1 /* $Id: logic.c,v 1.15 2007-04-20 16:37:35 quinn 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
22 // This file contains the primary business logic. Several parts of it should
23 // Eventually be factored into separate modules.
30 #include <sys/socket.h>
36 #include <yaz/marcdisp.h>
37 #include <yaz/comstack.h>
38 #include <yaz/tcpip.h>
39 #include <yaz/proto.h>
40 #include <yaz/readconf.h>
41 #include <yaz/pquery.h>
42 #include <yaz/otherinfo.h>
43 #include <yaz/yaz-util.h>
45 #include <yaz/query-charset.h>
46 #include <yaz/querytowrbuf.h>
47 #if YAZ_VERSIONL >= 0x020163
48 #include <yaz/oid_db.h>
57 #include <yaz/timing.h>
60 #include <netinet/in.h>
65 #include "termlists.h"
67 #include "relevance.h"
74 static void client_fatal(struct client *cl);
75 static void connection_destroy(struct connection *co);
76 static int client_prep_connection(struct client *cl);
77 static void ingest_records(struct client *cl, Z_Records *r);
78 void session_alert_watch(struct session *s, int what);
80 static struct connection *connection_freelist = 0;
81 static struct client *client_freelist = 0;
83 static char *client_states[] = {
87 "Client_Initializing",
92 "Client_Disconnected",
96 // Note: Some things in this structure will eventually move to configuration
97 struct parameters global_parameters =
107 "Index Data PazPar2",
117 static int send_apdu(struct client *c, Z_APDU *a)
119 struct connection *co = c->connection;
123 if (!z_APDU(global_parameters.odr_out, &a, 0, 0))
125 odr_perror(global_parameters.odr_out, "Encoding APDU");
128 buf = odr_getbuf(global_parameters.odr_out, &len, 0);
129 r = cs_put(co->link, buf, len);
132 yaz_log(YLOG_WARN, "cs_put: %s", cs_errmsg(cs_errno(co->link)));
137 fprintf(stderr, "cs_put incomplete (ParaZ does not handle that)\n");
140 odr_reset(global_parameters.odr_out); /* release the APDU structure */
141 co->state = Conn_Waiting;
145 // Set authentication token in init if one is set for the client
146 // TODO: Extend this to handle other schemes than open (should be simple)
147 static void init_authentication(struct client *cl, Z_InitRequest *req)
149 struct session_database *sdb = cl->database;
150 char *auth = session_setting_oneval(sdb, PZ_AUTHENTICATION);
154 Z_IdAuthentication *idAuth = odr_malloc(global_parameters.odr_out,
156 idAuth->which = Z_IdAuthentication_open;
157 idAuth->u.open = auth;
158 req->idAuthentication = idAuth;
162 static void send_init(IOCHAN i)
165 struct connection *co = iochan_getdata(i);
166 struct client *cl = co->client;
167 Z_APDU *a = zget_APDU(global_parameters.odr_out, Z_APDU_initRequest);
169 a->u.initRequest->implementationId = global_parameters.implementationId;
170 a->u.initRequest->implementationName = global_parameters.implementationName;
171 a->u.initRequest->implementationVersion =
172 global_parameters.implementationVersion;
173 ODR_MASK_SET(a->u.initRequest->options, Z_Options_search);
174 ODR_MASK_SET(a->u.initRequest->options, Z_Options_present);
175 ODR_MASK_SET(a->u.initRequest->options, Z_Options_namedResultSets);
177 ODR_MASK_SET(a->u.initRequest->protocolVersion, Z_ProtocolVersion_1);
178 ODR_MASK_SET(a->u.initRequest->protocolVersion, Z_ProtocolVersion_2);
179 ODR_MASK_SET(a->u.initRequest->protocolVersion, Z_ProtocolVersion_3);
181 init_authentication(cl, a->u.initRequest);
183 /* add virtual host if tunneling through Z39.50 proxy */
185 if (0 < strlen(global_parameters.zproxy_override)
186 && 0 < strlen(cl->database->database->url))
188 #if YAZ_VERSIONL >= 0x020163
189 yaz_oi_set_string_oid(&a->u.initRequest->otherInfo,
190 global_parameters.odr_out,
191 yaz_oid_userinfo_proxy,
192 1, cl->database->database->url);
194 yaz_oi_set_string_oidval(&a->u.initRequest->otherInfo,
195 global_parameters.odr_out, VAL_PROXY,
196 1, cl->database->database->url);
200 if (send_apdu(cl, a) >= 0)
202 iochan_setflags(i, EVENT_INPUT);
203 cl->state = Client_Initializing;
206 cl->state = Client_Error;
207 odr_reset(global_parameters.odr_out);
210 // Recursively traverse query structure to extract terms.
