f0f672c95d8b827c1eb292f7f22d913f12139b59
[pazpar2-moved-to-github.git] / src / session.c
1 /* This file is part of Pazpar2.
2    Copyright (C) 2006-2012 Index Data
3
4 Pazpar2 is free software; you can redistribute it and/or modify it under
5 the terms of the GNU General Public License as published by the Free
6 Software Foundation; either version 2, or (at your option) any later
7 version.
8
9 Pazpar2 is distributed in the hope that it will be useful, but WITHOUT ANY
10 WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
12 for more details.
13
14 You should have received a copy of the GNU General Public License
15 along with this program; if not, write to the Free Software
16 Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
17
18 */
19
20 /** \file session.c
21     \brief high-level logic; mostly user sessions and settings
22 */
23
24 #if HAVE_CONFIG_H
25 #include <config.h>
26 #endif
27
28 #include <time.h>
29 #include <stdlib.h>
30 #include <stdio.h>
31 #include <string.h>
32 #if HAVE_SYS_TIME_H
33 #include <sys/time.h>
34 #endif
35 #if HAVE_UNISTD_H
36 #include <unistd.h>
37 #endif
38 #ifdef WIN32
39 #include <windows.h>
40 #endif
41 #include <signal.h>
42 #include <ctype.h>
43 #include <assert.h>
44 #include <math.h>
45
46 #include <yaz/marcdisp.h>
47 #include <yaz/comstack.h>
48 #include <yaz/tcpip.h>
49 #include <yaz/proto.h>
50 #include <yaz/readconf.h>
51 #include <yaz/pquery.h>
52 #include <yaz/otherinfo.h>
53 #include <yaz/yaz-util.h>
54 #include <yaz/nmem.h>
55 #include <yaz/query-charset.h>
56 #include <yaz/querytowrbuf.h>
57 #include <yaz/oid_db.h>
58 #include <yaz/snprintf.h>
59 #include <yaz/gettimeofday.h>
60
61 #define USE_TIMING 0
62 #if USE_TIMING
63 #include <yaz/timing.h>
64 #endif
65
66 #include "ppmutex.h"
67 #include "parameters.h"
68 #include "session.h"
69 #include "eventl.h"
70 #include "http.h"
71 #include "termlists.h"
72 #include "reclists.h"
73 #include "relevance.h"
74 #include "database.h"
75 #include "client.h"
76 #include "settings.h"
77 #include "normalize7bit.h"
78
79 #include <libxml/tree.h>
80
81 #define TERMLIST_HIGH_SCORE 25
82
83 #define MAX_CHUNK 15
84
85 #define MAX(a,b) ((a)>(b)?(a):(b))
86
87 // Note: Some things in this structure will eventually move to configuration
88 struct parameters global_parameters = 
89 {
90     0,   // dump_records
91     0,   // debug_mode
92     0,   // predictable sessions
93 };
94
95 struct client_list {
96     struct client *client;
97     struct client_list *next;
98 };
99
100 /* session counting (1) , disable client counting (0) */
101 static YAZ_MUTEX g_session_mutex = 0;
102 static int no_sessions = 0;
103 static int no_session_total = 0;
104
105 static int session_use(int delta)
106 {
107     int sessions;
108     if (!g_session_mutex)
109         yaz_mutex_create(&g_session_mutex);
110     yaz_mutex_enter(g_session_mutex);
111     no_sessions += delta;
112     if (delta > 0)
113         no_session_total += delta;
114     sessions = no_sessions;
115     yaz_mutex_leave(g_session_mutex);
116     yaz_log(YLOG_DEBUG, "%s sessions=%d", delta == 0 ? "" : (delta > 0 ? "INC" : "DEC"), no_sessions);
117     return sessions;
118 }
119
120 int sessions_count(void)
121 {
122     return session_use(0);
123 }
124
125 int session_count_total(void)
126 {
127     int total = 0;
128     if (!g_session_mutex)
129         return 0;
130     yaz_mutex_enter(g_session_mutex);
131     total = no_session_total;
132     yaz_mutex_leave(g_session_mutex);
133     return total;
134 }
135
136 static void log_xml_doc(xmlDoc *doc)
137 {
138     FILE *lf = yaz_log_file();
139     xmlChar *result = 0;
140     int len = 0;
141 #if LIBXML_VERSION >= 20600
142     xmlDocDumpFormatMemory(doc, &result, &len, 1);
143 #else
144     xmlDocDumpMemory(doc, &result, &len);
145 #endif
146     if (lf && len)
147     {
148         (void) fwrite(result, 1, len, lf);
149         fprintf(lf, "\n");
150     }
151     xmlFree(result);
152 }
153
154 static void session_enter(struct session *s)
155 {
156     yaz_mutex_enter(s->session_mutex);
157 }
158
159 static void session_leave(struct session *s)
160 {
161     yaz_mutex_leave(s->session_mutex);
162 }
163
164 static void session_normalize_facet(struct session *s, const char *type,
165                                     const char *value,
166                                     WRBUF display_wrbuf,
167                                     WRBUF facet_wrbuf)
168 {
169     struct conf_service *service = s->service;
170     pp2_charset_token_t prt;
171     const char *facet_component;
172     int i;
173     const char *icu_chain_id = 0;
174
175     for (i = 0; i < service->num_metadata; i++)
176         if (!strcmp((service->metadata + i)->name, type))
177             icu_chain_id = (service->metadata + i)->facetrule;
178     if (!icu_chain_id)
179         icu_chain_id = "facet";
180     prt = pp2_charset_token_create(service->charsets, icu_chain_id);
181     if (!prt)
182     {
183         yaz_log(YLOG_FATAL, "Unknown ICU chain '%s' for facet of type '%s'",
184                 icu_chain_id, type);
185         wrbuf_destroy(facet_wrbuf);
186         wrbuf_destroy(display_wrbuf);
187         return;
188     }
189     pp2_charset_token_first(prt, value, 0);
190     while ((facet_component = pp2_charset_token_next(prt)))
191     {
192         const char *display_component;
193         if (*facet_component)
194         {
195             if (wrbuf_len(facet_wrbuf))
196                 wrbuf_puts(facet_wrbuf, " ");
197             wrbuf_puts(facet_wrbuf, facet_component);
198         }
199         display_component = pp2_get_display(prt);
200         if (display_component)
201         {
202             if (wrbuf_len(display_wrbuf))
203                 wrbuf_puts(display_wrbuf, " ");
204             wrbuf_puts(display_wrbuf, display_component);
205         }
206     }
207     pp2_charset_token_destroy(prt);
208 }
209
210 void add_facet(struct session *s, const char *type, const char *value, int count)
211 {
212     WRBUF facet_wrbuf = wrbuf_alloc();
213     WRBUF display_wrbuf = wrbuf_alloc();
214
215     session_normalize_facet(s, type, value, display_wrbuf, facet_wrbuf);
216  
217     if (wrbuf_len(facet_wrbuf))
218     {
219         int i;
220         for (i = 0; i < s->num_termlists; i++)
221             if (!strcmp(s->termlists[i].name, type))
222                 break;
223         if (i == s->num_termlists)
224         {
225             if (i == SESSION_MAX_TERMLISTS)
226             {
227                 session_log(s, YLOG_FATAL, "Too many termlists");
228                 wrbuf_destroy(facet_wrbuf);
229                 wrbuf_destroy(display_wrbuf);
230                 return;
231             }
232             
233             s->termlists[i].name = nmem_strdup(s->nmem, type);
234             s->termlists[i].termlist 
235                 = termlist_create(s->nmem, TERMLIST_HIGH_SCORE);
236             s->num_termlists = i + 1;
237         }
238         
239 #if 0
240         session_log(s, YLOG_DEBUG, "Facets for %s: %s norm:%s (%d)", type, value, wrbuf_cstr(facet_wrbuf), count);
241 #endif
242         termlist_insert(s->termlists[i].termlist, wrbuf_cstr(display_wrbuf),
243                         wrbuf_cstr(facet_wrbuf), count);
244     }
245     wrbuf_destroy(facet_wrbuf);
246     wrbuf_destroy(display_wrbuf);
247 }
248
249 static xmlDoc *record_to_xml(struct session *se,
250                              struct session_database *sdb, const char *rec)
251 {
252     struct database *db = sdb->database;
253     xmlDoc *rdoc = 0;
254
255     rdoc = xmlParseMemory(rec, strlen(rec));
256
257     if (!rdoc)
258     {
259         session_log(se, YLOG_FATAL, "Non-wellformed XML received from %s",
260                     db->id);
261         return 0;
262     }
263
264     if (global_parameters.dump_records)
265     {
266         session_log(se, YLOG_LOG, "Un-normalized record from %s", db->id);
267         log_xml_doc(rdoc);
268     }
269
270     return rdoc;
271 }
272
273 #define MAX_XSLT_ARGS 16
274
275 // Add static values from session database settings if applicable
276 static void insert_settings_parameters(struct session_database *sdb,
277                                        struct conf_service *service,
278                                        char **parms,
279                                        NMEM nmem)
280 {
281     int i;
282     int nparms = 0;
283     int offset = 0;
284
285     for (i = 0; i < service->num_metadata; i++)
286     {
287         struct conf_metadata *md = &service->metadata[i];
288         int setting;
289
290         if (md->setting == Metadata_setting_parameter &&
291             (setting = settings_lookup_offset(service, md->name)) >= 0)
292         {
