/* This file is part of Pazpar2.
- Copyright (C) 2006-2011 Index Data
+ Copyright (C) Index Data
Pazpar2 is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free
#include <yaz/querytowrbuf.h>
#include <yaz/oid_db.h>
#include <yaz/snprintf.h>
-#include <yaz/gettimeofday.h>
-#define USE_TIMING 0
+#define USE_TIMING 1
#if USE_TIMING
#include <yaz/timing.h>
#endif
#include "settings.h"
#include "normalize7bit.h"
-#define TERMLIST_HIGH_SCORE 25
+#include <libxml/tree.h>
#define MAX_CHUNK 15
#define MAX(a,b) ((a)>(b)?(a):(b))
// Note: Some things in this structure will eventually move to configuration
-struct parameters global_parameters =
+struct parameters global_parameters =
{
0, // dump_records
0, // debug_mode
/* session counting (1) , disable client counting (0) */
static YAZ_MUTEX g_session_mutex = 0;
static int no_sessions = 0;
-static int no_session_total = 0;
static int session_use(int delta)
{
yaz_mutex_create(&g_session_mutex);
yaz_mutex_enter(g_session_mutex);
no_sessions += delta;
- if (delta > 0)
- no_session_total += delta;
sessions = no_sessions;
yaz_mutex_leave(g_session_mutex);
- yaz_log(YLOG_DEBUG, "%s sessions=%d", delta == 0 ? "" : (delta > 0 ? "INC" : "DEC"), no_sessions);
+ yaz_log(YLOG_DEBUG, "%s sessions=%d", delta == 0 ? "" :
+ (delta > 0 ? "INC" : "DEC"), no_sessions);
return sessions;
}
return session_use(0);
}
-int session_count_total(void)
-{
- int total = 0;
- if (!g_session_mutex)
- return 0;
- yaz_mutex_enter(g_session_mutex);
- total = no_session_total;
- yaz_mutex_leave(g_session_mutex);
- return total;
-}
-
static void log_xml_doc(xmlDoc *doc)
{
FILE *lf = yaz_log_file();
xmlFree(result);
}
-static void session_enter(struct session *s)
+static void session_enter(struct session *s, const char *caller)
{
+ if (caller)
+ session_log(s, YLOG_DEBUG, "Session lock by %s", caller);
yaz_mutex_enter(s->session_mutex);
}
-static void session_leave(struct session *s)
+static void session_leave(struct session *s, const char *caller)
{
yaz_mutex_leave(s->session_mutex);
+ if (caller)
+ session_log(s, YLOG_DEBUG, "Session unlock by %s", caller);
}
-void add_facet(struct session *s, const char *type, const char *value, int count)
+static void session_normalize_facet(struct session *s,
+ const char *type, const char *value,
+ WRBUF display_wrbuf, WRBUF facet_wrbuf)
{
struct conf_service *service = s->service;
pp2_charset_token_t prt;
const char *facet_component;
- WRBUF facet_wrbuf = wrbuf_alloc();
- WRBUF display_wrbuf = wrbuf_alloc();
int i;
const char *icu_chain_id = 0;
prt = pp2_charset_token_create(service->charsets, icu_chain_id);
if (!prt)
{
- yaz_log(YLOG_FATAL, "Unknown ICU chain '%s' for facet of type '%s'",
+ session_log(s, YLOG_FATAL,
+ "Unknown ICU chain '%s' for facet of type '%s'",
icu_chain_id, type);
wrbuf_destroy(facet_wrbuf);
wrbuf_destroy(display_wrbuf);
}
}
pp2_charset_token_destroy(prt);
-
+}
+
+void add_facet(struct session *s, const char *type, const char *value, int count)
+{
+ WRBUF facet_wrbuf = wrbuf_alloc();
+ WRBUF display_wrbuf = wrbuf_alloc();
+
+ session_normalize_facet(s, type, value, display_wrbuf, facet_wrbuf);
+
if (wrbuf_len(facet_wrbuf))
{
int i;
wrbuf_destroy(display_wrbuf);
return;
}
-
+
s->termlists[i].name = nmem_strdup(s->nmem, type);
- s->termlists[i].termlist
- = termlist_create(s->nmem, TERMLIST_HIGH_SCORE);
+ s->termlists[i].termlist = termlist_create(s->nmem);