211 static void pull_terms(NMEM nmem, struct ccl_rpn_node *n, char **termlist, int *num)
223 pull_terms(nmem, n->u.p[0], termlist, num);
224 pull_terms(nmem, n->u.p[1], termlist, num);
227 nmem_strsplit(nmem, " ", n->u.t.term, &words, &numwords);
228 for (i = 0; i < numwords; i++)
229 termlist[(*num)++] = words[i];
236 // Extract terms from query into null-terminated termlist
237 static void extract_terms(NMEM nmem, struct ccl_rpn_node *query, char **termlist)
241 pull_terms(nmem, query, termlist, &num);
245 static void send_search(IOCHAN i)
247 struct connection *co = iochan_getdata(i);
248 struct client *cl = co->client;
249 struct session *se = cl->session;
250 struct session_database *sdb = cl->database;
251 Z_APDU *a = zget_APDU(global_parameters.odr_out, Z_APDU_searchRequest);
255 int ssub = 0, lslb = 100000, mspn = 10;
259 yaz_iconv_t iconv = 0;
261 yaz_log(YLOG_DEBUG, "Sending search to %s", cl->database->database->url);
263 // constructing RPN query
264 a->u.searchRequest->query = zquery = odr_malloc(global_parameters.odr_out,
266 zquery->which = Z_Query_type_1;
267 zquery->u.type_1 = p_query_rpn(global_parameters.odr_out, cl->pquery);
269 // converting to target encoding
270 if ((queryenc = session_setting_oneval(sdb, PZ_QUERYENCODING))){
271 iconv = yaz_iconv_open(queryenc, "UTF-8");
273 yaz_query_charset_convert_rpnquery(zquery->u.type_1,
274 global_parameters.odr_out,
276 yaz_iconv_close(iconv);
278 yaz_log(YLOG_WARN, "Query encoding failed %s %s",
279 cl->database->database->url, queryenc);
282 for (ndb = 0; sdb->database->databases[ndb]; ndb++)
284 databaselist = odr_malloc(global_parameters.odr_out, sizeof(char*) * ndb);
285 for (ndb = 0; sdb->database->databases[ndb]; ndb++)
286 databaselist[ndb] = sdb->database->databases[ndb];
288 if (!(piggyback = session_setting_oneval(sdb, PZ_PIGGYBACK)) || *piggyback == '1')
290 if ((recsyn = session_setting_oneval(sdb, PZ_REQUESTSYNTAX)))
292 #if YAZ_VERSIONL >= 0x020163
293 a->u.searchRequest->preferredRecordSyntax =
294 yaz_string_to_oid_odr(yaz_oid_std(),
295 CLASS_RECSYN, recsyn,
296 global_parameters.odr_out);
298 a->u.searchRequest->preferredRecordSyntax =
299 yaz_str_to_z3950oid(global_parameters.odr_out,
300 CLASS_RECSYN, recsyn);
303 a->u.searchRequest->smallSetUpperBound = &ssub;
304 a->u.searchRequest->largeSetLowerBound = &lslb;
305 a->u.searchRequest->mediumSetPresentNumber = &mspn;
307 a->u.searchRequest->resultSetName = "Default";
308 a->u.searchRequest->databaseNames = databaselist;
309 a->u.searchRequest->num_databaseNames = ndb;
312 { //scope for sending and logging queries
313 WRBUF wbquery = wrbuf_alloc();
314 yaz_query_to_wrbuf(wbquery, zquery);
317 if (send_apdu(cl, a) >= 0)
319 iochan_setflags(i, EVENT_INPUT);
320 cl->state = Client_Searching;
321 cl->requestid = se->requestid;
322 yaz_log(YLOG_LOG, "SearchRequest %s %s %s",
323 cl->database->database->url,
324 queryenc ? queryenc : "UTF-8",
325 wrbuf_cstr(wbquery));
328 cl->state = Client_Error;
329 yaz_log(YLOG_WARN, "Failed SearchRequest %s %s %s",
330 cl->database->database->url,
331 queryenc ? queryenc : "UTF-8",
332 wrbuf_cstr(wbquery));
335 wrbuf_destroy(wbquery);
338 odr_reset(global_parameters.odr_out);
341 static void send_present(IOCHAN i)
343 struct connection *co = iochan_getdata(i);
344 struct client *cl = co->client;
345 struct session_database *sdb = cl->database;
346 Z_APDU *a = zget_APDU(global_parameters.odr_out, Z_APDU_presentRequest);
348 int start = cl->records + 1;
351 toget = global_parameters.chunk;
352 if (toget > global_parameters.toget - cl->records)
353 toget = global_parameters.toget - cl->records;
354 if (toget > cl->hits - cl->records)
355 toget = cl->hits - cl->records;
357 yaz_log(YLOG_DEBUG, "Trying to present %d records\n", toget);
359 a->u.presentRequest->resultSetStartPoint = &start;
360 a->u.presentRequest->numberOfRecordsRequested = &toget;
362 a->u.presentRequest->resultSetId = "Default";
364 if ((recsyn = session_setting_oneval(sdb, PZ_REQUESTSYNTAX)))
366 #if YAZ_VERSIONL >= 0x020163
367 a->u.presentRequest->preferredRecordSyntax =
368 yaz_string_to_oid_odr(yaz_oid_std(),
369 CLASS_RECSYN, recsyn,
370 global_parameters.odr_out);
372 a->u.presentRequest->preferredRecordSyntax =
373 yaz_str_to_z3950oid(global_parameters.odr_out,
374 CLASS_RECSYN, recsyn);
378 if (send_apdu(cl, a) >= 0)
380 iochan_setflags(i, EVENT_INPUT);
381 cl->state = Client_Presenting;
384 cl->state = Client_Error;
385 odr_reset(global_parameters.odr_out);
388 static void do_initResponse(IOCHAN i, Z_APDU *a)
390 struct connection *co = iochan_getdata(i);
391 struct client *cl = co->client;
392 Z_InitResponse *r = a->u.initResponse;