293             const char *val = session_setting_oneval(sdb, setting);
294             if (val && nparms < MAX_XSLT_ARGS)
295             {
296                 char *buf;
297                 int len = strlen(val);
298                 buf = nmem_malloc(nmem, len + 3);
299                 buf[0] = '\'';
300                 strcpy(buf + 1, val);
301                 buf[len+1] = '\'';
302                 buf[len+2] = '\0';
303                 parms[offset++] = md->name;
304                 parms[offset++] = buf;
305                 nparms++;
306             }
307         }
308     }
309     parms[offset] = 0;
310 }
311
312 // Add static values from session database settings if applicable
313 static void insert_settings_values(struct session_database *sdb, xmlDoc *doc,
314     struct conf_service *service)
315 {
316     int i;
317
318     for (i = 0; i < service->num_metadata; i++)
319     {
320         struct conf_metadata *md = &service->metadata[i];
321         int offset;
322
323         if (md->setting == Metadata_setting_postproc &&
324             (offset = settings_lookup_offset(service, md->name)) >= 0)
325         {
326             const char *val = session_setting_oneval(sdb, offset);
327             if (val)
328             {
329                 xmlNode *r = xmlDocGetRootElement(doc);
330                 xmlNode *n = xmlNewTextChild(r, 0, (xmlChar *) "metadata",
331                                              (xmlChar *) val);
332                 xmlSetProp(n, (xmlChar *) "type", (xmlChar *) md->name);
333             }
334         }
335     }
336 }
337
338 static xmlDoc *normalize_record(struct session *se,
339                                 struct session_database *sdb,
340                                 struct conf_service *service,
341                                 const char *rec, NMEM nmem)
342 {
343     xmlDoc *rdoc = record_to_xml(se, sdb, rec);
344
345     if (rdoc)
346     {
347         char *parms[MAX_XSLT_ARGS*2+1];
348         
349         insert_settings_parameters(sdb, service, parms, nmem);
350         
351         if (normalize_record_transform(sdb->map, &rdoc, (const char **)parms))
352         {
353             session_log(se, YLOG_WARN, "Normalize failed from %s",
354                         sdb->database->id);
355         }
356         else
357         {
358             insert_settings_values(sdb, rdoc, service);
359             
360             if (global_parameters.dump_records)
361             {
362                 session_log(se, YLOG_LOG, "Normalized record from %s", 
363                             sdb->database->id);
364                 log_xml_doc(rdoc);
365             }
366         }
367     }
368     return rdoc;
369 }
370
371 void session_settings_dump(struct session *se,
372                            struct session_database *db,
373                            WRBUF w)
374 {
375     if (db->settings)
376     {
377         int i, num = db->num_settings;
378         for (i = 0; i < num; i++)
379         {
380             struct setting *s = db->settings[i];
381             for (;s ; s = s->next)
382             {
383                 wrbuf_puts(w, "<set name=\"");
384                 wrbuf_xmlputs(w, s->name);
385                 wrbuf_puts(w, "\" value=\"");
386                 wrbuf_xmlputs(w, s->value);
387                 wrbuf_puts(w, "\"/>");
388             }
389             if (db->settings[i])
390                 wrbuf_puts(w, "\n");
391         }
392     }
393 }
394
395 // Retrieve first defined value for 'name' for given database.
396 // Will be extended to take into account user associated with session
397 const char *session_setting_oneval(struct session_database *db, int offset)
398 {
399     if (offset >= db->num_settings || !db->settings[offset])
400         return "";
401     return db->settings[offset]->value;
402 }
403
404 // Prepare XSLT stylesheets for record normalization
405 // Structures are allocated on the session_wide nmem to avoid having
406 // to recompute this for every search. This would lead
407 // to leaking if a single session was to repeatedly change the PZ_XSLT
408 // setting. However, this is not a realistic use scenario.
409 static int prepare_map(struct session *se, struct session_database *sdb)
410 {
411     const char *s;
412
413     if (sdb->settings && sdb->settings[PZ_XSLT] && !sdb->map &&
414         (s = session_setting_oneval(sdb, PZ_XSLT)))        
415     {
416         char auto_stylesheet[256];
417
418         if (!strcmp(s, "auto"))
419         {
420             const char *request_syntax = session_setting_oneval(
421                 sdb, PZ_REQUESTSYNTAX);
422             if (request_syntax)
423             {
424                 char *cp;
425                 yaz_snprintf(auto_stylesheet, sizeof(auto_stylesheet),
426                              "%s.xsl", request_syntax);
427                 for (cp = auto_stylesheet; *cp; cp++)
428                 {
429                     /* deliberately only consider ASCII */
430                     if (*cp > 32 && *cp < 127)
431                         *cp = tolower(*cp);
432                 }
433                 s = auto_stylesheet;
434             }
435             else
436             {
437                 session_log(se, YLOG_WARN,
438                             "No pz:requestsyntax for auto stylesheet");
439             }
440         }
441         sdb->map = normalize_cache_get(se->normalize_cache,
442                                        se->service, s);
443         if (!sdb->map)
444             return -1;
445     }
446     return 0;
447 }
448
449 // called if watch should be removed because http_channel is to be destroyed
450 static void session_watch_cancel(void *data, struct http_channel *c,
451                                  void *data2)
452 {
453     struct session_watchentry *ent = data;
454
455     ent->fun = 0;
456     ent->data = 0;
457     ent->obs = 0;
458 }
459
460 // set watch. Returns 0=OK, -1 if watch is already set
461 int session_set_watch(struct session *s, int what, 
462                       session_watchfun fun, void *data,
463                       struct http_channel *chan)
464 {
465     int ret;
466     session_enter(s);
467     if (s->watchlist[what].fun)
468         ret = -1;
469     else
470     {
471         
472         s->watchlist[what].fun = fun;
473         s->watchlist[what].data = data;
474         s->watchlist[what].obs = http_add_observer(chan, &s->watchlist[what],
475                                                    session_watch_cancel);
476         ret = 0;
477     }
478     session_leave(s);
479     return ret;
480 }
481
482 void session_alert_watch(struct session *s, int what)
483 {
484     assert(s);
485     session_enter(s);
486     if (s->watchlist[what].fun)
487     {
488         /* our watch is no longer associated with http_channel */
489         void *data;
490         session_watchfun fun;
491
492         http_remove_observer(s->watchlist[what].obs);
493         fun  = s->watchlist[what].fun;
494         data = s->watchlist[what].data;
495
496         /* reset watch before fun is invoked - in case fun wants to set
497            it again */
498         s->watchlist[what].fun = 0;
499         s->watchlist[what].data = 0;
500         s->watchlist[what].obs = 0;
501
502         session_leave(s);
503         session_log(s, YLOG_DEBUG,
504                     "Alert Watch: %d calling function: %p", what, fun);
505         fun(data);
506     }
507     else
508         session_leave(s);
509 }
510
511 //callback for grep_databases
512 static void select_targets_callback(struct session *se,
513                                     struct session_database *db)
514 {
515     struct client *cl;
516     struct client_list *l;
517
518     for (l = se->clients_cached; l; l = l->next)
519         if (client_get_database(l->client) == db)
520             break;
521
522     if (l)
523         cl = l->client;
524     else
525     {
526         cl = client_create(db->database->id);
527         client_set_database(cl, db);
528
529         l = xmalloc(sizeof(*l));
530         l->client = cl;
531         l->next = se->clients_cached;
532         se->clients_cached = l;
533     }
534     /* set session always. If may be 0 if client is not active */
535     client_set_session(cl, se);
536
537     l = xmalloc(sizeof(*l));
538     l->client = cl;
539     l->next = se->clients_active;
540     se->clients_active = l;
541 }
542
543 static void session_reset_active_clients(struct session *se,
544                                          struct client_list *new_list)
545 {
546     struct client_list *l;
547
548     session_enter(se);
549     l = se->clients_active;
550     se->clients_active = new_list;