s->num_termlists = i + 1;
}
-
+
#if 0
- session_log(s, YLOG_DEBUG, "Facets for %s: %s norm:%s (%d)", type, value, wrbuf_cstr(facet_wrbuf), count);
+ session_log(s, YLOG_LOG, "Facets for %s: %s norm:%s (%d)", type, value, wrbuf_cstr(facet_wrbuf), count);
#endif
termlist_insert(s->termlists[i].termlist, wrbuf_cstr(display_wrbuf),
wrbuf_cstr(facet_wrbuf), count);
if (!rdoc)
{
- session_log(se, YLOG_FATAL, "Non-wellformed XML received from %s",
- db->id);
+ session_log(se, YLOG_WARN, "Non-wellformed XML");
return 0;
}
// Add static values from session database settings if applicable
static void insert_settings_values(struct session_database *sdb, xmlDoc *doc,
- struct conf_service *service)
+ xmlNode *root,
+ struct conf_service *service)
{
int i;
const char *val = session_setting_oneval(sdb, offset);
if (val)
{
- xmlNode *r = xmlDocGetRootElement(doc);
- xmlNode *n = xmlNewTextChild(r, 0, (xmlChar *) "metadata",
+ xmlNode *n = xmlNewTextChild(root, 0, (xmlChar *) "metadata",
(xmlChar *) val);
xmlSetProp(n, (xmlChar *) "type", (xmlChar *) md->name);
}
if (rdoc)
{
char *parms[MAX_XSLT_ARGS*2+1];
-
+
insert_settings_parameters(sdb, service, parms, nmem);
-
+
if (normalize_record_transform(sdb->map, &rdoc, (const char **)parms))
{
- session_log(se, YLOG_WARN, "Normalize failed from %s",
- sdb->database->id);
- }
- else
- {
- insert_settings_values(sdb, rdoc, service);
-
- if (global_parameters.dump_records)
- {
- session_log(se, YLOG_LOG, "Normalized record from %s",
- sdb->database->id);
- log_xml_doc(rdoc);
- }
+ session_log(se, YLOG_WARN, "Normalize failed");
}
}
return rdoc;
// setting. However, this is not a realistic use scenario.
static int prepare_map(struct session *se, struct session_database *sdb)
{
- const char *s;
-
- if (!sdb->settings)
+ if (sdb->settings && !sdb->map)
{
- session_log(se, YLOG_WARN, "No settings on %s", sdb->database->id);
- return -1;
- }
- if ((s = session_setting_oneval(sdb, PZ_XSLT)))
- {
- char auto_stylesheet[256];
+ const char *s;
- if (!strcmp(s, "auto"))
+ if (sdb->settings[PZ_XSLT] &&
+ (s = session_setting_oneval(sdb, PZ_XSLT)))
{
- const char *request_syntax = session_setting_oneval(
- sdb, PZ_REQUESTSYNTAX);
- if (request_syntax)
+ char auto_stylesheet[256];
+
+ if (!strcmp(s, "auto"))
{
- char *cp;
- yaz_snprintf(auto_stylesheet, sizeof(auto_stylesheet),
- "%s.xsl", request_syntax);
- for (cp = auto_stylesheet; *cp; cp++)
+ const char *request_syntax = session_setting_oneval(
+ sdb, PZ_REQUESTSYNTAX);
+ if (request_syntax)
{
- /* deliberately only consider ASCII */
- if (*cp > 32 && *cp < 127)
- *cp = tolower(*cp);
+ char *cp;
+ yaz_snprintf(auto_stylesheet, sizeof(auto_stylesheet),
+ "%s.xsl", request_syntax);
+ for (cp = auto_stylesheet; *cp; cp++)
+ {
+ /* deliberately only consider ASCII */
+ if (*cp > 32 && *cp < 127)
+ *cp = tolower(*cp);
+ }
+ s = auto_stylesheet;
+ }
+ else
+ {
+ session_log(se, YLOG_WARN,
+ "No pz:requestsyntax for auto stylesheet");
}
- s = auto_stylesheet;
- }
- else
- {
- session_log(se, YLOG_WARN,
- "No pz:requestsyntax for auto stylesheet");
}
+ sdb->map = normalize_cache_get(se->normalize_cache,
+ se->service, s);
+ if (!sdb->map)
+ return -1;
}
- sdb->map = normalize_cache_get(se->normalize_cache,
- se->service, s);
- if (!sdb->map)
- return -1;
- }
- return 0;
-}
-
-// This analyzes settings and recomputes any supporting data structures
-// if necessary.