394 yaz_log(YLOG_DEBUG, "Init response %s", cl->database->database->url);
398 cl->state = Client_Idle;
401 cl->state = Client_Failed; // FIXME need to do something to the connection
404 static void do_searchResponse(IOCHAN i, Z_APDU *a)
406 struct connection *co = iochan_getdata(i);
407 struct client *cl = co->client;
408 struct session *se = cl->session;
409 Z_SearchResponse *r = a->u.searchResponse;
411 yaz_log(YLOG_DEBUG, "Search response %s (status=%d)",
412 cl->database->database->url, *r->searchStatus);
414 if (*r->searchStatus)
416 cl->hits = *r->resultCount;
417 se->total_hits += cl->hits;
418 if (r->presentStatus && !*r->presentStatus && r->records)
420 yaz_log(YLOG_DEBUG, "Records in search response %s",
421 cl->database->database->url);
422 ingest_records(cl, r->records);
424 cl->state = Client_Idle;
429 cl->state = Client_Error;
431 Z_Records *recs = r->records;
432 if (recs->which == Z_Records_NSD)
435 "Search response: Non-surrogate diagnostic %s",
436 cl->database->database->url);
437 cl->diagnostic = *recs->u.nonSurrogateDiagnostic->condition;
438 cl->state = Client_Error;
444 static void do_closeResponse(IOCHAN i, Z_APDU *a)
446 struct connection *co = iochan_getdata(i);
447 struct client *cl = co->client;
448 /* Z_Close *r = a->u.close; */
450 yaz_log(YLOG_WARN, "Close response %s", cl->database->database->url);
452 cl->state = Client_Failed;
453 connection_destroy(co);
457 char *normalize_mergekey(char *buf, int skiparticle)
459 char *p = buf, *pout = buf;
464 char articles[] = "the den der die des an a "; // must end in space
466 while (*p && !isalnum(*p))
469 while (*p && *p != ' ' && pout - firstword < 62)
470 *(pout++) = tolower(*(p++));
473 if (!strstr(articles, firstword))
480 while (*p && !isalnum(*p))
483 *(pout++) = tolower(*(p++));
486 while (*p && !isalnum(*p))
493 while (pout > buf && *pout == ' ');
498 static void add_facet(struct session *s, const char *type, const char *value)
504 for (i = 0; i < s->num_termlists; i++)
505 if (!strcmp(s->termlists[i].name, type))
507 if (i == s->num_termlists)
509 if (i == SESSION_MAX_TERMLISTS)
511 yaz_log(YLOG_FATAL, "Too many termlists");
514 s->termlists[i].name = nmem_strdup(s->nmem, type);
515 s->termlists[i].termlist = termlist_create(s->nmem, s->expected_maxrecs, 15);
516 s->num_termlists = i + 1;
518 termlist_insert(s->termlists[i].termlist, value);
521 static xmlDoc *normalize_record(struct client *cl, Z_External *rec)
523 struct database_retrievalmap *m;
524 struct session_database *sdb = cl->database;
525 struct database *db = sdb->database;
529 // First normalize to XML
534 if (rec->which != Z_External_octet)
536 yaz_log(YLOG_WARN, "Unexpected external branch, probably BER %s",
540 buf = (char*) rec->u.octet_aligned->buf;
541 len = rec->u.octet_aligned->len;
542 if (yaz_marc_read_iso2709(sdb->yaz_marc, buf, len) < 0)
544 yaz_log(YLOG_WARN, "Failed to decode MARC %s", db->url);
548 yaz_marc_write_using_libxml2(sdb->yaz_marc, 1);
549 if (yaz_marc_write_xml(sdb->yaz_marc, &res,
550 "http://www.loc.gov/MARC21/slim", 0, 0) < 0)
552 yaz_log(YLOG_WARN, "Failed to encode as XML %s",
556 rdoc = xmlNewDoc((xmlChar *) "1.0");
557 xmlDocSetRootElement(rdoc, res);
563 "Unknown native_syntax in normalize_record from %s",
568 if (global_parameters.dump_records){
570 "Input Record (normalized) from %s\n----------------\n",
572 #if LIBXML_VERSION >= 20600
573 xmlDocFormatDump(stderr, rdoc, 1);
575 xmlDocDump(stderr, rdoc);
579 for (m = sdb->map; m; m = m->next){
585 new = xsltApplyStylesheet(m->stylesheet, rdoc, 0);
586 root= xmlDocGetRootElement(new);
587 if (!new || !root || !(root->children))
589 yaz_log(YLOG_WARN, "XSLT transformation failed from %s",
590 cl->database->database->url);
599 // do it another way to detect transformation errors right now
600 // but does not seem to work either!
602 xsltTransformContextPtr ctxt;
603 ctxt = xsltNewTransformContext(m->stylesheet, rdoc);
604 new = xsltApplyStylesheetUser(m->stylesheet, rdoc, 0, 0, 0, ctxt);
605 if ((ctxt->state == XSLT_STATE_ERROR) ||
606 (ctxt->state == XSLT_STATE_STOPPED)){
607 yaz_log(YLOG_WARN, "XSLT transformation failed from %s",
608 cl->database->database->url);
619 if (global_parameters.dump_records)
621 fprintf(stderr, "Record from %s\n----------------\n",
622 cl->database->database->url);
623 #if LIBXML_VERSION >= 20600
624 xmlDocFormatDump(stderr, rdoc, 1);
626 xmlDocDump(stderr, rdoc);
632 // Extract what appears to be years from buf, storing highest and
634 static int extract_years(const char *buf, int *first, int *last)
643 while (*buf && !isdigit(*buf))
646 for (e = buf; *e && isdigit(*e); e++)
650 int value = atoi(buf);
