1 /* $Id: pazpar2.c,v 1.36 2007-01-16 05:29:48 quinn Exp $ */
8 #include <sys/socket.h>
14 #include <yaz/marcdisp.h>
15 #include <yaz/comstack.h>
16 #include <yaz/tcpip.h>
17 #include <yaz/proto.h>
18 #include <yaz/readconf.h>
19 #include <yaz/pquery.h>
20 #include <yaz/yaz-util.h>
29 #include <yaz/timing.h>
32 #include <netinet/in.h>
37 #include "termlists.h"
39 #include "relevance.h"
44 static void client_fatal(struct client *cl);
45 static void connection_destroy(struct connection *co);
46 static int client_prep_connection(struct client *cl);
47 static void ingest_records(struct client *cl, Z_Records *r);
48 static struct conf_retrievalprofile *database_retrieval_profile(struct database *db);
49 void session_alert_watch(struct session *s, int what);
51 IOCHAN channel_list = 0; // Master list of connections we're handling events to
53 static struct connection *connection_freelist = 0;
54 static struct client *client_freelist = 0;
56 static struct host *hosts = 0; // The hosts we know about
57 static struct database *databases = 0; // The databases we know about
59 static char *client_states[] = {
63 "Client_Initializing",
68 "Client_Disconnected",
72 // Note: Some things in this structure will eventually move to configuration
73 struct parameters global_parameters =
81 "Index Data PazPar2 (MasterKey)",
93 static int send_apdu(struct client *c, Z_APDU *a)
95 struct connection *co = c->connection;
99 if (!z_APDU(global_parameters.odr_out, &a, 0, 0))
101 odr_perror(global_parameters.odr_out, "Encoding APDU");
104 buf = odr_getbuf(global_parameters.odr_out, &len, 0);
105 r = cs_put(co->link, buf, len);
108 yaz_log(YLOG_WARN, "cs_put: %s", cs_errmsg(cs_errno(co->link)));
113 fprintf(stderr, "cs_put incomplete (ParaZ does not handle that)\n");
116 odr_reset(global_parameters.odr_out); /* release the APDU structure */
117 co->state = Conn_Waiting;
122 static void send_init(IOCHAN i)
124 struct connection *co = iochan_getdata(i);
125 struct client *cl = co->client;
126 Z_APDU *a = zget_APDU(global_parameters.odr_out, Z_APDU_initRequest);
128 a->u.initRequest->implementationId = global_parameters.implementationId;
129 a->u.initRequest->implementationName = global_parameters.implementationName;
130 a->u.initRequest->implementationVersion =
131 global_parameters.implementationVersion;
132 ODR_MASK_SET(a->u.initRequest->options, Z_Options_search);
133 ODR_MASK_SET(a->u.initRequest->options, Z_Options_present);
134 ODR_MASK_SET(a->u.initRequest->options, Z_Options_namedResultSets);
136 ODR_MASK_SET(a->u.initRequest->protocolVersion, Z_ProtocolVersion_1);
137 ODR_MASK_SET(a->u.initRequest->protocolVersion, Z_ProtocolVersion_2);
138 ODR_MASK_SET(a->u.initRequest->protocolVersion, Z_ProtocolVersion_3);
139 if (send_apdu(cl, a) >= 0)
141 iochan_setflags(i, EVENT_INPUT);
142 cl->state = Client_Initializing;
145 cl->state = Client_Error;
146 odr_reset(global_parameters.odr_out);
149 static void send_search(IOCHAN i)
151 struct connection *co = iochan_getdata(i);
152 struct client *cl = co->client;
153 struct session *se = cl->session;
154 struct database *db = cl->database;
155 Z_APDU *a = zget_APDU(global_parameters.odr_out, Z_APDU_searchRequest);
156 int ndb, cerror, cpos;
159 struct ccl_rpn_node *cn;
160 int ssub = 0, lslb = 100000, mspn = 10;
162 yaz_log(YLOG_DEBUG, "Sending search");
164 cn = ccl_find_str(global_parameters.ccl_filter, se->query, &cerror, &cpos);
167 a->u.searchRequest->query = zquery = odr_malloc(global_parameters.odr_out,
169 zquery->which = Z_Query_type_1;
170 zquery->u.type_1 = ccl_rpn_query(global_parameters.odr_out, cn);
173 for (ndb = 0; db->databases[ndb]; ndb++)
175 databaselist = odr_malloc(global_parameters.odr_out, sizeof(char*) * ndb);
176 for (ndb = 0; db->databases[ndb]; ndb++)
177 databaselist[ndb] = db->databases[ndb];
179 a->u.presentRequest->preferredRecordSyntax =