551     session_leave(se);
552
553     while (l)
554     {
555         struct client_list *l_next = l->next;
556
557         client_lock(l->client); 
558         client_set_session(l->client, 0); /* mark client inactive */
559         client_unlock(l->client);
560
561         xfree(l);
562         l = l_next;
563     }
564 }
565
566 static void session_remove_cached_clients(struct session *se)
567 {
568     struct client_list *l;
569
570     session_reset_active_clients(se, 0);
571
572     session_enter(se);
573     l = se->clients_cached;
574     se->clients_cached = 0;
575     session_leave(se);
576
577     while (l)
578     {
579         struct client_list *l_next = l->next;
580         client_lock(l->client);
581         client_set_session(l->client, 0);
582         client_set_database(l->client, 0);
583         client_unlock(l->client);
584         client_destroy(l->client);
585         xfree(l);
586         l = l_next;
587     }
588 }
589
590 // Associates a set of clients with a session;
591 // Note: Session-databases represent databases with per-session 
592 // setting overrides
593 static int select_targets(struct session *se, const char *filter)
594 {
595     return session_grep_databases(se, filter, select_targets_callback);
596 }
597
598 int session_active_clients(struct session *s)
599 {
600     struct client_list *l;
601     int res = 0;
602
603     for (l = s->clients_active; l; l = l->next)
604         if (client_is_active(l->client))
605             res++;
606
607     return res;
608 }
609
610 int session_is_preferred_clients_ready(struct session *s)
611 {
612     struct client_list *l;
613     int res = 0;
614
615     for (l = s->clients_active; l; l = l->next)
616         if (client_is_active_preferred(l->client))
617             res++;
618     session_log(s, YLOG_DEBUG, "Has %d active preferred clients.", res);
619     return res == 0;
620 }
621
622 void session_sort(struct session *se, const char *field, int increasing)
623 {
624     struct session_sorted_results *sr;
625     struct client_list *l;
626
627     session_enter(se);
628
629     /* see if we already have sorted for this critieria */
630     for (sr = se->sorted_results; sr; sr = sr->next)
631     {
632         if (!strcmp(field, sr->field) && increasing == sr->increasing)
633             break;
634     }
635     if (sr)
636     {
637         yaz_log(YLOG_LOG, "search_sort: field=%s increasing=%d already fetched",
638                 field, increasing);
639         session_leave(se);
640         return;
641     }
642     yaz_log(YLOG_LOG, "search_sort: field=%s increasing=%d must fetch",
643             field, increasing);
644     sr = nmem_malloc(se->nmem, sizeof(*sr));
645     sr->field = nmem_strdup(se->nmem, field);
646     sr->increasing = increasing;
647     sr->next = se->sorted_results;
648     se->sorted_results = sr;
649     
650     for (l = se->clients_active; l; l = l->next)
651     {
652         struct client *cl = l->client;
653         if (client_get_state(cl) == Client_Connecting ||
654             client_get_state(cl) == Client_Idle ||
655             client_get_state(cl) == Client_Working)
656             client_start_search(cl);
657     }
658     session_leave(se);
659 }
660
661 enum pazpar2_error_code session_search(struct session *se,
662                                        const char *query,
663                                        const char *startrecs,
664                                        const char *maxrecs,
665                                        const char *filter,
666                                        const char *limit,
667                                        const char **addinfo,
668                                        const char *sort_field, int increasing)
669 {
670     int live_channels = 0;
671     int no_working = 0;
672     int no_failed_query = 0;
673     int no_failed_limit = 0;
674     struct client_list *l, *l0;
675     struct timeval tval;
676     facet_limits_t facet_limits;
677
678     session_log(se, YLOG_DEBUG, "Search");
679
680     *addinfo = 0;
681
682     if (se->settings_modified)
683         session_remove_cached_clients(se);
684     else
685         session_reset_active_clients(se, 0);
686     
687     session_enter(se);
688     reclist_destroy(se->reclist);
689     se->reclist = 0;
690     se->settings_modified = 0;
691     relevance_destroy(&se->relevance);
692     nmem_reset(se->nmem);
693     se->total_records = se->total_merged = 0;
694     se->num_termlists = 0;
695
696     /* reset list of sorted results and clear to relevance search */
697     se->sorted_results = nmem_malloc(se->nmem, sizeof(*se->sorted_results));
698     se->sorted_results->field = nmem_strdup(se->nmem, sort_field);
699     se->sorted_results->increasing = increasing;
700     se->sorted_results->next = 0;
701     
702     live_channels = select_targets(se, filter);
703     if (!live_channels)
704     {
705         session_leave(se);
706         return PAZPAR2_NO_TARGETS;
707     }
708     se->reclist = reclist_create(se->nmem);
709
710     yaz_gettimeofday(&tval);
711     
712     tval.tv_sec += 5;
713
714     facet_limits = facet_limits_create(limit);
715     if (!facet_limits)
716     {
717         *addinfo = "limit";
718         session_leave(se);
719         return PAZPAR2_MALFORMED_PARAMETER_VALUE;
720     }
721
722     l0 = se->clients_active;
723     se->clients_active = 0;
724     session_leave(se);
725
726     for (l = l0; l; l = l->next)
727     {
728         int parse_ret;
729         struct client *cl = l->client;
730
731         if (prepare_map(se, client_get_database(cl)) < 0)
732             continue;
733
734         parse_ret = client_parse_query(cl, query, facet_limits, startrecs,
735             maxrecs);
736         if (parse_ret == -1)
737             no_failed_query++;
738         else if (parse_ret == -2)
739             no_failed_limit++;
740         else
741         {
742             int r =
743                 client_prep_connection(cl, se->service->z3950_operation_timeout,
744                                        se->service->z3950_session_timeout,
745                                        se->service->server->iochan_man,
746                                        &tval);
747             if (parse_ret == 1 && r == 2)
748             {
749                 session_log(se, YLOG_LOG, "client REUSE %s", client_get_id(cl));
750                 client_reingest(cl);
751             }
752             else if (r)
753             {
754                 session_log(se, YLOG_LOG, "client NEW %s", client_get_id(cl));
755                 client_start_search(cl);
756             }
757             no_working++;
758         }
759     }
760     facet_limits_destroy(facet_limits);
761     session_reset_active_clients(se, l0);
762
763     if (no_working == 0)
764     {
765         if (no_failed_query > 0)
766         {
767             *addinfo = "query";
768             return PAZPAR2_MALFORMED_PARAMETER_VALUE;
769         }
770         else if (no_failed_limit > 0)
771         {
772             *addinfo = "limit";
773             return PAZPAR2_MALFORMED_PARAMETER_VALUE;
774         }
775         else
776             return PAZPAR2_NO_TARGETS;
777     }
778     yaz_log(YLOG_LOG, "session_start_search done");
779     return PAZPAR2_NO_ERROR;
780 }
781
782 // Creates a new session_database object for a database
783 static void session_init_databases_fun(void *context, struct database *db)
784 {
785     struct session *se = (struct session *) context;
786     struct session_database *new = nmem_malloc(se->session_nmem, sizeof(*new));
787     int i;
788
789     new->database = db;
790     
791     new->map = 0;
792     assert(db->settings);
793     new->settings = nmem_malloc(se->session_nmem,
794                                 sizeof(struct settings *) * db->num_settings);
795     new->num_settings = db->num_settings;
796     for (i = 0; i < db->num_settings; i++)
797     {
798         struct setting *setting = db->settings[i];
799         new->settings[i] = setting;
800     }
801     new->next = se->databases;
802     se->databases = new;
803 }
804
805 // Doesn't free memory associated with sdb -- nmem takes care of that
806 static void session_database_destroy(struct session_database *sdb)
807 {
808     sdb->map = 0;
809 }
810
811 // Initialize session_database list -- this represents this session's view
812 // of the database list -- subject to modification by the settings ws command
813 void session_init_databases(struct session *se)
814 {
815     se->databases = 0;
816     predef_grep_databases(se, se->service, session_init_databases_fun);
817 }
818
819 // Probably session_init_databases_fun should be refactored instead of
820 // called here.
