZ_NATIVESYNTAX))
289 && !strncmp(s, "iso2709", 7))
291 char *encoding = "marc-8s", *e;
294 // See if a native encoding is specified
295 if ((e = strchr(s, ';')))
298 sdb->yaz_marc = yaz_marc_create();
299 yaz_marc_subfield_str(sdb->yaz_marc, "\t");
301 cm = yaz_iconv_open("utf-8", encoding);
305 "Unable to map from %s to UTF-8 for target %s",
306 encoding, sdb->database->url);
309 yaz_marc_iconv(sdb->yaz_marc, cm);
314 // Prepare XSLT stylesheets for record normalization
315 // Structures are allocated on the session_wide nmem to avoid having
316 // to recompute this for every search. This would lead
317 // to leaking if a single session was to repeatedly change the PZ_XSLT
318 // setting. However, this is not a realistic use scenario.
319 static int prepare_map(struct session *se, struct session_database *sdb)
325 yaz_log(YLOG_WARN, "No settings on %s", sdb->database->url);
328 if ((s = session_setting_oneval(sdb, PZ_XSLT)))
331 struct database_retrievalmap **m = &sdb->map;
334 nmem_strsplit(se->session_nmem, ",", s, &stylesheets, &num);
335 for (i = 0; i < num; i++)
337 (*m) = nmem_malloc(se->session_nmem, sizeof(**m));
339 if (!((*m)->stylesheet = conf_load_stylesheet(stylesheets[i])))
341 yaz_log(YLOG_FATAL, "Unable to load stylesheet: %s",
349 yaz_log(YLOG_WARN, "No Normalization stylesheet for target %s",
354 // This analyzes settings and recomputes any supporting data structures
356 static int prepare_session_database(struct session *se,
357 struct session_database *sdb)
362 "No settings associated with %s", sdb->database->url);
365 if (sdb->settings[PZ_NATIVESYNTAX] && !sdb->yaz_marc)
367 if (prepare_yazmarc(sdb) < 0)
370 if (sdb->settings[PZ_XSLT] && !sdb->map)
372 if (prepare_map(se, sdb) < 0)
379 void session_set_watch(struct session *s, int what,
380 session_watchfun fun, void *data)
382 s->watchlist[what].fun = fun;
383 s->watchlist[what].data = data;
386 void session_alert_watch(struct session *s, int what)
388 if (!s->watchlist[what].fun)
390 (*s->watchlist[what].fun)(s->watchlist[what].data);
391 s->watchlist[what].fun = 0;
392 s->watchlist[what].data = 0;
395 //callback for grep_databases
396 static void select_targets_callback(void *context, struct session_database *db)
398 struct session *se = (struct session*) context;
399 struct client *cl = client_create();
400 client_set_database(cl, db);
401 client_set_session(cl, se);
404 // Associates a set of clients with a session;
405 // Note: Session-databases represent databases with per-session
407 int select_targets(struct session *se, struct database_criterion *crit)
410 client_destroy(se->clients);
412 return session_grep_databases(se, crit, select_targets_callback);
415 int session_active_clients(struct session *s)
420 for (c = s->clients; c; c = client_next_in_session(c))
421 if (client_is_active(c))
427 // parses crit1=val1,crit2=val2|val3,...