180 yaz_oidval_to_z3950oid(global_parameters.odr_out,
181 CLASS_RECSYN, VAL_USMARC);
182 a->u.searchRequest->smallSetUpperBound = &ssub;
183 a->u.searchRequest->largeSetLowerBound = &lslb;
184 a->u.searchRequest->mediumSetPresentNumber = &mspn;
185 a->u.searchRequest->resultSetName = "Default";
186 a->u.searchRequest->databaseNames = databaselist;
187 a->u.searchRequest->num_databaseNames = ndb;
189 if (send_apdu(cl, a) >= 0)
191 iochan_setflags(i, EVENT_INPUT);
192 cl->state = Client_Searching;
193 cl->requestid = se->requestid;
196 cl->state = Client_Error;
198 odr_reset(global_parameters.odr_out);
201 static void send_present(IOCHAN i)
203 struct connection *co = iochan_getdata(i);
204 struct client *cl = co->client;
205 Z_APDU *a = zget_APDU(global_parameters.odr_out, Z_APDU_presentRequest);
207 int start = cl->records + 1;
209 toget = global_parameters.chunk;
210 if (toget > global_parameters.toget - cl->records)
211 toget = global_parameters.toget - cl->records;
212 if (toget > cl->hits - cl->records)
213 toget = cl->hits - cl->records;
215 yaz_log(YLOG_DEBUG, "Trying to present %d records\n", toget);
217 a->u.presentRequest->resultSetStartPoint = &start;
218 a->u.presentRequest->numberOfRecordsRequested = &toget;
220 a->u.presentRequest->resultSetId = "Default";
222 a->u.presentRequest->preferredRecordSyntax =
223 yaz_oidval_to_z3950oid(global_parameters.odr_out,
224 CLASS_RECSYN, VAL_USMARC);
226 if (send_apdu(cl, a) >= 0)
228 iochan_setflags(i, EVENT_INPUT);
229 cl->state = Client_Presenting;
232 cl->state = Client_Error;
233 odr_reset(global_parameters.odr_out);
236 static void do_initResponse(IOCHAN i, Z_APDU *a)
238 struct connection *co = iochan_getdata(i);
239 struct client *cl = co->client;
240 Z_InitResponse *r = a->u.initResponse;
242 yaz_log(YLOG_DEBUG, "Received init response");
246 cl->state = Client_Idle;
249 cl->state = Client_Failed; // FIXME need to do something to the connection
252 static void do_searchResponse(IOCHAN i, Z_APDU *a)
254 struct connection *co = iochan_getdata(i);
255 struct client *cl = co->client;
256 struct session *se = cl->session;
257 Z_SearchResponse *r = a->u.searchResponse;
259 yaz_log(YLOG_DEBUG, "Searchresponse (status=%d)", *r->searchStatus);
261 if (*r->searchStatus)
263 cl->hits = *r->resultCount;
264 se->total_hits += cl->hits;
265 if (r->presentStatus && !*r->presentStatus && r->records)
267 yaz_log(YLOG_DEBUG, "Records in search response");
268 cl->records += *r->numberOfRecordsReturned;
269 ingest_records(cl, r->records);
271 cl->state = Client_Idle;
276 cl->state = Client_Error;
278 Z_Records *recs = r->records;
279 if (recs->which == Z_Records_NSD)
281 yaz_log(YLOG_WARN, "Non-surrogate diagnostic");
282 cl->diagnostic = *recs->u.nonSurrogateDiagnostic->condition;
283 cl->state = Client_Error;
289 char *normalize_mergekey(char *buf, int skiparticle)
291 char *p = buf, *pout = buf;
296 char articles[] = "the den der die des an a "; // must end in space
298 while (*p && !isalnum(*p))
301 while (*p && *p != ' ' && pout - firstword < 62)
302 *(pout++) = tolower(*(p++));
305 if (!strstr(articles, firstword))
312 while (*p && !isalnum(*p))
315 *(pout++) = tolower(*(p++));
318 while (*p && !isalnum(*p))
325 while (pout > buf && *pout == ' ');
332 // FIXME needs to be generalized. Should flexibly generate X lists per search
333 static void extract_subject(struct session *s, const char *rec)
335 const char *field, *subfield;
337 while ((field = find_field(rec, "650")))
340 if ((subfield = find_subfield(field, 'a')))
346 ef = index(subfield, '\n');
349 if ((e = index(subfield, '\t')) && e < ef)
351 while (ef > subfield && !isalpha(*(ef - 1)) && *(ef - 1) != ')')
355 memcpy(buf, subfield, len);
359 termlist_insert(s->termlist, buf);
366 static void add_facet(struct session *s, const char *type, const char *value)
372 for (i = 0; i < s->num_termlists; i++)