821 static struct session_database *load_session_database(struct session *se, 
822                                                       char *id)
823 {
824     struct database *db = new_database(id, se->session_nmem);
825
826     session_init_databases_fun((void*) se, db);
827
828     // New sdb is head of se->databases list
829     return se->databases;
830 }
831
832 // Find an existing session database. If not found, load it
833 static struct session_database *find_session_database(struct session *se, 
834                                                       char *id)
835 {
836     struct session_database *sdb;
837
838     for (sdb = se->databases; sdb; sdb = sdb->next)
839         if (!strcmp(sdb->database->id, id))
840             return sdb;
841     return load_session_database(se, id);
842 }
843
844 // Apply a session override to a database
845 void session_apply_setting(struct session *se, char *dbname, char *setting,
846                            char *value)
847 {
848     struct session_database *sdb = find_session_database(se, dbname);
849     struct conf_service *service = se->service;
850     struct setting *new = nmem_malloc(se->session_nmem, sizeof(*new));
851     int offset = settings_create_offset(service, setting);
852
853     expand_settings_array(&sdb->settings, &sdb->num_settings, offset,
854                           se->session_nmem);
855     new->precedence = 0;
856     new->target = dbname;
857     new->name = setting;
858     new->value = value;
859     new->next = sdb->settings[offset];
860     sdb->settings[offset] = new;
861
862     se->settings_modified = 1;
863
864     // Force later recompute of settings-driven data structures
865     // (happens when a search starts and client connections are prepared)
866     switch (offset)
867     {
868     case PZ_XSLT:
869         if (sdb->map)
870         {
871             sdb->map = 0;
872         }
873         break;
874     }
875 }
876
877 void session_destroy(struct session *se)
878 {
879     struct session_database *sdb;
880     session_log(se, YLOG_DEBUG, "Destroying");
881     session_use(-1);
882     session_remove_cached_clients(se);
883
884     for (sdb = se->databases; sdb; sdb = sdb->next)
885         session_database_destroy(sdb);
886     normalize_cache_destroy(se->normalize_cache);
887     relevance_destroy(&se->relevance);
888     reclist_destroy(se->reclist);
889     nmem_destroy(se->nmem);
890     service_destroy(se->service);
891     yaz_mutex_destroy(&se->session_mutex);
892 }
893
894 size_t session_get_memory_status(struct session *session) {
895     size_t session_nmem;
896     if (session == 0)
897         return 0;
898     session_enter(session);
899     session_nmem = nmem_total(session->nmem);
900     session_leave(session);
901     return session_nmem;
902 }
903
904
905 struct session *new_session(NMEM nmem, struct conf_service *service,
906                             unsigned session_id)
907 {
908     int i;
909     struct session *session = nmem_malloc(nmem, sizeof(*session));
910
911     char tmp_str[50];
912
913     sprintf(tmp_str, "session#%u", session_id);
914
915     session->session_id = session_id;
916     session_log(session, YLOG_DEBUG, "New");
917     session->service = service;
918     session->relevance = 0;
919     session->total_records = 0;
920     session->number_of_warnings_unknown_elements = 0;
921     session->number_of_warnings_unknown_metadata = 0;
922     session->num_termlists = 0;
923     session->reclist = 0;
924     session->clients_active = 0;
925     session->clients_cached = 0;
926     session->settings_modified = 0;
927     session->session_nmem = nmem;
928     session->nmem = nmem_create();
929     session->databases = 0;
930     for (i = 0; i <= SESSION_WATCH_MAX; i++)
931     {
932         session->watchlist[i].data = 0;
933         session->watchlist[i].fun = 0;
934     }
935     session->normalize_cache = normalize_cache_create();
936     session->session_mutex = 0;
937     pazpar2_mutex_create(&session->session_mutex, tmp_str);
938     session_use(1);
939     return session;
940 }
941
942 const char * client_get_suggestions_xml(struct client *cl, WRBUF wrbuf);
943
944 static struct hitsbytarget *hitsbytarget_nb(struct session *se,
945                                             int *count, NMEM nmem)
946 {
947     struct hitsbytarget *res = 0;
948     struct client_list *l;
949     size_t sz = 0;
950
951     for (l = se->clients_active; l; l = l->next)
952         sz++;
953
954     res = nmem_malloc(nmem, sizeof(*res) * sz);
955     *count = 0;
956     for (l = se->clients_active; l; l = l->next)
957     {
958         struct client *cl = l->client;
959         WRBUF w = wrbuf_alloc();
960         const char *name = session_setting_oneval(client_get_database(cl),
961                                                   PZ_NAME);
962
963         res[*count].id = client_get_id(cl);
964         res[*count].name = *name ? name : "Unknown";
965         res[*count].hits = client_get_hits(cl);
966         res[*count].approximation = client_get_approximation(cl);
967         res[*count].records = client_get_num_records(cl);
968         res[*count].filtered = client_get_num_records_filtered(cl);
969         res[*count].diagnostic =
970             client_get_diagnostic(cl, &res[*count].addinfo);
971         res[*count].state = client_get_state_str(cl);
972         res[*count].connected  = client_get_connection(cl) ? 1 : 0;
973         session_settings_dump(se, client_get_database(cl), w);
974         res[*count].settings_xml = nmem_strdup(nmem, wrbuf_cstr(w));
975         wrbuf_rewind(w);
976         wrbuf_puts(w, "");
977         res[*count].suggestions_xml = nmem_strdup(nmem, client_get_suggestions_xml(cl, w));
978         wrbuf_destroy(w);
979         (*count)++;
980     }
981     return res;
982 }
983
984 struct hitsbytarget *get_hitsbytarget(struct session *se, int *count, NMEM nmem)
985 {
986     struct hitsbytarget *p;
987     session_enter(se);
988     p = hitsbytarget_nb(se, count, nmem);
989     session_leave(se);
990     return p;
991 }
992     
993 struct termlist_score **get_termlist_score(struct session *se,
994                                            const char *name, int *num)
995 {
996     int i;
997     struct termlist_score **tl = 0;
998
999     session_enter(se);
1000     for (i = 0; i < se->num_termlists; i++)
1001         if (!strcmp((const char *) se->termlists[i].name, name))
1002         {
1003             tl = termlist_highscore(se->termlists[i].termlist, num);
1004             break;
1005         }
1006     session_leave(se);
1007     return tl;
1008 }
1009
1010 // Compares two hitsbytarget nodes by hitcount
1011 static int cmp_ht(const void *p1, const void *p2)
1012 {
1013     const struct hitsbytarget *h1 = p1;
1014     const struct hitsbytarget *h2 = p2;
1015     return h2->hits - h1->hits;
1016 }
1017
1018 static int targets_termlist_nb(WRBUF wrbuf, struct session *se, int num,
1019                                NMEM nmem, int version)
1020 {
1021     struct hitsbytarget *ht;
1022     int count, i;
1023
1024     ht = hitsbytarget_nb(se, &count, nmem);
1025     qsort(ht, count, sizeof(struct hitsbytarget), cmp_ht);
1026     for (i = 0; i < count && i < num && ht[i].hits > 0; i++)
1027     {
1028
1029         // do only print terms which have display names
1030     
1031         wrbuf_puts(wrbuf, "<term>\n");
1032
1033         wrbuf_puts(wrbuf, "<id>");
1034         wrbuf_xmlputs(wrbuf, ht[i].id);
1035         wrbuf_puts(wrbuf, "</id>\n");
1036         
1037         wrbuf_puts(wrbuf, "<name>");
1038         if (!ht[i].name || !ht[i].name[0])
1039             wrbuf_xmlputs(wrbuf, "NO TARGET NAME");
1040         else
1041             wrbuf_xmlputs(wrbuf, ht[i].name);
1042         wrbuf_puts(wrbuf, "</name>\n");
1043         
1044         wrbuf_printf(wrbuf, "<frequency>" ODR_INT_PRINTF "</frequency>\n",
1045                      ht[i].hits);
1046
1047         if (version >= 2) {
1048             // Should not print if we know it isn't a approximation.