428 static struct database_criterion *parse_filter(NMEM m, const char *buf)
430 struct database_criterion *res = 0;
437 nmem_strsplit(m, ",", buf, &values, &num);
438 for (i = 0; i < num; i++)
443 struct database_criterion *new = nmem_malloc(m, sizeof(*new));
444 char *eq = strchr(values[i], '=');
447 yaz_log(YLOG_WARN, "Missing equal-sign in filter");
451 new->name = values[i];
452 nmem_strsplit(m, "|", eq, &subvalues, &subnum);
454 for (subi = 0; subi < subnum; subi++)
456 struct database_criterion_value *newv
457 = nmem_malloc(m, sizeof(*newv));
458 newv->value = subvalues[subi];
459 newv->next = new->values;
468 char *search(struct session *se, char *query, char *filter)
470 int live_channels = 0;
472 struct database_criterion *criteria;
474 yaz_log(YLOG_DEBUG, "Search");
476 nmem_reset(se->nmem);
477 criteria = parse_filter(se->nmem, filter);
479 live_channels = select_targets(se, criteria);
482 int maxrecs = live_channels * global_parameters.toget;
483 se->num_termlists = 0;
484 se->reclist = reclist_create(se->nmem, maxrecs);
485 // This will be initialized in send_search()
486 se->total_records = se->total_hits = se->total_merged = 0;
487 se->expected_maxrecs = maxrecs;
494 for (cl = se->clients; cl; cl = client_next_in_session(cl))
496 if (prepare_session_database(se, client_get_database(cl)) < 0)
497 return "CONFIG_ERROR";
498 // Query must parse for all targets
499 if (client_parse_query(cl, query) < 0)
503 for (cl = se->clients; cl; cl = client_next_in_session(cl))
504 client_prep_connection(cl);
509 // Creates a new session_database object for a database
510 static void session_init_databases_fun(void *context, struct database *db)
512 struct session *se = (struct session *) context;
513 struct session_database *new = nmem_malloc(se->session_nmem, sizeof(*new));
514 int num = settings_num();
521 if (global_parameters.server && global_parameters.server->icu_chn)
522 new->pct = pp2_charset_create(global_parameters.server->icu_chn);
524 new->pct = pp2_charset_create(0);
526 new->pct = pp2_charset_create(0);
531 = nmem_malloc(se->session_nmem, sizeof(struct settings *) * num);
532 memset(new->settings, 0, sizeof(struct settings*) * num);
536 for (i = 0; i < num; i++)
537 new->settings[i] = db->settings[i];
539 new->next = se->databases;
543 // Doesn't free memory associated with sdb -- nmem takes care of that
544 static void session_database_destroy(struct session_database *sdb)
546 struct database_retrievalmap *m;
548 for (m = sdb->map; m; m = m->next)
549 xsltFreeStylesheet(m->stylesheet);
551 yaz_marc_destroy(sdb->yaz_marc);
553 pp2_charset_destroy(sdb->pct);
556 // Initialize session_database list -- this represents this session's view
557 // of the database list -- subject to modification by the settings ws command
558 void session_init_databases(struct session *se)
561 grep_databases(se, 0, session_init_databases_fun);
564 // Probably session_init_databases_fun should be refactored instead of
566 static struct session_database *load_session_database(struct session *se,
569 struct database *db = find_database(id, 0);
571 session_init_databases_fun((void*) se, db);
572 // New sdb is head of se->databases list
573 return se->databases;
576 // Find an existing session database. If not found, load it
577 static struct session_database *find_session_database(struct session *se,
580 struct session_database *sdb;
582 for (sdb = se->databases; sdb; sdb = sdb->next)
583 if (!strcmp(sdb->database->url, id))
585 return load_session_database(se, id);
588 // Apply a session override to a database
589 void session_apply_setting(struct session *se, char *dbname, char *setting,
592 struct session_database *sdb = find_session_database(se, dbname);
593 struct setting *new = nmem_malloc(se->session_nmem, sizeof(*new));
594 int offset = settings_offset_cprefix(setting);
598 yaz_log(YLOG_WARN, "Unknown setting %s", setting);
601 // Jakub: This breaks the filter setting.