373 if (!strcmp(s->termlists[i].name, type))
375 if (i == s->num_termlists)
377 if (i == SESSION_MAX_TERMLISTS)
379 yaz_log(YLOG_FATAL, "Too many termlists");
382 s->termlists[i].name = nmem_strdup(s->nmem, type);
383 s->termlists[i].termlist = termlist_create(s->nmem, s->expected_maxrecs, 15);
384 s->num_termlists = i + 1;
386 termlist_insert(s->termlists[i].termlist, value);
389 int yaz_marc_write_xml();
391 static xmlDoc *normalize_record(struct client *cl, Z_External *rec)
393 struct conf_retrievalprofile *rprofile = cl->database->rprofile;
394 struct conf_retrievalmap *m;
398 // First normalize to XML
399 if (rprofile->native_syntax == Nativesyn_iso2709)
403 if (rec->which != Z_External_octet)
405 yaz_log(YLOG_WARN, "Unexpected external branch, probably BER");
408 buf = (char*) rec->u.octet_aligned->buf;
409 len = rec->u.octet_aligned->len;
410 if (yaz_marc_read_iso2709(rprofile->yaz_marc, buf, len) < 0)
412 yaz_log(YLOG_WARN, "Failed to decode MARC");
415 if (yaz_marc_write_xml(rprofile->yaz_marc, &res,
416 "http://www.loc.gov/MARC21/slim", 0, 0) < 0)
418 yaz_log(YLOG_WARN, "Failed to encode as XML");
421 rdoc = xmlNewDoc("1.0");
422 xmlDocSetRootElement(rdoc, res);
426 yaz_log(YLOG_FATAL, "Unknown native_syntax in normalize_record");
429 for (m = rprofile->maplist; m; m = m->next)
432 if (m->type != Map_xslt)
434 yaz_log(YLOG_WARN, "Unknown map type");
437 if (!(new = xsltApplyStylesheet(m->stylesheet, rdoc, 0)))
439 yaz_log(YLOG_WARN, "XSLT transformation failed");
445 if (global_parameters.dump_records)
447 fprintf(stderr, "Record:\n----------------\n");
448 #if LIBXML_VERSION >= 20600
449 xmlDocFormatDump(stderr, rdoc, 1);
451 xmlDocDump(stderr, rdoc);
457 // Extract what appears to be years from buf, storing highest and
459 static int extract_years(const char *buf, int *first, int *last)
468 while (*buf && !isdigit(*buf))
471 for (e = buf; *e && isdigit(*e); e++)
475 int value = atoi(buf);
476 if (*first < 0 || value < *first)
478 if (*last < 0 || value > *last)
486 static struct record *ingest_record(struct client *cl, Z_External *rec)
488 xmlDoc *xdoc = normalize_record(cl, rec);
491 struct record_cluster *cluster;
492 struct session *se = cl->session;
493 xmlChar *mergekey, *mergekey_norm;
496 struct conf_service *service = global_parameters.server->service;
501 root = xmlDocGetRootElement(xdoc);
502 if (!(mergekey = xmlGetProp(root, "mergekey")))
504 yaz_log(YLOG_WARN, "No mergekey found in record");
509 res = nmem_malloc(se->nmem, sizeof(struct record));
511 res->metadata = nmem_malloc(se->nmem,
512 sizeof(struct record_metadata*) * service->num_metadata);
513 memset(res->metadata, 0, sizeof(struct record_metadata*) * service->num_metadata);
515 mergekey_norm = nmem_strdup(se->nmem, (char*) mergekey);
517 normalize_mergekey(mergekey_norm, 0);
519 cluster = reclist_insert(se->reclist, res, mergekey_norm, &se->total_merged);
522 /* no room for record */
526 relevance_newrec(se->relevance, cluster);
529 for (n = root->children; n; n = n->next)
537 if (n->type != XML_ELEMENT_NODE)
539 if (!strcmp(n->name, "metadata"))
541 struct conf_metadata *md = 0;
542 struct conf_sortkey *sk = 0;
543 struct record_metadata **wheretoput, *newm;
547 type = xmlGetProp(n, "type");
548 value = xmlNodeListGetString(xdoc, n->children, 0);
549 // First, find out what field we're looking at
550 for (imeta = 0; imeta < service->num_metadata; imeta++)
551 if (!strcmp(type, service->metadata[imeta].name))
553 md = &service->metadata[imeta];
554 if (md->sortkey_offset >= 0)
555 sk = &service->sortkeys[md->sortkey_offset];
560 yaz_log(YLOG_WARN, "Ignoring unknown metadata element: %s", type);
564 // Find out where we are putting it
565 if (md->merge == Metadata_merge_no)
566 wheretoput = &res->metadata[imeta];
568 wheretoput = &cluster->metadata[imeta];
571 newm = nmem_malloc(se->nmem, sizeof(struct record_metadata));