651 if (*first < 0 || value < *first)
653 if (*last < 0 || value > *last)
661 static struct record *ingest_record(struct client *cl, Z_External *rec)
663 xmlDoc *xdoc = normalize_record(cl, rec);
666 struct record_cluster *cluster;
667 struct session *se = cl->session;
668 xmlChar *mergekey, *mergekey_norm;
671 struct conf_service *service = global_parameters.server->service;
676 root = xmlDocGetRootElement(xdoc);
677 if (!(mergekey = xmlGetProp(root, (xmlChar *) "mergekey")))
679 yaz_log(YLOG_WARN, "No mergekey found in record");
684 res = nmem_malloc(se->nmem, sizeof(struct record));
687 res->metadata = nmem_malloc(se->nmem,
688 sizeof(struct record_metadata*) * service->num_metadata);
689 memset(res->metadata, 0, sizeof(struct record_metadata*) * service->num_metadata);
691 mergekey_norm = (xmlChar *) nmem_strdup(se->nmem, (char*) mergekey);
693 normalize_mergekey((char *) mergekey_norm, 0);
695 cluster = reclist_insert(se->reclist,
696 global_parameters.server->service,
697 res, (char *) mergekey_norm,
699 if (global_parameters.dump_records)
700 yaz_log(YLOG_LOG, "Cluster id %d from %s (#%d)", cluster->recid,
701 cl->database->database->url, cl->records);
704 /* no room for record */
708 relevance_newrec(se->relevance, cluster);
710 for (n = root->children; n; n = n->next)
718 if (n->type != XML_ELEMENT_NODE)
720 if (!strcmp((const char *) n->name, "metadata"))
722 struct conf_metadata *md = 0;
723 struct conf_sortkey *sk = 0;
724 struct record_metadata **wheretoput, *newm;
728 type = xmlGetProp(n, (xmlChar *) "type");
729 value = xmlNodeListGetString(xdoc, n->children, 0);
734 // First, find out what field we're looking at
735 for (imeta = 0; imeta < service->num_metadata; imeta++)
736 if (!strcmp((const char *) type, service->metadata[imeta].name))
738 md = &service->metadata[imeta];
739 if (md->sortkey_offset >= 0)
740 sk = &service->sortkeys[md->sortkey_offset];
745 yaz_log(YLOG_WARN, "Ignoring unknown metadata element: %s", type);
749 // Find out where we are putting it
750 if (md->merge == Metadata_merge_no)
751 wheretoput = &res->metadata[imeta];
753 wheretoput = &cluster->metadata[imeta];
756 newm = nmem_malloc(se->nmem, sizeof(struct record_metadata));
758 if (md->type == Metadata_type_generic)
761 for (p = (char *) value; *p && isspace(*p); p++)
763 for (pe = p + strlen(p) - 1;
764 pe > p && strchr(" ,/.:([", *pe); pe--)
766 newm->data.text = nmem_strdup(se->nmem, p);
769 else if (md->type == Metadata_type_year)
771 if (extract_years((char *) value, &first, &last) < 0)
776 yaz_log(YLOG_WARN, "Unknown type in metadata element %s", type);
779 if (md->type == Metadata_type_year && md->merge != Metadata_merge_range)
781 yaz_log(YLOG_WARN, "Only range merging supported for years");
784 if (md->merge == Metadata_merge_unique)
786 struct record_metadata *mnode;
787 for (mnode = *wheretoput; mnode; mnode = mnode->next)
788 if (!strcmp((const char *) mnode->data.text, newm->data.text))
792 newm->next = *wheretoput;
796 else if (md->merge == Metadata_merge_longest)
799 strlen(newm->data.text) > strlen((*wheretoput)->data.text))
804 char *s = nmem_strdup(se->nmem, newm->data.text);
805 if (!cluster->sortkeys[md->sortkey_offset])
806 cluster->sortkeys[md->sortkey_offset] =
807 nmem_malloc(se->nmem, sizeof(union data_types));
808 normalize_mergekey(s,
809 (sk->type == Metadata_sortkey_skiparticle));
810 cluster->sortkeys[md->sortkey_offset]->text = s;
814 else if (md->merge == Metadata_merge_all || md->merge == Metadata_merge_no)
816 newm->next = *wheretoput;
819 else if (md->merge == Metadata_merge_range)
821 assert(md->type == Metadata_type_year);
825 (*wheretoput)->data.number.min = first;
826 (*wheretoput)->data.number.max = last;
828 cluster->sortkeys[md->sortkey_offset] = &newm->data;
832 if (first < (*wheretoput)->data.number.min)
833 (*wheretoput)->data.number.min = first;
834 if (last > (*wheretoput)->data.number.max)
835 (*wheretoput)->data.number.max = last;
840 union data_types *sdata = cluster->sortkeys[md->sortkey_offset];
841 yaz_log(YLOG_LOG, "SK range: %d-%d", sdata->number.min, sdata->number.max);
846 yaz_log(YLOG_WARN, "Don't know how to merge on element name %s", md->name);
849 relevance_countwords(se->relevance, cluster,
850 (char *) value, md->rank);
853 if (md->type == Metadata_type_year)
856 sprintf(year, "%d", last);
857 add_facet(se, (char *) type, year);
860 sprintf(year, "%d", first);
861 add_facet(se, (char *) type, year);
865 add_facet(se, (char *) type, (char *) value);
872 yaz_log(YLOG_WARN, "Unexpected element %s in internal record", n->name);
881 relevance_donerecord(se->relevance, cluster);
887 // Retrieve first defined value for 'name' for given database.