-static int prepare_session_database(struct session *se,
- struct session_database *sdb)
-{
- if (!sdb->settings)
- {
- session_log(se, YLOG_WARN,
- "No settings associated with %s", sdb->database->id);
- return -1;
- }
- if (sdb->settings[PZ_XSLT] && !sdb->map)
- {
- if (prepare_map(se, sdb) < 0)
- return -1;
}
return 0;
}
}
// set watch. Returns 0=OK, -1 if watch is already set
-int session_set_watch(struct session *s, int what,
+int session_set_watch(struct session *s, int what,
session_watchfun fun, void *data,
struct http_channel *chan)
{
int ret;
- session_enter(s);
+ session_enter(s, "session_set_watch");
if (s->watchlist[what].fun)
ret = -1;
else
{
-
+
s->watchlist[what].fun = fun;
s->watchlist[what].data = data;
s->watchlist[what].obs = http_add_observer(chan, &s->watchlist[what],
session_watch_cancel);
ret = 0;
}
- session_leave(s);
+ session_leave(s, "session_set_watch");
return ret;
}
void session_alert_watch(struct session *s, int what)
{
assert(s);
- session_enter(s);
+ session_enter(s, "session_alert_watch");
if (s->watchlist[what].fun)
{
/* our watch is no longer associated with http_channel */
s->watchlist[what].data = 0;
s->watchlist[what].obs = 0;
- session_leave(s);
+ session_leave(s, "session_alert_watch");
session_log(s, YLOG_DEBUG,
"Alert Watch: %d calling function: %p", what, fun);
fun(data);
}
else
- session_leave(s);
+ session_leave(s,"session_alert_watch");
}
//callback for grep_databases
static void select_targets_callback(struct session *se,
struct session_database *db)
{
- struct client *cl = client_create(db->database->id);
+ struct client *cl;
struct client_list *l;
- client_set_database(cl, db);
+ for (l = se->clients_cached; l; l = l->next)
+ if (client_get_database(l->client) == db)
+ break;
+ if (l)
+ cl = l->client;
+ else
+ {
+ cl = client_create(db->database->id);
+ client_set_database(cl, db);
+
+ l = xmalloc(sizeof(*l));
+ l->client = cl;
+ l->next = se->clients_cached;
+ se->clients_cached = l;
+ }
+ /* set session always. If may be 0 if client is not active */
client_set_session(cl, se);
l = xmalloc(sizeof(*l));
l->client = cl;
- l->next = se->clients;
- se->clients = l;
+ l->next = se->clients_active;
+ se->clients_active = l;
}
-static void session_remove_clients(struct session *se)
+static void session_reset_active_clients(struct session *se,
+ struct client_list *new_list)
{
struct client_list *l;
- session_enter(se);
- l = se->clients;
- se->clients = 0;
- session_leave(se);
+ session_enter(se, "session_reset_active_clients");
+ l = se->clients_active;
+ se->clients_active = new_list;
+ session_leave(se, "session_reset_active_clients");
+
+ while (l)
+ {
+ struct client_list *l_next = l->next;
+
+ client_lock(l->client);
+ client_set_session(l->client, 0); /* mark client inactive */
+ client_unlock(l->client);
+
+ xfree(l);
+ l = l_next;
+ }
+}
+
+static void session_remove_cached_clients(struct session *se)
+{
+ struct client_list *l;
+
+ session_reset_active_clients(se, 0);
+
+ session_enter(se, "session_remove_cached_clients");
+ l = se->clients_cached;
+ se->clients_cached = 0;
+ session_leave(se, "session_remove_cached_clients");
while (l)
{
}
// Associates a set of clients with a session;
-// Note: Session-databases represent databases with per-session
+// Note: Session-databases represent databases with per-session
// setting overrides
static int select_targets(struct session *se, const char *filter)
{
struct client_list *l;
int res = 0;
- for (l = s->clients; l; l = l->next)