602 /*if (offset == PZ_ID)
604 yaz_log(YLOG_WARN, "No need to set pz:id setting. Ignoring");
608 new->target = dbname;
611 new->next = sdb->settings[offset];
612 sdb->settings[offset] = new;
614 // Force later recompute of settings-driven data structures
615 // (happens when a search starts and client connections are prepared)
618 case PZ_NATIVESYNTAX:
621 yaz_marc_destroy(sdb->yaz_marc);
628 struct database_retrievalmap *m;
629 // We don't worry about the map structure -- it's in nmem
630 for (m = sdb->map; m; m = m->next)
631 xsltFreeStylesheet(m->stylesheet);
638 void destroy_session(struct session *s)
640 struct session_database *sdb;
642 yaz_log(YLOG_LOG, "Destroying session");
644 client_destroy(s->clients);
645 for (sdb = s->databases; sdb; sdb = sdb->next)
646 session_database_destroy(sdb);
647 nmem_destroy(s->nmem);
648 wrbuf_destroy(s->wrbuf);
651 struct session *new_session(NMEM nmem)
654 struct session *session = nmem_malloc(nmem, sizeof(*session));
656 yaz_log(YLOG_DEBUG, "New Pazpar2 session");
658 session->total_hits = 0;
659 session->total_records = 0;
660 session->num_termlists = 0;
661 session->reclist = 0;
662 session->requestid = -1;
663 session->clients = 0;
664 session->expected_maxrecs = 0;
665 session->session_nmem = nmem;
666 session->nmem = nmem_create();
667 session->wrbuf = wrbuf_alloc();
668 session_init_databases(session);
669 for (i = 0; i <= SESSION_WATCH_MAX; i++)
671 session->watchlist[i].data = 0;
672 session->watchlist[i].fun = 0;
678 struct hitsbytarget *hitsbytarget(struct session *se, int *count)
680 static struct hitsbytarget res[1000]; // FIXME MM
684 for (cl = se->clients; cl; cl = client_next_in_session(cl))
686 char *name = session_setting_oneval(client_get_database(cl), PZ_NAME);
688 res[*count].id = client_get_database(cl)->database->url;
689 res[*count].name = *name ? name : "Unknown";
690 res[*count].hits = client_get_hits(cl);
691 res[*count].records = client_get_num_records(cl);
692 res[*count].diagnostic = client_get_diagnostic(cl);
693 res[*count].state = client_get_state_str(cl);
694 res[*count].connected = client_get_connection(cl) ? 1 : 0;
701 struct termlist_score **termlist(struct session *s, const char *name, int *num)
705 for (i = 0; i < s->num_termlists; i++)
706 if (!strcmp((const char *) s->termlists[i].name, name))
707 return termlist_highscore(s->termlists[i].termlist, num);
711 #ifdef MISSING_HEADERS
712 void report_nmem_stats(void)
714 size_t in_use, is_free;
716 nmem_get_memory_in_use(&in_use);
717 nmem_get_memory_free(&is_free);
719 yaz_log(YLOG_LOG, "nmem stat: use=%ld free=%ld",
720 (long) in_use, (long) is_free);
724 struct record_cluster *show_single(struct session *s, int id)
726 struct record_cluster *r;
728 reclist_rewind(s->reclist);
729 while ((r = reclist_read_record(s->reclist)))
735 struct record_cluster **show(struct session *s, struct reclist_sortparms *sp,
736 int start, int *num, int *total, int *sumhits,
739 struct record_cluster **recs = nmem_malloc(nmem_show, *num
740 * sizeof(struct record_cluster *));
741 struct reclist_sortparms *spp;
744 yaz_timing_t t = yaz_timing_create();
747 for (spp = sp; spp; spp = spp->next)
748 if (spp->type == Metadata_sortkey_relevance)
750 relevance_prepare_read(s->relevance, s->reclist);
753 reclist_sort(s->reclist, sp);
755 *total = s->reclist->num_records;
756 *sumhits = s->total_hits;