888 // Will be extended to take into account user associated with session
889 char *session_setting_oneval(struct session_database *db, int offset)
891 if (!db->settings[offset])
893 return db->settings[offset]->value;
896 static void ingest_records(struct client *cl, Z_Records *r)
899 yaz_timing_t t = yaz_timing_create();
902 struct session *s = cl->session;
903 Z_NamePlusRecordList *rlist;
906 if (r->which != Z_Records_DBOSD)
908 rlist = r->u.databaseOrSurDiagnostics;
909 for (i = 0; i < rlist->num_records; i++)
911 Z_NamePlusRecord *npr = rlist->records[i];
914 if (npr->which != Z_NamePlusRecord_databaseRecord)
917 "Unexpected record type, probably diagnostic %s",
918 cl->database->database->url);
922 rec = ingest_record(cl, npr->u.databaseRecord);
926 if (s->watchlist[SESSION_WATCH_RECORDS].fun && rlist->num_records)
927 session_alert_watch(s, SESSION_WATCH_RECORDS);
931 yaz_log(YLOG_LOG, "ingest_records %6.5f %3.2f %3.2f",
932 yaz_timing_get_real(t), yaz_timing_get_user(t),
933 yaz_timing_get_sys(t));
934 yaz_timing_destroy(&t);
938 static void do_presentResponse(IOCHAN i, Z_APDU *a)
940 struct connection *co = iochan_getdata(i);
941 struct client *cl = co->client;
942 Z_PresentResponse *r = a->u.presentResponse;
945 Z_Records *recs = r->records;
946 if (recs->which == Z_Records_NSD)
948 yaz_log(YLOG_WARN, "Non-surrogate diagnostic %s",
949 cl->database->database->url);
950 cl->diagnostic = *recs->u.nonSurrogateDiagnostic->condition;
951 cl->state = Client_Error;
955 if (!*r->presentStatus && cl->state != Client_Error)
957 yaz_log(YLOG_DEBUG, "Good Present response %s",
958 cl->database->database->url);
959 ingest_records(cl, r->records);
960 cl->state = Client_Idle;
962 else if (*r->presentStatus)
964 yaz_log(YLOG_WARN, "Bad Present response %s",
965 cl->database->database->url);
966 cl->state = Client_Error;
970 static void handler(IOCHAN i, int event)
972 struct connection *co = iochan_getdata(i);
973 struct client *cl = co->client;
974 struct session *se = 0;
980 yaz_log(YLOG_WARN, "Destroying orphan connection");
981 connection_destroy(co);
985 if (co->state == Conn_Connecting && event & EVENT_OUTPUT)
988 socklen_t errlen = sizeof(errcode);
990 if (getsockopt(cs_fileno(co->link), SOL_SOCKET, SO_ERROR, &errcode,
991 &errlen) < 0 || errcode != 0)
998 yaz_log(YLOG_DEBUG, "Connect OK");
999 co->state = Conn_Open;
1001 cl->state = Client_Connected;
1005 else if (event & EVENT_INPUT)
1007 int len = cs_get(co->link, &co->ibuf, &co->ibufsize);
1011 yaz_log(YLOG_WARN|YLOG_ERRNO, "Error reading from %s",
1012 cl->database->database->url);
1013 connection_destroy(co);
1018 yaz_log(YLOG_WARN, "EOF reading from %s", cl->database->database->url);
1019 connection_destroy(co);
1022 else if (len > 1) // We discard input if we have no connection
1024 co->state = Conn_Open;
1026 if (cl && (cl->requestid == se->requestid || cl->state == Client_Initializing))
1030 odr_reset(global_parameters.odr_in);
1031 odr_setbuf(global_parameters.odr_in, co->ibuf, len, 0);
1032 if (!z_APDU(global_parameters.odr_in, &a, 0, 0))
1039 case Z_APDU_initResponse:
1040 do_initResponse(i, a);
1042 case Z_APDU_searchResponse:
1043 do_searchResponse(i, a);
1045 case Z_APDU_presentResponse:
1046 do_presentResponse(i, a);
1049 do_closeResponse(i, a);
1053 "Unexpected Z39.50 response from %s",
1054 cl->database->database->url);
1058 // We aren't expecting staggered output from target
1059 // if (cs_more(t->link))
1060 // iochan_setevent(i, EVENT_INPUT);
1062 else // we throw away response and go to idle mode
1064 yaz_log(YLOG_DEBUG, "Ignoring result of expired operation");
1065 cl->state = Client_Idle;
1068 /* if len==1 we do nothing but wait for more input */
1071 if (cl->state == Client_Connected) {
1075 if (cl->state == Client_Idle)
1077 if (cl->requestid != se->requestid && cl->pquery) {
1080 else if (cl->hits > 0 && cl->records < global_parameters.toget &&
1081 cl->records < cl->hits) {
1087 // Disassociate connection from client
1088 static void connection_release(struct connection *co)
1090 struct client *cl = co->client;
1092 yaz_log(YLOG_DEBUG, "Connection release %s", co->host->hostport);
1099 // Close connection and recycle structure
1100 static void connection_destroy(struct connection *co)
1102 struct host *h = co->host;
1104 iochan_destroy(co->iochan);
1106 yaz_log(YLOG_DEBUG, "Connection destroy %s", co->host->hostport);
1107 if (h->connections == co)
1108 h->connections = co->next;
1111 struct connection *pco;
1112 for (pco = h->connections; pco && pco->next != co; pco = pco->next)
1115 pco->next = co->next;
1121 if (co->client->state != Client_Idle)
1122 co->client->state = Client_Disconnected;
1123 co->client->connection = 0;
1125 co->next = connection_freelist;
1126 connection_freelist = co;
1129 // Creates a new connection for client, associated with the host of
1130 // client's database
1131 static struct connection *connection_create(struct client *cl)
1133 struct connection *new;
1139 if (!(link = cs_create(tcpip_type, 0, PROTO_Z3950)))
1141 yaz_log(YLOG_FATAL|YLOG_ERRNO, "Failed to create comstack");
1145 if (0 == strlen(global_parameters.zproxy_override)){
1146 /* no Z39.50 proxy needed - direct connect */
1147 yaz_log(YLOG_DEBUG, "Connection create %s", cl->database->database->url);
1149 if (!(addr = cs_straddr(link, cl->database->database->host->ipport)))
1151 yaz_log(YLOG_WARN|YLOG_ERRNO,
1152 "Lookup of IP address %s failed",
1153 cl->database->database->host->ipport);
1158 /* Z39.50 proxy connect */
1159 yaz_log(YLOG_DEBUG, "Connection create %s proxy %s",
1160 cl->database->database->url, global_parameters.zproxy_override);
1162 if (!(addr = cs_straddr(link, global_parameters.zproxy_override)))
1164 yaz_log(YLOG_WARN|YLOG_ERRNO,
1165 "Lookup of IP address %s failed",
1166 global_parameters.zproxy_override);
1171 res = cs_connect(link, addr);
1174 yaz_log(YLOG_WARN|YLOG_ERRNO, "cs_connect %s", cl->database->database->url);
1178 if ((new = connection_freelist))
1179 connection_freelist = new->next;
1182 new = xmalloc(sizeof (struct connection));
1186 new->state = Conn_Connecting;
1187 new->host = cl->database->database->host;
1188 new->next = new->host->connections;
1189 new->host->connections = new;
1191 cl->connection = new;
1194 new->iochan = iochan_create(cs_fileno(link), handler, 0);