1049             wrbuf_printf(wrbuf, "<approximation>" ODR_INT_PRINTF "</approximation>\n", ht[i].approximation);
1050             wrbuf_printf(wrbuf, "<records>%d</records>\n", ht[i].records - ht[i].filtered);
1051             wrbuf_printf(wrbuf, "<filtered>%d</filtered>\n", ht[i].filtered);
1052         }
1053
1054         wrbuf_puts(wrbuf, "<state>");
1055         wrbuf_xmlputs(wrbuf, ht[i].state);
1056         wrbuf_puts(wrbuf, "</state>\n");
1057         
1058         wrbuf_printf(wrbuf, "<diagnostic>%d</diagnostic>\n", 
1059                      ht[i].diagnostic);
1060         wrbuf_puts(wrbuf, "</term>\n");
1061     }
1062     return count;
1063 }
1064
1065 void perform_termlist(struct http_channel *c, struct session *se,
1066                       const char *name, int num, int version)
1067 {
1068     int i, j;
1069     NMEM nmem_tmp = nmem_create();
1070     char **names;
1071     int num_names = 0;
1072
1073     if (!name)
1074         name = "*";
1075
1076     nmem_strsplit(nmem_tmp, ",", name, &names, &num_names);
1077
1078     session_enter(se);
1079
1080     for (j = 0; j < num_names; j++)
1081     {
1082         const char *tname;
1083         int must_generate_empty = 1; /* bug 5350 */
1084
1085         for (i = 0; i < se->num_termlists; i++)
1086         {
1087             tname = se->termlists[i].name;
1088             if (!strcmp(names[j], tname) || !strcmp(names[j], "*"))
1089             {
1090                 struct termlist_score **p = 0;
1091                 int len;
1092
1093                 wrbuf_puts(c->wrbuf, "<list name=\"");
1094                 wrbuf_xmlputs(c->wrbuf, tname);
1095                 wrbuf_puts(c->wrbuf, "\">\n");
1096                 must_generate_empty = 0;
1097
1098                 p = termlist_highscore(se->termlists[i].termlist, &len);
1099                 if (p)
1100                 {
1101                     int i;
1102                     for (i = 0; i < len && i < num; i++)
1103                     {
1104                         // prevent sending empty term elements
1105                         if (!p[i]->display_term || !p[i]->display_term[0])
1106                             continue;
1107                         
1108                         wrbuf_puts(c->wrbuf, "<term>");
1109                         wrbuf_puts(c->wrbuf, "<name>");
1110                         wrbuf_xmlputs(c->wrbuf, p[i]->display_term);
1111                         wrbuf_puts(c->wrbuf, "</name>");
1112                         
1113                         wrbuf_printf(c->wrbuf, 
1114                                      "<frequency>%d</frequency>", 
1115                                      p[i]->frequency);
1116                         wrbuf_puts(c->wrbuf, "</term>\n");
1117                     }
1118                 }
1119                 wrbuf_puts(c->wrbuf, "</list>\n");
1120             }
1121         }
1122         tname = "xtargets";
1123         if (!strcmp(names[j], tname) || !strcmp(names[j], "*"))
1124         {
1125             wrbuf_puts(c->wrbuf, "<list name=\"");
1126             wrbuf_xmlputs(c->wrbuf, tname);
1127             wrbuf_puts(c->wrbuf, "\">\n");
1128
1129             targets_termlist_nb(c->wrbuf, se, num, c->nmem, version);
1130             wrbuf_puts(c->wrbuf, "</list>\n");
1131             must_generate_empty = 0;
1132         }
1133         if (must_generate_empty)
1134         {
1135             wrbuf_puts(c->wrbuf, "<list name=\"");
1136             wrbuf_xmlputs(c->wrbuf, names[j]);
1137             wrbuf_puts(c->wrbuf, "\"/>\n");
1138         }
1139     }
1140     session_leave(se);
1141     nmem_destroy(nmem_tmp);
1142 }
1143
1144 #ifdef MISSING_HEADERS
1145 void report_nmem_stats(void)
1146 {
1147     size_t in_use, is_free;
1148
1149     nmem_get_memory_in_use(&in_use);
1150     nmem_get_memory_free(&is_free);
1151
1152     yaz_log(YLOG_LOG, "nmem stat: use=%ld free=%ld", 
1153             (long) in_use, (long) is_free);
1154 }
1155 #endif
1156
1157 struct record_cluster *show_single_start(struct session *se, const char *id,
1158                                          struct record_cluster **prev_r,
1159                                          struct record_cluster **next_r)
1160 {
1161     struct record_cluster *r = 0;
1162
1163     session_enter(se);
1164     *prev_r = 0;
1165     *next_r = 0;
1166     if (se->reclist)
1167     {
1168         reclist_enter(se->reclist);
1169         while ((r = reclist_read_record(se->reclist)))
1170         {
1171             if (!strcmp(r->recid, id))
1172             {
1173                 *next_r = reclist_read_record(se->reclist);
1174                 break;
1175             }
1176             *prev_r = r;
1177         }
1178         reclist_leave(se->reclist);
1179     }
1180     if (!r)
1181         session_leave(se);
1182     return r;
1183 }
1184
1185 void show_single_stop(struct session *se, struct record_cluster *rec)
1186 {
1187     session_leave(se);
1188 }
1189
1190 struct record_cluster **show_range_start(struct session *se,
1191                                          struct reclist_sortparms *sp, 
1192                                          int start, int *num, int *total, Odr_int *sumhits, Odr_int *approx_hits)
1193 {
1194     struct record_cluster **recs;
1195     struct reclist_sortparms *spp;
1196     int i;
1197 #if USE_TIMING    
1198     yaz_timing_t t = yaz_timing_create();
1199 #endif
1200     session_enter(se);
1201     recs = nmem_malloc(se->nmem, *num * sizeof(struct record_cluster *));
1202     if (!se->relevance)
1203     {
1204         *num = 0;
1205         *total = 0;
1206         *sumhits = 0;
1207         recs = 0;
1208     }
1209     else
1210     {
1211         struct client_list *l;
1212         
1213         for (spp = sp; spp; spp = spp->next)
1214             if (spp->type == Metadata_sortkey_relevance)
1215             {
1216                 relevance_prepare_read(se->relevance, se->reclist);
1217                 break;
1218             }
1219         reclist_sort(se->reclist, sp);
1220         
1221         reclist_enter(se->reclist);
1222         *total = reclist_get_num_records(se->reclist);
1223
1224         *sumhits = 0;
1225         for (l = se->clients_active; l; l = l->next) {
1226             *sumhits += client_get_hits(l->client);
1227             *approx_hits += client_get_approximation(l->client);
1228         }
1229         for (i = 0; i < start; i++)
1230             if (!reclist_read_record(se->reclist))
1231             {
1232                 *num = 0;
1233                 recs = 0;
1234                 break;
1235             }
1236         
1237         for (i = 0; i < *num; i++)
1238         {
1239             struct record_cluster *r = reclist_read_record(se->reclist);
1240             if (!r)
1241             {
1242                 *num = i;
1243                 break;
1244             }
1245             recs[i] = r;
1246         }
1247         reclist_leave(se->reclist);
1248     }
1249 #if USE_TIMING
1250     yaz_timing_stop(t);
1251     yaz_log(YLOG_LOG, "show %6.5f %3.2f %3.2f", 
1252             yaz_timing_get_real(t), yaz_timing_get_user(t),
1253             yaz_timing_get_sys(t));
1254     yaz_timing_destroy(&t);
1255 #endif
1256     return recs;
1257 }
1258
1259 void show_range_stop(struct session *se, struct record_cluster **recs)
1260 {
1261     session_leave(se);
1262 }
1263
1264 void statistics(struct session *se, struct statistics *stat)
1265 {
1266     struct client_list *l;
1267     int count = 0;
1268
1269     memset(stat, 0, sizeof(*stat));
1270     stat->num_hits = 0;