573 if (md->type == Metadata_type_generic)
576 newm->data.text = nmem_strdup(se->nmem, value);
577 for (p = newm->data.text + strlen(newm->data.text) - 1;
578 p > newm->data.text && strchr(" ,/.", *p); p--)
582 else if (md->type == Metadata_type_year)
584 if (extract_years(value, &first, &last) < 0)
589 yaz_log(YLOG_WARN, "Unknown type in metadata element %s", type);
592 if (md->type == Metadata_type_year && md->merge != Metadata_merge_range)
594 yaz_log(YLOG_WARN, "Only range merging supported for years");
597 if (md->merge == Metadata_merge_unique)
599 struct record_metadata *mnode;
600 for (mnode = *wheretoput; mnode; mnode = mnode->next)
601 if (!strcmp(mnode->data.text, newm->data.text))
605 newm->next = *wheretoput;
609 else if (md->merge == Metadata_merge_longest)
612 strlen(newm->data.text) > strlen((*wheretoput)->data.text))
617 char *s = nmem_strdup(se->nmem, newm->data.text);
618 if (!cluster->sortkeys[md->sortkey_offset])
619 cluster->sortkeys[md->sortkey_offset] =
620 nmem_malloc(se->nmem, sizeof(union data_types));
621 normalize_mergekey(s,
622 (sk->type == Metadata_sortkey_skiparticle));
623 cluster->sortkeys[md->sortkey_offset]->text = s;
627 else if (md->merge == Metadata_merge_all || md->merge == Metadata_merge_no)
629 newm->next = *wheretoput;
632 else if (md->merge == Metadata_merge_range)
634 assert(md->type == Metadata_type_year);
638 (*wheretoput)->data.number.min = first;
639 (*wheretoput)->data.number.max = last;
641 cluster->sortkeys[md->sortkey_offset] = &newm->data;
645 if (first < (*wheretoput)->data.number.min)
646 (*wheretoput)->data.number.min = first;
647 if (last > (*wheretoput)->data.number.max)
648 (*wheretoput)->data.number.max = last;
653 union data_types *sdata = cluster->sortkeys[md->sortkey_offset];
654 yaz_log(YLOG_LOG, "SK range: %d-%d", sdata->number.min, sdata->number.max);
659 yaz_log(YLOG_WARN, "Don't know how to merge on element name %s", md->name);
662 relevance_countwords(se->relevance, cluster, value, md->rank);
664 add_facet(se, type, value);
670 yaz_log(YLOG_WARN, "Unexpected element %s in internal record", n->name);
675 relevance_donerecord(se->relevance, cluster);
681 static void ingest_records(struct client *cl, Z_Records *r)
684 yaz_timing_t t = yaz_timing_create();
687 struct session *s = cl->session;
688 Z_NamePlusRecordList *rlist;
691 if (r->which != Z_Records_DBOSD)
693 rlist = r->u.databaseOrSurDiagnostics;
694 for (i = 0; i < rlist->num_records; i++)
696 Z_NamePlusRecord *npr = rlist->records[i];
698 if (npr->which != Z_NamePlusRecord_databaseRecord)
700 yaz_log(YLOG_WARN, "Unexpected record type, probably diagnostic");
704 rec = ingest_record(cl, npr->u.databaseRecord);
708 if (s->watchlist[SESSION_WATCH_RECORDS].fun && rlist->num_records)
709 session_alert_watch(s, SESSION_WATCH_RECORDS);
713 yaz_log(YLOG_LOG, "ingest_records %6.5f %3.2f %3.2f",
714 yaz_timing_get_real(t), yaz_timing_get_user(t),
715 yaz_timing_get_sys(t));
716 yaz_timing_destroy(&t);
720 static void do_presentResponse(IOCHAN i, Z_APDU *a)
722 struct connection *co = iochan_getdata(i);
723 struct client *cl = co->client;
724 Z_PresentResponse *r = a->u.presentResponse;
727 Z_Records *recs = r->records;
728 if (recs->which == Z_Records_NSD)
730 yaz_log(YLOG_WARN, "Non-surrogate diagnostic");
731 cl->diagnostic = *recs->u.nonSurrogateDiagnostic->condition;
732 cl->state = Client_Error;
736 if (!*r->presentStatus && cl->state != Client_Error)
738 yaz_log(YLOG_DEBUG, "Good Present response");
739 cl->records += *r->numberOfRecordsReturned;
740 ingest_records(cl, r->records);
741 cl->state = Client_Idle;
743 else if (*r->presentStatus)
745 yaz_log(YLOG_WARN, "Bad Present response");
746 cl->state = Client_Error;
750 static void handler(IOCHAN i, int event)
752 struct connection *co = iochan_getdata(i);
753 struct client *cl = co->client;
754 struct session *se = 0;