758 for (i = 0; i < start; i++)
759 if (!reclist_read_record(s->reclist))
766 for (i = 0; i < *num; i++)
768 struct record_cluster *r = reclist_read_record(s->reclist);
778 yaz_log(YLOG_LOG, "show %6.5f %3.2f %3.2f",
779 yaz_timing_get_real(t), yaz_timing_get_user(t),
780 yaz_timing_get_sys(t));
781 yaz_timing_destroy(&t);
786 void statistics(struct session *se, struct statistics *stat)
791 memset(stat, 0, sizeof(*stat));
792 for (cl = se->clients; cl; cl = client_next_in_session(cl))
794 if (!client_get_connection(cl))
795 stat->num_no_connection++;
796 switch (client_get_state(cl))
798 case Client_Connecting: stat->num_connecting++; break;
799 case Client_Initializing: stat->num_initializing++; break;
800 case Client_Searching: stat->num_searching++; break;
801 case Client_Presenting: stat->num_presenting++; break;
802 case Client_Idle: stat->num_idle++; break;
803 case Client_Failed: stat->num_failed++; break;
804 case Client_Error: stat->num_error++; break;
809 stat->num_hits = se->total_hits;
810 stat->num_records = se->total_records;
812 stat->num_clients = count;
815 void start_http_listener(void)
818 struct conf_server *ser = global_parameters.server;
820 if (*global_parameters.listener_override)
821 strcpy(hp, global_parameters.listener_override);
824 strcpy(hp, ser->host ? ser->host : "");
829 sprintf(hp + strlen(hp), "%d", ser->port);
835 void start_proxy(void)
838 struct conf_server *ser = global_parameters.server;
840 if (*global_parameters.proxy_override)
841 strcpy(hp, global_parameters.proxy_override);
842 else if (ser->proxy_host || ser->proxy_port)
844 strcpy(hp, ser->proxy_host ? ser->proxy_host : "");
849 sprintf(hp + strlen(hp), "%d", ser->proxy_port);
855 http_set_proxyaddr(hp, ser->myurl ? ser->myurl : "");
859 // Master list of connections we're handling events to
860 static IOCHAN channel_list = 0;
861 void pazpar2_add_channel(IOCHAN chan)
863 chan->next = channel_list;
867 void pazpar2_event_loop()
869 event_loop(&channel_list);
872 struct record *ingest_record(struct client *cl, Z_External *rec,
875 xmlDoc *xdoc = normalize_record(client_get_database(cl), rec);
877 struct record *record;
878 struct record_cluster *cluster;
879 struct session *se = client_get_session(cl);
880 xmlChar *mergekey, *mergekey_norm;
883 struct conf_service *service = global_parameters.server->service;
888 root = xmlDocGetRootElement(xdoc);
889 if (!(mergekey = xmlGetProp(root, (xmlChar *) "mergekey")))
891 yaz_log(YLOG_WARN, "No mergekey found in record");
896 record = record_create(se->nmem,
897 service->num_metadata, service->num_sortkeys, cl);
899 mergekey_norm = (xmlChar *) nmem_strdup(se->nmem, (char*) mergekey);
901 normalize7bit_mergekey((char *) mergekey_norm, 0);
903 cluster = reclist_insert(se->reclist,
904 global_parameters.server->service,
905 record, (char *) mergekey_norm,
907 if (global_parameters.dump_records)
908 yaz_log(YLOG_LOG, "Cluster id %d from %s (#%d)", cluster->recid,
909 client_get_database(cl)->database->url, record_no);
912 /* no room for record */
916 relevance_newrec(se->relevance, cluster);
919 // now parsing XML record and adding data to cluster or record metadata
920 for (n = root->children; n; n = n->next)
928 if (n->type != XML_ELEMENT_NODE)
930 if (!strcmp((const char *) n->name, "metadata"))
932 struct conf_metadata *ser_md = 0;
933 struct conf_sortkey *ser_sk = 0;