1271     for (l = se->clients_active; l; l = l->next)
1272     {
1273         struct client *cl = l->client;
1274         if (!client_get_connection(cl))
1275             stat->num_no_connection++;
1276         stat->num_hits += client_get_hits(cl);
1277         switch (client_get_state(cl))
1278         {
1279         case Client_Connecting: stat->num_connecting++; break;
1280         case Client_Working: stat->num_working++; break;
1281         case Client_Idle: stat->num_idle++; break;
1282         case Client_Failed: stat->num_failed++; break;
1283         case Client_Error: stat->num_error++; break;
1284         default: break;
1285         }
1286         count++;
1287     }
1288     stat->num_records = se->total_records;
1289
1290     stat->num_clients = count;
1291 }
1292
1293 static struct record_metadata *record_metadata_init(
1294     NMEM nmem, const char *value, enum conf_metadata_type type,
1295     struct _xmlAttr *attr)
1296 {
1297     struct record_metadata *rec_md = record_metadata_create(nmem);
1298     struct record_metadata_attr **attrp = &rec_md->attributes;
1299     
1300     for (; attr; attr = attr->next)
1301     {
1302         if (attr->children && attr->children->content)
1303         {
1304             if (strcmp((const char *) attr->name, "type"))
1305             {  /* skip the "type" attribute.. Its value is already part of
1306                   the element in output (md-%s) and so repeating it here
1307                   is redundant */
1308                 *attrp = nmem_malloc(nmem, sizeof(**attrp));
1309                 (*attrp)->name =
1310                     nmem_strdup(nmem, (const char *) attr->name);
1311                 (*attrp)->value =
1312                     nmem_strdup(nmem, (const char *) attr->children->content);
1313                 attrp = &(*attrp)->next;
1314             }
1315         }
1316     }
1317     *attrp = 0;
1318
1319     if (type == Metadata_type_generic)
1320     {
1321         char *p = nmem_strdup(nmem, value);
1322
1323         p = normalize7bit_generic(p, " ,/.:([");
1324         
1325         rec_md->data.text.disp = p;
1326         rec_md->data.text.sort = 0;
1327     }
1328     else if (type == Metadata_type_year || type == Metadata_type_date)
1329     {
1330         int first, last;
1331         int longdate = 0;
1332
1333         if (type == Metadata_type_date)
1334             longdate = 1;
1335         if (extract7bit_dates((char *) value, &first, &last, longdate) < 0)
1336             return 0;
1337
1338         rec_md->data.number.min = first;
1339         rec_md->data.number.max = last;
1340     }
1341     else
1342         return 0;
1343     return rec_md;
1344 }
1345
1346 static int get_mergekey_from_doc(xmlDoc *doc, xmlNode *root, const char *name,
1347                                  struct conf_service *service, WRBUF norm_wr)
1348 {
1349     xmlNode *n;
1350     int no_found = 0;
1351     for (n = root->children; n; n = n->next)
1352     {
1353         if (n->type != XML_ELEMENT_NODE)
1354             continue;
1355         if (!strcmp((const char *) n->name, "metadata"))
1356         {
1357             xmlChar *type = xmlGetProp(n, (xmlChar *) "type");
1358             if (type == NULL) {
1359                 yaz_log(YLOG_FATAL, "Missing type attribute on metadata element. Skipping!");
1360             }
1361             else if (!strcmp(name, (const char *) type))
1362             {
1363                 xmlChar *value = xmlNodeListGetString(doc, n->children, 1);
1364                 if (value)
1365                 {
1366                     const char *norm_str;
1367                     pp2_charset_token_t prt =
1368                         pp2_charset_token_create(service->charsets, "mergekey");
1369                     
1370                     pp2_charset_token_first(prt, (const char *) value, 0);
1371                     if (wrbuf_len(norm_wr) > 0)
1372                         wrbuf_puts(norm_wr, " ");
1373                     wrbuf_puts(norm_wr, name);
1374                     while ((norm_str =
1375                             pp2_charset_token_next(prt)))
1376                     {
1377                         if (*norm_str)
1378                         {
1379                             wrbuf_puts(norm_wr, " ");
1380                             wrbuf_puts(norm_wr, norm_str);
1381                         }
1382                     }
1383                     xmlFree(value);
1384                     pp2_charset_token_destroy(prt);
1385                     no_found++;
1386                 }
1387             }
1388             xmlFree(type);
1389         }
1390     }
1391     return no_found;
1392 }
1393
1394 static const char *get_mergekey(xmlDoc *doc, struct client *cl, int record_no,
1395                                 struct conf_service *service, NMEM nmem)
1396 {
1397     char *mergekey_norm = 0;
1398     xmlNode *root = xmlDocGetRootElement(doc);
1399     WRBUF norm_wr = wrbuf_alloc();
1400
1401     /* consider mergekey from XSL first */
1402     xmlChar *mergekey = xmlGetProp(root, (xmlChar *) "mergekey");
1403     if (mergekey)
1404     {
1405         const char *norm_str;
1406         pp2_charset_token_t prt =
1407             pp2_charset_token_create(service->charsets, "mergekey");
1408
1409         pp2_charset_token_first(prt, (const char *) mergekey, 0);
1410         while ((norm_str = pp2_charset_token_next(prt)))
1411         {
1412             if (*norm_str)
1413             {
1414                 if (wrbuf_len(norm_wr))
1415                     wrbuf_puts(norm_wr, " ");
1416                 wrbuf_puts(norm_wr, norm_str);
1417             }
1418         }
1419         pp2_charset_token_destroy(prt);
1420         xmlFree(mergekey);
1421     }
1422     else
1423     {
1424         /* no mergekey defined in XSL. Look for mergekey metadata instead */
1425         int field_id;
1426         for (field_id = 0; field_id < service->num_metadata; field_id++)
1427         {
1428             struct conf_metadata *ser_md = &service->metadata[field_id];
1429             if (ser_md->mergekey != Metadata_mergekey_no)
1430             {
1431                 int r = get_mergekey_from_doc(doc, root, ser_md->name,
1432                                               service, norm_wr);
1433                 if (r == 0 && ser_md->mergekey == Metadata_mergekey_required)
1434                 {
1435                     /* no mergekey on this one and it is required.. 
1436                        Generate unique key instead */
1437                     wrbuf_rewind(norm_wr);
1438                     break;
1439                 }
1440             }
1441         }
1442     }
1443
1444     /* generate unique key if none is not generated already or is empty */
1445     if (wrbuf_len(norm_wr) == 0)
1446     {
1447         wrbuf_printf(norm_wr, "position: %s-%d",
1448                      client_get_id(cl), record_no);
1449     }
1450     else
1451     {
1452         const char *lead = "content: ";
1453         wrbuf_insert(norm_wr, 0, lead, strlen(lead));
1454     }
1455     if (wrbuf_len(norm_wr) > 0)
1456         mergekey_norm = nmem_strdup(nmem, wrbuf_cstr(norm_wr));
1457     wrbuf_destroy(norm_wr);
1458     return mergekey_norm;
1459 }
1460
1461 /** \brief see if metadata for pz:recordfilter exists 
1462     \param root xml root element of normalized record
1463     \param sdb session database for client
1464     \retval 0 if there is no metadata for pz:recordfilter
1465     \retval 1 if there is metadata for pz:recordfilter
1466
1467     If there is no pz:recordfilter defined, this function returns 1
1468     as well.