760 yaz_log(YLOG_WARN, "Destroying orphan connection");
761 connection_destroy(co);
765 if (co->state == Conn_Connecting && event & EVENT_OUTPUT)
768 socklen_t errlen = sizeof(errcode);
770 if (getsockopt(cs_fileno(co->link), SOL_SOCKET, SO_ERROR, &errcode,
771 &errlen) < 0 || errcode != 0)
778 yaz_log(YLOG_DEBUG, "Connect OK");
779 co->state = Conn_Open;
781 cl->state = Client_Connected;
785 else if (event & EVENT_INPUT)
787 int len = cs_get(co->link, &co->ibuf, &co->ibufsize);
791 yaz_log(YLOG_WARN|YLOG_ERRNO, "Error reading from Z server");
792 connection_destroy(co);
797 yaz_log(YLOG_WARN, "EOF reading from Z server");
798 connection_destroy(co);
801 else if (len > 1) // We discard input if we have no connection
803 co->state = Conn_Open;
805 if (cl && (cl->requestid == se->requestid || cl->state == Client_Initializing))
809 odr_reset(global_parameters.odr_in);
810 odr_setbuf(global_parameters.odr_in, co->ibuf, len, 0);
811 if (!z_APDU(global_parameters.odr_in, &a, 0, 0))
818 case Z_APDU_initResponse:
819 do_initResponse(i, a);
821 case Z_APDU_searchResponse:
822 do_searchResponse(i, a);
824 case Z_APDU_presentResponse:
825 do_presentResponse(i, a);
828 yaz_log(YLOG_WARN, "Unexpected result from server");
832 // We aren't expecting staggered output from target
833 // if (cs_more(t->link))
834 // iochan_setevent(i, EVENT_INPUT);
836 else // we throw away response and go to idle mode
838 yaz_log(YLOG_DEBUG, "Ignoring result of expired operation");
839 cl->state = Client_Idle;
842 /* if len==1 we do nothing but wait for more input */
845 if (cl->state == Client_Connected) {
849 if (cl->state == Client_Idle)
851 if (cl->requestid != se->requestid && *se->query) {
854 else if (cl->hits > 0 && cl->records < global_parameters.toget &&
855 cl->records < cl->hits) {
861 // Disassociate connection from client
862 static void connection_release(struct connection *co)
864 struct client *cl = co->client;
866 yaz_log(YLOG_DEBUG, "Connection release %s", co->host->hostport);
873 // Close connection and recycle structure
874 static void connection_destroy(struct connection *co)
876 struct host *h = co->host;
878 iochan_destroy(co->iochan);
880 yaz_log(YLOG_DEBUG, "Connection destroy %s", co->host->hostport);
881 if (h->connections == co)
882 h->connections = co->next;
885 struct connection *pco;
886 for (pco = h->connections; pco && pco->next != co; pco = pco->next)
889 pco->next = co->next;
895 if (co->client->state != Client_Idle)
896 co->client->state = Client_Disconnected;
897 co->client->connection = 0;
899 co->next = connection_freelist;
900 connection_freelist = co;
903 // Creates a new connection for client, associated with the host of
905 static struct connection *connection_create(struct client *cl)
907 struct connection *new;
912 yaz_log(YLOG_DEBUG, "Connection create %s", cl->database->url);
913 if (!(link = cs_create(tcpip_type, 0, PROTO_Z3950)))
915 yaz_log(YLOG_FATAL|YLOG_ERRNO, "Failed to create comstack");
919 if (!(addr = cs_straddr(link, cl->database->host->ipport)))
921 yaz_log(YLOG_WARN|YLOG_ERRNO, "Lookup of IP address %s failed?",
922 cl->database->host->ipport);
926 res = cs_connect(link, addr);
929 yaz_log(YLOG_WARN|YLOG_ERRNO, "cs_connect %s", cl->database->url);
933 if ((new = connection_freelist))
934 connection_freelist = new->next;
937 new = xmalloc(sizeof (struct connection));
941 new->state = Conn_Connecting;
942 new->host = cl->database->host;
943 new->next = new->host->connections;
944 new->host->connections = new;
946 cl->connection = new;
949 new->iochan = iochan_create(cs_fileno(link), handler, 0);
950 iochan_setdata(new->iochan, new);
951 new->iochan->next = channel_list;
952 channel_list = new->iochan;
956 // Close connection and set state to error
957 static void client_fatal(struct client *cl)
959 yaz_log(YLOG_WARN, "Fatal error from %s", cl->database->url);