934 struct record_metadata **wheretoput = 0;
935 struct record_metadata *rec_md = 0;
936 int md_field_id = -1;
937 int sk_field_id = -1;
940 type = xmlGetProp(n, (xmlChar *) "type");
941 value = xmlNodeListGetString(xdoc, n->children, 0);
947 = conf_service_metadata_field_id(service, (const char *) type);
951 "Ignoring unknown metadata element: %s", type);
955 ser_md = &service->metadata[md_field_id];
957 if (ser_md->sortkey_offset >= 0){
958 sk_field_id = ser_md->sortkey_offset;
959 ser_sk = &service->sortkeys[sk_field_id];
962 // Find out where we are putting it - based on merge or not
963 if (ser_md->merge == Metadata_merge_no)
964 wheretoput = &record->metadata[md_field_id];
966 wheretoput = &cluster->metadata[md_field_id];
968 // create new record_metadata
969 rec_md = record_metadata_create(se->nmem);
971 // and polulate with data:
972 // type based charmapping decisions follow here
973 if (ser_md->type == Metadata_type_generic)
976 char * p = (char *) value;
977 p = normalize7bit_generic(p, " ,/.:([");
979 rec_md->data.text = nmem_strdup(se->nmem, p);
982 else if (ser_md->type == Metadata_type_year)
984 if (extract7bit_years((char *) value, &first, &last) < 0)
990 "Unknown type in metadata element %s", type);
994 // and polulate with data:
995 // assign cluster or record based on merge action
996 if (ser_md->merge == Metadata_merge_unique)
998 struct record_metadata *mnode;
999 for (mnode = *wheretoput; mnode; mnode = mnode->next)
1000 if (!strcmp((const char *) mnode->data.text,
1005 rec_md->next = *wheretoput;
1006 *wheretoput = rec_md;
1009 else if (ser_md->merge == Metadata_merge_longest)
1012 || strlen(rec_md->data.text)
1013 > strlen((*wheretoput)->data.text))
1015 *wheretoput = rec_md;
1018 char *s = nmem_strdup(se->nmem, rec_md->data.text);
1019 if (!cluster->sortkeys[sk_field_id])
1020 cluster->sortkeys[sk_field_id] =
1021 nmem_malloc(se->nmem,
1022 sizeof(union data_types));
1023 normalize7bit_mergekey(s,
1024 (ser_sk->type == Metadata_sortkey_skiparticle));
1025 cluster->sortkeys[sk_field_id]->text = s;
1029 else if (ser_md->merge == Metadata_merge_all
1030 || ser_md->merge == Metadata_merge_no)
1032 rec_md->next = *wheretoput;
1033 *wheretoput = rec_md;
1035 else if (ser_md->merge == Metadata_merge_range)
1039 *wheretoput = rec_md;
1040 (*wheretoput)->data.number.min = first;
1041 (*wheretoput)->data.number.max = last;
1043 cluster->sortkeys[sk_field_id]
1048 if (first < (*wheretoput)->data.number.min)
1049 (*wheretoput)->data.number.min = first;
1050 if (last > (*wheretoput)->data.number.max)
1051 (*wheretoput)->data.number.max = last;
1056 union data_types *sdata
1057 = cluster->sortkeys[sk_field_id];
1058 yaz_log(YLOG_LOG, "SK range: %d-%d",
1059 sdata->number.min, sdata->number.max);
1065 // ranking of _all_ fields enabled ...
1067 relevance_countwords(se->relevance, cluster,
1068 (char *) value, ser_md->rank);
1070 // construct facets ...
1071 if (ser_md->termlist)
1073 if (ser_md->type == Metadata_type_year)
1076 sprintf(year, "%d", last);
1077 add_facet(se, (char *) type, year);
1080 sprintf(year, "%d", first);
1081 add_facet(se, (char *) type, year);
1085 add_facet(se, (char *) type, (char *) value);
1095 "Unexpected element %s in internal record", n->name);
1104 relevance_donerecord(se->relevance, cluster);
1105 se->total_records++;
1115 * indent-tabs-mode: nil
1117 * vim: shiftwidth=4 tabstop=8 expandtab