1469 */
1470     
1471 static int check_record_filter(xmlNode *root, struct session_database *sdb)
1472 {
1473     int match = 0;
1474     xmlNode *n;
1475     const char *s;
1476     s = session_setting_oneval(sdb, PZ_RECORDFILTER);
1477
1478     if (!s || !*s)
1479         return 1;
1480
1481     for (n = root->children; n; n = n->next)
1482     {
1483         if (n->type != XML_ELEMENT_NODE)
1484             continue;
1485         if (!strcmp((const char *) n->name, "metadata"))
1486         {
1487             xmlChar *type = xmlGetProp(n, (xmlChar *) "type");
1488             if (type)
1489             {
1490                 size_t len;
1491                 int substring;
1492                 const char *eq;
1493
1494                 if ((eq = strchr(s, '=')))
1495                     substring = 0;
1496                 else if ((eq = strchr(s, '~')))
1497                     substring = 1;
1498                 if (eq)
1499                     len = eq - s;
1500                 else
1501                     len = strlen(s);
1502                 if (len == strlen((const char *)type) &&
1503                     !memcmp((const char *) type, s, len))
1504                 {
1505                     xmlChar *value = xmlNodeGetContent(n);
1506                     if (value && *value)
1507                     {
1508                         if (!eq ||
1509                             (substring && strstr((const char *) value, eq+1)) ||
1510                             (!substring && !strcmp((const char *) value, eq + 1)))
1511                             match = 1;
1512                     }
1513                     xmlFree(value);
1514                 }
1515                 xmlFree(type);
1516             }
1517         }
1518     }
1519     return match;
1520 }
1521
1522
1523 static int ingest_to_cluster(struct client *cl,
1524                              xmlDoc *xdoc,
1525                              xmlNode *root,
1526                              int record_no,
1527                              const char *mergekey_norm);
1528
1529 /** \brief ingest XML record
1530     \param cl client holds the result set for record
1531     \param rec record buffer (0 terminated)
1532     \param record_no record position (1, 2, ..)
1533     \param nmem working NMEM
1534     \retval 0 OK
1535     \retval -1 failure
1536     \retval -2 Filtered
1537 */
1538 int ingest_record(struct client *cl, const char *rec,
1539                   int record_no, NMEM nmem)
1540 {
1541     struct session *se = client_get_session(cl);
1542     int ret = 0;
1543     struct session_database *sdb = client_get_database(cl);
1544     struct conf_service *service = se->service;
1545     xmlDoc *xdoc = normalize_record(se, sdb, service, rec, nmem);
1546     xmlNode *root;
1547     const char *mergekey_norm;
1548     
1549     if (!xdoc)
1550         return -1;
1551     
1552     root = xmlDocGetRootElement(xdoc);
1553     
1554     if (!check_record_filter(root, sdb))
1555     {
1556         session_log(se, YLOG_LOG, "Filtered out record no %d from %s", record_no, sdb->database->id);
1557         xmlFreeDoc(xdoc);
1558         return -2;
1559     }
1560     
1561     mergekey_norm = get_mergekey(xdoc, cl, record_no, service, nmem);
1562     if (!mergekey_norm)
1563     {
1564         session_log(se, YLOG_WARN, "Got no mergekey");
1565         xmlFreeDoc(xdoc);
1566         return -1;
1567     }
1568     session_enter(se);
1569     if (client_get_session(cl) == se)
1570         ret = ingest_to_cluster(cl, xdoc, root, record_no, mergekey_norm);
1571     session_leave(se);
1572     
1573     xmlFreeDoc(xdoc);
1574     return ret;
1575 }
1576
1577 static int check_limit_local(struct client *cl,
1578                              struct record *record,
1579                              int record_no)
1580 {
1581     int skip_record = 0;
1582     struct session *se = client_get_session(cl);
1583     struct conf_service *service = se->service;
1584     NMEM nmem_tmp = nmem_create();
1585     struct session_database *sdb = client_get_database(cl);
1586     int l = 0;
1587     while (!skip_record)
1588     {
1589         struct conf_metadata *ser_md = 0;
1590         struct record_metadata *rec_md = 0;
1591         int md_field_id;
1592         char **values = 0;
1593         int i, num_v = 0;
1594         
1595         const char *name =
1596             client_get_facet_limit_local(cl, sdb, &l, nmem_tmp, &num_v,
1597                                          &values);
1598         if (!name)
1599             break;
1600         
1601         md_field_id = conf_service_metadata_field_id(service, name);
1602         if (md_field_id < 0)
1603         {
1604             skip_record = 1;
1605             break;
1606         }
1607         ser_md = &service->metadata[md_field_id];
1608         rec_md = record->metadata[md_field_id];
1609         yaz_log(YLOG_LOG, "check limit local %s", name);
1610         for (i = 0; i < num_v; )
1611         {
1612             if (rec_md)
1613             {
1614                 if (ser_md->type == Metadata_type_year 
1615                     || ser_md->type == Metadata_type_date)
1616                 {
1617                     int y = atoi(values[i]);
1618                     if (y >= rec_md->data.number.min 
1619                         && y <= rec_md->data.number.max)
1620                         break;
1621                 }
1622                 else
1623                 {
1624                     yaz_log(YLOG_LOG, "cmp: '%s' '%s'",
1625                             rec_md->data.text.disp, values[i]);
1626                     if (!strcmp(rec_md->data.text.disp, values[i]))
1627                     {
1628                         break;
1629                     }
1630                 }
1631                 rec_md = rec_md->next;
1632             }
1633             else
1634             {
1635                 rec_md = record->metadata[md_field_id];
1636                 i++;
1637             }
1638         }
1639         if (i == num_v)
1640         {
1641             skip_record = 1;
1642             break;
1643         }
1644     }
1645     nmem_destroy(nmem_tmp);
1646     return skip_record;
1647 }
1648                              
1649 static int ingest_to_cluster(struct client *cl,
1650                              xmlDoc *xdoc,
1651                              xmlNode *root,
1652                              int record_no,
1653                              const char *mergekey_norm)
1654 {
1655     xmlNode *n;
1656     xmlChar *type = 0;
1657     xmlChar *value = 0;
1658     struct session *se = client_get_session(cl);
1659     struct conf_service *service = se->service;
1660     int term_factor = 1;
1661     struct record_cluster *cluster;
1662     struct session_database *sdb = client_get_database(cl);
1663     struct record *record = record_create(se->nmem, 
1664                                           service->num_metadata,
1665                                           service->num_sortkeys, cl,
1666                                           record_no);
1667
1668     for (n = root->children; n; n = n->next)
1669     {
1670         if (type)
1671             xmlFree(type);
1672         if (value)
1673             xmlFree(value);
1674         type = value = 0;
1675         
1676         if (n->type != XML_ELEMENT_NODE)
1677             continue;
1678         if (!strcmp((const char *) n->name, "metadata"))
1679         {
1680             struct conf_metadata *ser_md = 0;
1681             struct record_metadata **wheretoput = 0;
1682             struct record_metadata *rec_md = 0;
1683             int md_field_id = -1;
1684             
1685             type = xmlGetProp(n, (xmlChar *) "type");
1686             value = xmlNodeListGetString(xdoc, n->children, 1);
1687             
1688             if (!type || !value || !*value)
1689                 continue;
1690             
1691             md_field_id 
1692                 = conf_service_metadata_field_id(service, (const char *) type);
1693             if (md_field_id < 0)
1694             {
1695                 if (se->number_of_warnings_unknown_metadata == 0)
1696                 {
1697                     session_log(se, YLOG_WARN, 
1698                             "Ignoring unknown metadata element: %s", type);
1699                 }
1700                 se->number_of_warnings_unknown_metadata++;
1701                 continue;
1702             }
1703            
1704             ser_md = &service->metadata[md_field_id];
1705
1706             // non-merged metadata
1707             rec_md = record_metadata_init(se->nmem, (const char *) value,
1708                                           ser_md->type, n->properties);
1709             if (!rec_md)
1710             {
1711                 session_log(se, YLOG_WARN, "bad metadata data '%s' "
1712                             "for element '%s'", value, type);
1713                 continue;
1714             }
1715             wheretoput = &record->metadata[md_field_id];
1716             while (*wheretoput)
1717                 wheretoput = &(*wheretoput)->next;
1718             *wheretoput = rec_md;
1719         }
1720     }
1721
1722     if (check_limit_local(cl, record, record_no))
1723     {
1724         session_log(se, YLOG_LOG, "Facet filtered out record no %d from %s",