960 connection_destroy(cl->connection);
961 cl->state = Client_Error;
964 // Ensure that client has a connection associated
965 static int client_prep_connection(struct client *cl)
967 struct connection *co;
968 struct session *se = cl->session;
969 struct host *host = cl->database->host;
973 yaz_log(YLOG_DEBUG, "Client prep %s", cl->database->url);
977 // See if someone else has an idle connection
978 // We should look at timestamps here to select the longest-idle connection
979 for (co = host->connections; co; co = co->next)
980 if (co->state == Conn_Open && (!co->client || co->client->session != se))
984 connection_release(co);
989 co = connection_create(cl);
993 if (co->state == Conn_Connecting)
995 cl->state = Client_Connecting;
996 iochan_setflag(co->iochan, EVENT_OUTPUT);
998 else if (co->state == Conn_Open)
1000 if (cl->state == Client_Error || cl->state == Client_Disconnected)
1001 cl->state = Client_Idle;
1002 iochan_setflag(co->iochan, EVENT_OUTPUT);
1010 // This function will most likely vanish when a proper target profile mechanism is
1012 void load_simpletargets(const char *fn)
1014 FILE *f = fopen(fn, "r");
1019 yaz_log(YLOG_WARN|YLOG_ERRNO, "open %s", fn);
1023 while (fgets(line, 255, f))
1027 struct database *database;
1029 if (strncmp(line, "target ", 7))
1032 url[strlen(url) - 1] = '\0';
1033 yaz_log(YLOG_DEBUG, "Target: %s", url);
1034 if ((db = strchr(url, '/')))
1039 for (host = hosts; host; host = host->next)
1040 if (!strcmp(url, host->hostport))
1044 struct addrinfo *addrinfo, hints;
1047 unsigned char addrbuf[4];
1050 host = xmalloc(sizeof(struct host));
1051 host->hostport = xstrdup(url);
1052 host->connections = 0;
1054 if ((port = strchr(url, ':')))
1060 hints.ai_family = PF_INET;
1061 hints.ai_socktype = SOCK_STREAM;
1062 hints.ai_protocol = IPPROTO_TCP;
1063 hints.ai_addrlen = 0;
1065 hints.ai_canonname = 0;
1067 // This is not robust code. It assumes that getaddrinfo returns AF_INET
1069 if ((res = getaddrinfo(url, port, &hints, &addrinfo)))
1071 yaz_log(YLOG_WARN, "Failed to resolve %s: %s", url, gai_strerror(res));
1072 xfree(host->hostport);
1076 assert(addrinfo->ai_family == PF_INET);
1077 memcpy(addrbuf, &((struct sockaddr_in*)addrinfo->ai_addr)->sin_addr.s_addr, 4);
1078 sprintf(ipport, "%u.%u.%u.%u:%s",
1079 addrbuf[0], addrbuf[1], addrbuf[2], addrbuf[3], port);
1080 host->ipport = xstrdup(ipport);
1081 freeaddrinfo(addrinfo);
1085 database = xmalloc(sizeof(struct database));
1086 database->host = host;
1087 database->url = xmalloc(strlen(url) + strlen(db) + 2);
1088 strcpy(database->url, url);
1089 strcat(database->url, "/");
1090 strcat(database->url, db);
1092 database->databases = xmalloc(2 * sizeof(char *));
1093 database->databases[0] = xstrdup(db);
1094 database->databases[1] = 0;
1095 database->errors = 0;
1096 database->qprofile = 0;
1097 database->rprofile = database_retrieval_profile(database);
1098 database->next = databases;
1099 databases = database;
1105 static void pull_terms(NMEM nmem, struct ccl_rpn_node *n, char **termlist, int *num)
1113 pull_terms(nmem, n->u.p[0], termlist, num);
1114 pull_terms(nmem, n->u.p[1], termlist, num);
1117 termlist[(*num)++] = nmem_strdup(nmem, n->u.t.term);
1124 // Extract terms from query into null-terminated termlist
1125 static int extract_terms(NMEM nmem, char *query, char **termlist)
1128 struct ccl_rpn_node *n;
1131 n = ccl_find_str(global_parameters.ccl_filter, query, &error, &pos);
1134 pull_terms(nmem, n, termlist, &num);
1140 static struct client *client_create(void)
1143 if (client_freelist)
1145 r = client_freelist;
1146 client_freelist = client_freelist->next;
1149 r = xmalloc(sizeof(struct client));
1158 r->state = Client_Disconnected;
1163 void client_destroy(struct client *c)
1165 struct session *se = c->session;
1166 if (c == se->clients)
1167 se->clients = c->next;