1195 iochan_setdata(new->iochan, new);
1196 pazpar2_add_channel(new->iochan);
1200 // Close connection and set state to error
1201 static void client_fatal(struct client *cl)
1203 yaz_log(YLOG_WARN, "Fatal error from %s", cl->database->database->url);
1204 connection_destroy(cl->connection);
1205 cl->state = Client_Error;
1208 // Ensure that client has a connection associated
1209 static int client_prep_connection(struct client *cl)
1211 struct connection *co;
1212 struct session *se = cl->session;
1213 struct host *host = cl->database->database->host;
1215 co = cl->connection;
1217 yaz_log(YLOG_DEBUG, "Client prep %s", cl->database->database->url);
1221 // See if someone else has an idle connection
1222 // We should look at timestamps here to select the longest-idle connection
1223 for (co = host->connections; co; co = co->next)
1224 if (co->state == Conn_Open && (!co->client || co->client->session != se))
1228 connection_release(co);
1229 cl->connection = co;
1233 co = connection_create(cl);
1237 if (co->state == Conn_Connecting)
1239 cl->state = Client_Connecting;
1240 iochan_setflag(co->iochan, EVENT_OUTPUT);
1242 else if (co->state == Conn_Open)
1244 if (cl->state == Client_Error || cl->state == Client_Disconnected)
1245 cl->state = Client_Idle;
1246 iochan_setflag(co->iochan, EVENT_OUTPUT);
1254 // Initialize YAZ Map structures for MARC-based targets
1255 static int prepare_yazmarc(struct session_database *sdb)
1261 yaz_log(YLOG_WARN, "No settings for %s", sdb->database->url);
1264 if ((s = session_setting_oneval(sdb, PZ_NATIVESYNTAX)) && !strncmp(s, "iso2709", 7))
1266 char *encoding = "marc-8s", *e;
1269 // See if a native encoding is specified
1270 if ((e = strchr(s, ';')))
1273 sdb->yaz_marc = yaz_marc_create();
1274 yaz_marc_subfield_str(sdb->yaz_marc, "\t");
1276 cm = yaz_iconv_open("utf-8", encoding);
1280 "Unable to map from %s to UTF-8 for target %s",
1281 encoding, sdb->database->url);
1284 yaz_marc_iconv(sdb->yaz_marc, cm);
1289 // Prepare XSLT stylesheets for record normalization
1290 // Structures are allocated on the session_wide nmem to avoid having
1291 // to recompute this for every search. This would lead
1292 // to leaking if a single session was to repeatedly change the PZ_XSLT
1293 // setting. However, this is not a realistic use scenario.
1294 static int prepare_map(struct session *se, struct session_database *sdb)
1300 yaz_log(YLOG_WARN, "No settings on %s", sdb->database->url);
1303 if ((s = session_setting_oneval(sdb, PZ_XSLT)))
1306 struct database_retrievalmap **m = &sdb->map;
1309 nmem_strsplit(se->session_nmem, ",", s, &stylesheets, &num);
1310 for (i = 0; i < num; i++)
1312 (*m) = nmem_malloc(se->session_nmem, sizeof(**m));
1314 if (!((*m)->stylesheet = conf_load_stylesheet(stylesheets[i])))
1316 yaz_log(YLOG_FATAL, "Unable to load stylesheet: %s",
1324 yaz_log(YLOG_WARN, "No Normalization stylesheet for target %s",
1325 sdb->database->url);
1329 // This analyzes settings and recomputes any supporting data structures
1331 static int prepare_session_database(struct session *se, struct session_database *sdb)
1335 yaz_log(YLOG_WARN, "No settings associated with %s", sdb->database->url);
1338 if (sdb->settings[PZ_NATIVESYNTAX] && !sdb->yaz_marc)
1340 if (prepare_yazmarc(sdb) < 0)
1343 if (sdb->settings[PZ_XSLT] && !sdb->map)
1345 if (prepare_map(se, sdb) < 0)
1351 // Initialize CCL map for a target
1352 static CCL_bibset prepare_cclmap(struct client *cl)
1354 struct session_database *sdb = cl->database;
1360 res = ccl_qual_mk();
1361 for (s = sdb->settings[PZ_CCLMAP]; s; s = s->next)
1363 char *p = strchr(s->name + 3, ':');
1366 yaz_log(YLOG_WARN, "Malformed cclmap name: %s", s->name);
1371 ccl_qual_fitem(res, s->value, p);
1376 // Parse the query given the settings specific to this client
1377 static int client_parse_query(struct client *cl, const char *query)
1379 struct session *se = cl->session;
1380 struct ccl_rpn_node *cn;
1382 CCL_bibset ccl_map = prepare_cclmap(cl);
1386 cn = ccl_find_str(ccl_map, query, &cerror, &cpos);
1387 ccl_qual_rm(&ccl_map);
1390 cl->state = Client_Error;
1391 yaz_log(YLOG_WARN, "Failed to parse query for %s",
1392 cl->database->database->url);
1395 wrbuf_rewind(se->wrbuf);
1396 ccl_pquery(se->wrbuf, cn);
1397 wrbuf_putc(se->wrbuf, '\0');
1400 cl->pquery = xstrdup(wrbuf_buf(se->wrbuf));
1404 // Initialize relevance structure with query terms
1406 extract_terms(se->nmem, cn, p);
1407 se->relevance = relevance_create(se->nmem, (const char **) p,
1408 se->expected_maxrecs);
1415 static struct client *client_create(void)
1418 if (client_freelist)
1420 r = client_freelist;
1421 client_freelist = client_freelist->next;
1424 r = xmalloc(sizeof(struct client));
1434 r->state = Client_Disconnected;
1439 void client_destroy(struct client *c)
1441 struct session *se = c->session;
1442 if (c == se->clients)
1443 se->clients = c->next;
1447 for (cc = se->clients; cc && cc->next != c; cc = cc->next)
1453 connection_release(c->connection);
1454 c->next = client_freelist;
1455 client_freelist = c;
1458 void session_set_watch(struct session *s, int what, session_watchfun fun, void *data)
1460 s->watchlist[what].fun = fun;
1461 s->watchlist[what].data = data;
1464 void session_alert_watch(struct session *s, int what)
1466 if (!s->watchlist[what].fun)
1468 (*s->watchlist[what].fun)(s->watchlist[what].data);
1469 s->watchlist[what].fun = 0;
1470 s->watchlist[what].data = 0;
1473 //callback for grep_databases
1474 static void select_targets_callback(void *context, struct session_database *db)
1476 struct session *se = (struct session*) context;
1477 struct client *cl = client_create();
1480 cl->next = se->clients;
1484 // Associates a set of clients with a session;
1485 // Note: Session-databases represent databases with per-session setting overrides
1486 int select_targets(struct session *se, struct database_criterion *crit)
1489 client_destroy(se->clients);
1491 return session_grep_databases(se, crit, select_targets_callback);
1494 int session_active_clients(struct session *s)
1499 for (c = s->clients; c; c = c->next)
1500 if (c->connection && (c->state == Client_Connecting ||
1501 c->state == Client_Initializing ||
1502 c->state == Client_Searching ||
1503 c->state == Client_Presenting))
1509 // parses crit1=val1,crit2=val2|val3,...