1725                     record_no, sdb->database->id);
1726         if (type)
1727             xmlFree(type);
1728         if (value)
1729             xmlFree(value);
1730         return -2;
1731     }
1732     cluster = reclist_insert(se->reclist, service, record,
1733                              mergekey_norm, &se->total_merged);
1734     if (!cluster)
1735         return -1;
1736
1737     {
1738         const char *use_term_factor_str =
1739             session_setting_oneval(sdb, PZ_TERMLIST_TERM_FACTOR);
1740         if (use_term_factor_str && use_term_factor_str[0] == '1')
1741         {
1742             int maxrecs = client_get_maxrecs(cl);
1743             int hits = (int) client_get_hits(cl);
1744             term_factor = MAX(hits, maxrecs) /  MAX(1, maxrecs);
1745             assert(term_factor >= 1);
1746             yaz_log(YLOG_DEBUG, "Using term factor: %d (%d / %d)", term_factor, MAX(hits, maxrecs), MAX(1, maxrecs));
1747         }
1748     }
1749
1750     if (global_parameters.dump_records)
1751         session_log(se, YLOG_LOG, "Cluster id %s from %s (#%d)", cluster->recid,
1752                     sdb->database->id, record_no);
1753
1754
1755     relevance_newrec(se->relevance, cluster);
1756     
1757     // now parsing XML record and adding data to cluster or record metadata
1758     for (n = root->children; n; n = n->next)
1759     {
1760         pp2_charset_token_t prt;
1761         if (type)
1762             xmlFree(type);
1763         if (value)
1764             xmlFree(value);
1765         type = value = 0;
1766         
1767         if (n->type != XML_ELEMENT_NODE)
1768             continue;
1769         if (!strcmp((const char *) n->name, "metadata"))
1770         {
1771             struct conf_metadata *ser_md = 0;
1772             struct conf_sortkey *ser_sk = 0;
1773             struct record_metadata **wheretoput = 0;
1774             struct record_metadata *rec_md = 0;
1775             int md_field_id = -1;
1776             int sk_field_id = -1;
1777             
1778             type = xmlGetProp(n, (xmlChar *) "type");
1779             value = xmlNodeListGetString(xdoc, n->children, 1);
1780             
1781             if (!type || !value || !*value)
1782                 continue;
1783             
1784             md_field_id 
1785                 = conf_service_metadata_field_id(service, (const char *) type);
1786             if (md_field_id < 0)
1787                 continue;
1788             
1789             ser_md = &service->metadata[md_field_id];
1790             
1791             if (ser_md->sortkey_offset >= 0)
1792             {
1793                 sk_field_id = ser_md->sortkey_offset;
1794                 ser_sk = &service->sortkeys[sk_field_id];
1795             }
1796
1797             // merged metadata
1798             rec_md = record_metadata_init(se->nmem, (const char *) value,
1799                                           ser_md->type, 0);
1800             if (!rec_md)
1801                 continue;
1802
1803             wheretoput = &cluster->metadata[md_field_id];
1804
1805             // and polulate with data:
1806             // assign cluster or record based on merge action
1807             if (ser_md->merge == Metadata_merge_unique)
1808             {
1809                 while (*wheretoput)
1810                 {
1811                     if (!strcmp((const char *) (*wheretoput)->data.text.disp, 
1812                                 rec_md->data.text.disp))
1813                         break;
1814                     wheretoput = &(*wheretoput)->next;
1815                 }
1816                 if (!*wheretoput)
1817                     *wheretoput = rec_md;
1818             }
1819             else if (ser_md->merge == Metadata_merge_longest)
1820             {
1821                 if (!*wheretoput 
1822                     || strlen(rec_md->data.text.disp) 
1823                     > strlen((*wheretoput)->data.text.disp))
1824                 {
1825                     *wheretoput = rec_md;
1826                     if (ser_sk)
1827                     {
1828                         const char *sort_str = 0;
1829                         int skip_article = 
1830                             ser_sk->type == Metadata_sortkey_skiparticle;
1831
1832                         if (!cluster->sortkeys[sk_field_id])
1833                             cluster->sortkeys[sk_field_id] = 
1834                                 nmem_malloc(se->nmem, 
1835                                             sizeof(union data_types));
1836                          
1837                         prt =
1838                             pp2_charset_token_create(service->charsets, "sort");
1839
1840                         pp2_charset_token_first(prt, rec_md->data.text.disp,
1841                                                 skip_article);
1842
1843                         pp2_charset_token_next(prt);
1844                          
1845                         sort_str = pp2_get_sort(prt);
1846                          
1847                         cluster->sortkeys[sk_field_id]->text.disp = 
1848                             rec_md->data.text.disp;
1849                         if (!sort_str)
1850                         {
1851                             sort_str = rec_md->data.text.disp;
1852                             session_log(se, YLOG_WARN, 
1853                                     "Could not make sortkey. Bug #1858");
1854                         }
1855                         cluster->sortkeys[sk_field_id]->text.sort = 
1856                             nmem_strdup(se->nmem, sort_str);
1857                         pp2_charset_token_destroy(prt);
1858                     }
1859                 }
1860             }
1861             else if (ser_md->merge == Metadata_merge_all)
1862             {
1863                 while (*wheretoput)
1864                     wheretoput = &(*wheretoput)->next;
1865                 *wheretoput = rec_md;
1866             }
1867             else if (ser_md->merge == Metadata_merge_range)
1868             {
1869                 if (!*wheretoput)
1870                 {
1871                     *wheretoput = rec_md;
1872                     if (ser_sk)
1873                         cluster->sortkeys[sk_field_id] 
1874                             = &rec_md->data;
1875                 }
1876                 else
1877                 {
1878                     int this_min = rec_md->data.number.min;
1879                     int this_max = rec_md->data.number.max;
1880                     if (this_min < (*wheretoput)->data.number.min)
1881                         (*wheretoput)->data.number.min = this_min;
1882                     if (this_max > (*wheretoput)->data.number.max)
1883                         (*wheretoput)->data.number.max = this_max;
1884                 }
1885             }
1886
1887
1888             // ranking of _all_ fields enabled ... 
1889             if (ser_md->rank)
1890                 relevance_countwords(se->relevance, cluster, 
1891                                      (char *) value, ser_md->rank,
1892                                      ser_md->name);
1893
1894             // construct facets ... unless the client already has reported them
1895             if (ser_md->termlist && !client_has_facet(cl, (char *) type))
1896             {
1897                 if (ser_md->type == Metadata_type_year)
1898                 {
1899                     char year[64];
1900                     sprintf(year, "%d", rec_md->data.number.max);
1901
1902                     add_facet(se, (char *) type, year, term_factor);
1903                     if (rec_md->data.number.max != rec_md->data.number.min)
1904                     {
1905                         sprintf(year, "%d", rec_md->data.number.min);
1906                         add_facet(se, (char *) type, year, term_factor);
1907                     }
1908                 }
1909                 else
1910                     add_facet(se, (char *) type, (char *) value, term_factor);
1911             }
1912
1913             // cleaning up
1914             xmlFree(type);
1915             xmlFree(value);
1916             type = value = 0;
1917         }
1918         else
1919         {
1920             if (se->number_of_warnings_unknown_elements == 0)
1921                 session_log(se, YLOG_WARN,
1922                         "Unexpected element in internal record: %s", n->name);
1923             se->number_of_warnings_unknown_elements++;
1924         }
1925     }
1926     if (type)
1927         xmlFree(type);
1928     if (value)
1929         xmlFree(value);
1930
1931     relevance_donerecord(se->relevance, cluster);
1932     se->total_records++;
1933
1934     return 0;
1935 }
1936
1937 void session_log(struct session *s, int level, const char *fmt, ...)
1938 {
1939     char buf[1024];
1940     va_list ap;
1941     va_start(ap, fmt);
1942
1943     yaz_vsnprintf(buf, sizeof(buf)-30, fmt, ap);
1944     yaz_log(level, "Session %u: %s", s->session_id, buf);
1945
1946     va_end(ap);
1947 }
1948
1949 /*
1950  * Local variables:
1951  * c-basic-offset: 4
1952  * c-file-style: "Stroustrup"
1953  * indent-tabs-mode: nil
1954  * End:
1955  * vim: shiftwidth=4 tabstop=8 expandtab
1956  */
1957