1171 for (cc = se->clients; cc && cc->next != c; cc = cc->next)
1177 connection_release(c->connection);
1178 c->next = client_freelist;
1179 client_freelist = c;
1182 void session_set_watch(struct session *s, int what, session_watchfun fun, void *data)
1184 s->watchlist[what].fun = fun;
1185 s->watchlist[what].data = data;
1188 void session_alert_watch(struct session *s, int what)
1190 if (!s->watchlist[what].fun)
1192 (*s->watchlist[what].fun)(s->watchlist[what].data);
1193 s->watchlist[what].fun = 0;
1194 s->watchlist[what].data = 0;
1197 // This needs to be extended with selection criteria
1198 static struct conf_retrievalprofile *database_retrieval_profile(struct database *db)
1202 yaz_log(YLOG_FATAL, "Must load configuration (-f)");
1205 if (!config->retrievalprofiles)
1207 yaz_log(YLOG_FATAL, "No retrieval profiles defined");
1209 return config->retrievalprofiles;
1212 // This should be extended with parameters to control selection criteria
1213 // Associates a set of clients with a session;
1214 int select_targets(struct session *se)
1216 struct database *db;
1220 client_destroy(se->clients);
1221 for (db = databases; db; db = db->next)
1223 struct client *cl = client_create();
1226 cl->next = se->clients;
1233 int session_active_clients(struct session *s)
1238 for (c = s->clients; c; c = c->next)
1239 if (c->connection && (c->state == Client_Connecting ||
1240 c->state == Client_Initializing ||
1241 c->state == Client_Searching ||
1242 c->state == Client_Presenting))
1248 char *search(struct session *se, char *query)
1250 int live_channels = 0;
1253 yaz_log(YLOG_DEBUG, "Search");
1255 strcpy(se->query, query);
1257 nmem_reset(se->nmem);
1258 for (cl = se->clients; cl; cl = cl->next)
1264 if (client_prep_connection(cl))
1270 int maxrecs = live_channels * global_parameters.toget;
1271 se->num_termlists = 0;
1272 se->reclist = reclist_create(se->nmem, maxrecs);
1273 extract_terms(se->nmem, query, p);
1274 se->relevance = relevance_create(se->nmem, (const char **) p, maxrecs);
1275 se->total_records = se->total_hits = se->total_merged = 0;
1276 se->expected_maxrecs = maxrecs;
1284 void destroy_session(struct session *s)
1286 yaz_log(YLOG_LOG, "Destroying session");
1288 client_destroy(s->clients);
1289 nmem_destroy(s->nmem);
1290 wrbuf_free(s->wrbuf, 1);
1293 struct session *new_session()
1296 struct session *session = xmalloc(sizeof(*session));
1298 yaz_log(YLOG_DEBUG, "New pazpar2 session");
1300 session->total_hits = 0;
1301 session->total_records = 0;
1302 session->num_termlists = 0;
1303 session->reclist = 0;
1304 session->requestid = -1;
1305 session->clients = 0;
1306 session->expected_maxrecs = 0;
1307 session->query[0] = '\0';
1308 session->nmem = nmem_create();
1309 session->wrbuf = wrbuf_alloc();
1310 for (i = 0; i <= SESSION_WATCH_MAX; i++)
1312 session->watchlist[i].data = 0;
1313 session->watchlist[i].fun = 0;
1316 select_targets(session);
1321 struct hitsbytarget *hitsbytarget(struct session *se, int *count)
1323 static struct hitsbytarget res[1000]; // FIXME MM
1327 for (cl = se->clients; cl; cl = cl->next)
1329 strcpy(res[*count].id, cl->database->host->hostport);
1330 res[*count].hits = cl->hits;
1331 res[*count].records = cl->records;
1332 res[*count].diagnostic = cl->diagnostic;
1333 res[*count].state = client_states[cl->state];
1334 res[*count].connected = cl->connection ? 1 : 0;
1341 struct termlist_score **termlist(struct session *s, const char *name, int *num)
1345 for (i = 0; i < s->num_termlists; i++)
1346 if (!strcmp(s->termlists[i].name, name))
1347 return termlist_highscore(s->termlists[i].termlist, num);
1351 #ifdef MISSING_HEADERS
1352 void report_nmem_stats(void)
1354 size_t in_use, is_free;
1356 nmem_get_memory_in_use(&in_use);
1357 nmem_get_memory_free(&is_free);
1359 yaz_log(YLOG_LOG, "nmem stat: use=%ld free=%ld",
1360 (long) in_use, (long) is_free);
1364 struct record_cluster *show_single(struct session *s, int id)