1510 static struct database_criterion *parse_filter(NMEM m, const char *buf)
1512 struct database_criterion *res = 0;
1519 nmem_strsplit(m, ",", buf, &values, &num);
1520 for (i = 0; i < num; i++)
1525 struct database_criterion *new = nmem_malloc(m, sizeof(*new));
1526 char *eq = strchr(values[i], '=');
1529 yaz_log(YLOG_WARN, "Missing equal-sign in filter");
1533 new->name = values[i];
1534 nmem_strsplit(m, "|", eq, &subvalues, &subnum);
1536 for (subi = 0; subi < subnum; subi++)
1538 struct database_criterion_value *newv = nmem_malloc(m, sizeof(*newv));
1539 newv->value = subvalues[subi];
1540 newv->next = new->values;
1549 char *search(struct session *se, char *query, char *filter)
1551 int live_channels = 0;
1553 struct database_criterion *criteria;
1555 yaz_log(YLOG_DEBUG, "Search");
1557 nmem_reset(se->nmem);
1558 criteria = parse_filter(se->nmem, filter);
1560 live_channels = select_targets(se, criteria);
1563 int maxrecs = live_channels * global_parameters.toget;
1564 se->num_termlists = 0;
1565 se->reclist = reclist_create(se->nmem, maxrecs);
1566 // This will be initialized in send_search()
1567 se->total_records = se->total_hits = se->total_merged = 0;
1568 se->expected_maxrecs = maxrecs;
1575 for (cl = se->clients; cl; cl = cl->next)
1577 if (prepare_session_database(se, cl->database) < 0)
1578 return "CONFIG_ERROR";
1579 if (client_parse_query(cl, query) < 0) // Query must parse for all targets
1583 for (cl = se->clients; cl; cl = cl->next)
1585 client_prep_connection(cl);
1591 // Apply a session override to a database
1592 void session_apply_setting(struct session *se, char *dbname, char *setting, char *value)
1594 struct session_database *sdb;
1596 for (sdb = se->databases; sdb; sdb = sdb->next)
1597 if (!strcmp(dbname, sdb->database->url))
1599 struct setting *new = nmem_malloc(se->session_nmem, sizeof(*new));
1600 int offset = settings_offset(setting);
1604 yaz_log(YLOG_WARN, "Unknown setting %s", setting);
1607 new->precedence = 0;
1608 new->target = dbname;
1609 new->name = setting;
1611 new->next = sdb->settings[offset];
1612 sdb->settings[offset] = new;
1614 // Force later recompute of settings-driven data structures
1615 // (happens when a search starts and client connections are prepared)
1618 case PZ_NATIVESYNTAX:
1621 yaz_marc_destroy(sdb->yaz_marc);
1628 struct database_retrievalmap *m;
1629 // We don't worry about the map structure -- it's in nmem
1630 for (m = sdb->map; m; m = m->next)
1631 xsltFreeStylesheet(m->stylesheet);
1639 yaz_log(YLOG_WARN, "Unknown database in setting override: %s", dbname);
1642 void session_init_databases_fun(void *context, struct database *db)
1644 struct session *se = (struct session *) context;
1645 struct session_database *new = nmem_malloc(se->session_nmem, sizeof(*new));
1646 int num = settings_num();
1652 new->settings = nmem_malloc(se->session_nmem, sizeof(struct settings *) * num);
1653 for (i = 0; i < num; i++)
1654 new->settings[i] = db->settings[i];
1655 new->next = se->databases;
1656 se->databases = new;
1659 // Initialize session_database list -- this represents this session's view
1660 // of the database list -- subject to modification by the settings ws command
1661 void session_init_databases(struct session *se)
1664 grep_databases(se, 0, session_init_databases_fun);
1667 void destroy_session(struct session *s)
1669 yaz_log(YLOG_LOG, "Destroying session");
1671 client_destroy(s->clients);
1672 nmem_destroy(s->nmem);
1673 wrbuf_destroy(s->wrbuf);
1676 struct session *new_session(NMEM nmem)
1679 struct session *session = nmem_malloc(nmem, sizeof(*session));
1681 yaz_log(YLOG_DEBUG, "New Pazpar2 session");
1683 session->total_hits = 0;
1684 session->total_records = 0;
1685 session->num_termlists = 0;
1686 session->reclist = 0;
1687 session->requestid = -1;
1688 session->clients = 0;
1689 session->expected_maxrecs = 0;
1690 session->session_nmem = nmem;
1691 session->nmem = nmem_create();
1692 session->wrbuf = wrbuf_alloc();
1693 session_init_databases(session);
1694 for (i = 0; i <= SESSION_WATCH_MAX; i++)
1696 session->watchlist[i].data = 0;
1697 session->watchlist[i].fun = 0;
1703 struct hitsbytarget *hitsbytarget(struct session *se, int *count)
1705 static struct hitsbytarget res[1000]; // FIXME MM
1709 for (cl = se->clients; cl; cl = cl->next)