1366 struct record_cluster *r;
1368 reclist_rewind(s->reclist);
1369 while ((r = reclist_read_record(s->reclist)))
1375 struct record_cluster **show(struct session *s, struct reclist_sortparms *sp, int start,
1376 int *num, int *total, int *sumhits, NMEM nmem_show)
1378 struct record_cluster **recs = nmem_malloc(nmem_show, *num
1379 * sizeof(struct record_cluster *));
1380 struct reclist_sortparms *spp;
1383 yaz_timing_t t = yaz_timing_create();
1386 for (spp = sp; spp; spp = spp->next)
1387 if (spp->type == Metadata_sortkey_relevance)
1389 relevance_prepare_read(s->relevance, s->reclist);
1392 reclist_sort(s->reclist, sp);
1394 *total = s->reclist->num_records;
1395 *sumhits = s->total_hits;
1397 for (i = 0; i < start; i++)
1398 if (!reclist_read_record(s->reclist))
1405 for (i = 0; i < *num; i++)
1407 struct record_cluster *r = reclist_read_record(s->reclist);
1417 yaz_log(YLOG_LOG, "show %6.5f %3.2f %3.2f",
1418 yaz_timing_get_real(t), yaz_timing_get_user(t),
1419 yaz_timing_get_sys(t));
1420 yaz_timing_destroy(&t);
1425 void statistics(struct session *se, struct statistics *stat)
1430 memset(stat, 0, sizeof(*stat));
1431 for (cl = se->clients; cl; cl = cl->next)
1433 if (!cl->connection)
1434 stat->num_no_connection++;
1437 case Client_Connecting: stat->num_connecting++; break;
1438 case Client_Initializing: stat->num_initializing++; break;
1439 case Client_Searching: stat->num_searching++; break;
1440 case Client_Presenting: stat->num_presenting++; break;
1441 case Client_Idle: stat->num_idle++; break;
1442 case Client_Failed: stat->num_failed++; break;
1443 case Client_Error: stat->num_error++; break;
1448 stat->num_hits = se->total_hits;
1449 stat->num_records = se->total_records;
1451 stat->num_clients = count;
1454 static CCL_bibset load_cclfile(const char *fn)
1456 CCL_bibset res = ccl_qual_mk();
1457 if (ccl_qual_fname(res, fn) < 0)
1459 yaz_log(YLOG_FATAL|YLOG_ERRNO, "%s", fn);
1465 static void start_http_listener(void)
1468 struct conf_server *ser = global_parameters.server;
1470 if (*global_parameters.listener_override)
1471 strcpy(hp, global_parameters.listener_override);
1474 strcpy(hp, ser->host ? ser->host : "");
1479 sprintf(hp + strlen(hp), "%d", ser->port);
1485 static void start_proxy(void)
1488 struct conf_server *ser = global_parameters.server;
1490 if (*global_parameters.proxy_override)
1491 strcpy(hp, global_parameters.proxy_override);
1492 else if (ser->proxy_host || ser->proxy_port)
1494 strcpy(hp, ser->proxy_host ? ser->proxy_host : "");
1495 if (ser->proxy_port)
1499 sprintf(hp + strlen(hp), "%d", ser->proxy_port);
1505 http_set_proxyaddr(hp);
1508 int main(int argc, char **argv)
1513 if (signal(SIGPIPE, SIG_IGN) == SIG_ERR)
1514 yaz_log(YLOG_WARN|YLOG_ERRNO, "signal");
1516 yaz_log_init(YLOG_DEFAULT_LEVEL, "pazpar2", 0);
1518 while ((ret = options("f:x:h:p:C:s:d", argv, argc, &arg)) != -2)
1522 if (!read_config(arg))
1526 strcpy(global_parameters.listener_override, arg);
1529 global_parameters.ccl_filter = load_cclfile(arg);
1532 strcpy(global_parameters.proxy_override, arg);
1535 load_simpletargets(arg);
1538 global_parameters.dump_records = 1;
1541 fprintf(stderr, "Usage: pazpar2\n"
1543 " -h [host:]port (REST protocol listener)\n"
1545 " -s simpletargetfile\n"
1546 " -p hostname[:portno] (HTTP proxy)\n");
1553 yaz_log(YLOG_FATAL, "Load config with -f");
1556 global_parameters.server = config->servers;
1558 start_http_listener();
1560 global_parameters.ccl_filter = load_cclfile("../etc/default.bib");
1561 global_parameters.yaz_marc = yaz_marc_create();
1562 yaz_marc_subfield_str(global_parameters.yaz_marc, "\t");
1563 global_parameters.odr_in = odr_createmem(ODR_DECODE);
1564 global_parameters.odr_out = odr_createmem(ODR_ENCODE);
1566 event_loop(&channel_list);
1574 * indent-tabs-mode: nil
1576 * vim: shiftwidth=4 tabstop=8 expandtab