1711 char *name = session_setting_oneval(cl->database, PZ_NAME);
1713 res[*count].id = cl->database->database->url;
1714 res[*count].name = *name ? name : "Unknown";
1715 res[*count].hits = cl->hits;
1716 res[*count].records = cl->records;
1717 res[*count].diagnostic = cl->diagnostic;
1718 res[*count].state = client_states[cl->state];
1719 res[*count].connected = cl->connection ? 1 : 0;
1726 struct termlist_score **termlist(struct session *s, const char *name, int *num)
1730 for (i = 0; i < s->num_termlists; i++)
1731 if (!strcmp((const char *) s->termlists[i].name, name))
1732 return termlist_highscore(s->termlists[i].termlist, num);
1736 #ifdef MISSING_HEADERS
1737 void report_nmem_stats(void)
1739 size_t in_use, is_free;
1741 nmem_get_memory_in_use(&in_use);
1742 nmem_get_memory_free(&is_free);
1744 yaz_log(YLOG_LOG, "nmem stat: use=%ld free=%ld",
1745 (long) in_use, (long) is_free);
1749 struct record_cluster *show_single(struct session *s, int id)
1751 struct record_cluster *r;
1753 reclist_rewind(s->reclist);
1754 while ((r = reclist_read_record(s->reclist)))
1760 struct record_cluster **show(struct session *s, struct reclist_sortparms *sp, int start,
1761 int *num, int *total, int *sumhits, NMEM nmem_show)
1763 struct record_cluster **recs = nmem_malloc(nmem_show, *num
1764 * sizeof(struct record_cluster *));
1765 struct reclist_sortparms *spp;
1768 yaz_timing_t t = yaz_timing_create();
1771 for (spp = sp; spp; spp = spp->next)
1772 if (spp->type == Metadata_sortkey_relevance)
1774 relevance_prepare_read(s->relevance, s->reclist);
1777 reclist_sort(s->reclist, sp);
1779 *total = s->reclist->num_records;
1780 *sumhits = s->total_hits;
1782 for (i = 0; i < start; i++)
1783 if (!reclist_read_record(s->reclist))
1790 for (i = 0; i < *num; i++)
1792 struct record_cluster *r = reclist_read_record(s->reclist);
1802 yaz_log(YLOG_LOG, "show %6.5f %3.2f %3.2f",
1803 yaz_timing_get_real(t), yaz_timing_get_user(t),
1804 yaz_timing_get_sys(t));
1805 yaz_timing_destroy(&t);
1810 void statistics(struct session *se, struct statistics *stat)
1815 memset(stat, 0, sizeof(*stat));
1816 for (cl = se->clients; cl; cl = cl->next)
1818 if (!cl->connection)
1819 stat->num_no_connection++;
1822 case Client_Connecting: stat->num_connecting++; break;
1823 case Client_Initializing: stat->num_initializing++; break;
1824 case Client_Searching: stat->num_searching++; break;
1825 case Client_Presenting: stat->num_presenting++; break;
1826 case Client_Idle: stat->num_idle++; break;
1827 case Client_Failed: stat->num_failed++; break;
1828 case Client_Error: stat->num_error++; break;
1833 stat->num_hits = se->total_hits;
1834 stat->num_records = se->total_records;
1836 stat->num_clients = count;
1839 void start_http_listener(void)
1842 struct conf_server *ser = global_parameters.server;
1844 if (*global_parameters.listener_override)
1845 strcpy(hp, global_parameters.listener_override);
1848 strcpy(hp, ser->host ? ser->host : "");
1853 sprintf(hp + strlen(hp), "%d", ser->port);
1859 void start_proxy(void)
1862 struct conf_server *ser = global_parameters.server;
1864 if (*global_parameters.proxy_override)
1865 strcpy(hp, global_parameters.proxy_override);
1866 else if (ser->proxy_host || ser->proxy_port)
1868 strcpy(hp, ser->proxy_host ? ser->proxy_host : "");
1869 if (ser->proxy_port)
1873 sprintf(hp + strlen(hp), "%d", ser->proxy_port);
1879 http_set_proxyaddr(hp, ser->myurl ? ser->myurl : "");
1882 void start_zproxy(void)
1884 struct conf_server *ser = global_parameters.server;
1886 if (*global_parameters.zproxy_override){
1887 yaz_log(YLOG_LOG, "Z39.50 proxy %s",
1888 global_parameters.zproxy_override);
1892 else if (ser->zproxy_host || ser->zproxy_port)
1896 strcpy(hp, ser->zproxy_host ? ser->zproxy_host : "");
1897 if (ser->zproxy_port)
1904 sprintf(hp + strlen(hp), "%d", ser->zproxy_port);
1906 strcpy(global_parameters.zproxy_override, hp);
1907 yaz_log(YLOG_LOG, "Z39.50 proxy %s",
1908 global_parameters.zproxy_override);
1915 // Master list of connections we're handling events to
1916 static IOCHAN channel_list = 0;
1917 void pazpar2_add_channel(IOCHAN chan)
1919 chan->next = channel_list;
1920 channel_list = chan;
1923 void pazpar2_event_loop()
1925 event_loop(&channel_list);
1931 * indent-tabs-mode: nil
1933 * vim: shiftwidth=4 tabstop=8 expandtab