+ for (l = s->clients_active; l; l = l->next)
if (client_is_active(l->client))
res++;
struct client_list *l;
int res = 0;
- for (l = s->clients; l; l = l->next)
+ for (l = s->clients_active; l; l = l->next)
if (client_is_active_preferred(l->client))
res++;
session_log(s, YLOG_DEBUG, "Has %d active preferred clients.", res);
return res == 0;
}
-void session_sort(struct session *se, const char *field, int increasing)
+static void session_clear_set(struct session *se, struct reclist_sortparms *sp)
+{
+ reclist_destroy(se->reclist);
+ if (nmem_total(se->nmem))
+ session_log(se, YLOG_DEBUG, "NMEN operation usage %zd",
+ nmem_total(se->nmem));
+ nmem_reset(se->nmem);
+ se->total_records = se->total_merged = 0;
+ se->num_termlists = 0;
+
+ relevance_clear(se->relevance);
+
+ /* reset list of sorted results and clear to relevance search */
+ se->sorted_results = nmem_malloc(se->nmem, sizeof(*se->sorted_results));
+ se->sorted_results->name = nmem_strdup(se->nmem, sp->name);
+ se->sorted_results->increasing = sp->increasing;
+ se->sorted_results->type = sp->type;
+ se->sorted_results->next = 0;
+
+ session_log(se, YLOG_DEBUG, "clear_set session_sort: field=%s increasing=%d type=%d configured",
+ sp->name, sp->increasing, sp->type);
+
+ se->reclist = reclist_create(se->nmem);
+}
+
+static void session_sort_unlocked(struct session *se,
+ struct reclist_sortparms *sp,
+ const char *mergekey,
+ const char *rank)
{
- struct session_sorted_results *sr;
struct client_list *l;
+ const char *field = sp->name;
+ int increasing = sp->increasing;
+ int type = sp->type;
+ int clients_research = 0;
- session_enter(se);
+ session_log(se, YLOG_DEBUG, "session_sort field=%s increasing=%d type=%d",
+ field, increasing, type);
- /* see if we already have sorted for this critieria */
- for (sr = se->sorted_results; sr; sr = sr->next)
+ if (rank && (!se->rank || strcmp(se->rank, rank)))
{
- if (!strcmp(field, sr->field) && increasing == sr->increasing)
- break;
+ /* new rank must research/reingest anyway */
+ assert(rank);
+ xfree(se->rank);
+ se->rank = *rank ? xstrdup(rank) : 0;
+ clients_research = 1;
+ session_log(se, YLOG_DEBUG, "session_sort: new rank = %s",
+ rank);
}
- if (sr)
+ if (mergekey && (!se->mergekey || strcmp(se->mergekey, mergekey)))
{
- yaz_log(YLOG_LOG, "search_sort: field=%s increasing=%d already fetched",
- field, increasing);
- session_leave(se);
- return;
+ /* new mergekey must research/reingest anyway */
+ assert(mergekey);
+ xfree(se->mergekey);
+ se->mergekey = *mergekey ? xstrdup(mergekey) : 0;
+ clients_research = 1;
+ session_log(se, YLOG_DEBUG, "session_sort: new mergekey = %s",
+ mergekey);
+ }
+ if (clients_research == 0)
+ {
+ struct reclist_sortparms *sr;
+ for (sr = se->sorted_results; sr; sr = sr->next)
+ if (!reclist_sortparms_cmp(sr, sp))
+ break;
+ if (sr)
+ {
+ session_log(se, YLOG_DEBUG, "session_sort: field=%s increasing=%d type=%d already fetched",
+ field, increasing, type);
+ return;
+ }
+ }
+ session_log(se, YLOG_DEBUG, "session_sort: field=%s increasing=%d type=%d must fetch",
+ field, increasing, type);
+
+ // We need to reset reclist on every sort that changes the records, not just for position
+ // So if just one client requires new searching, we need to clear set.
+ // Ask each of the client if sorting requires re-search due to native sort
+ // If it does it will require us to
+ for (l = se->clients_active; l; l = l->next)
+ {
+ struct client *cl = l->client;
+ // Assume no re-search is required.
+ client_parse_init(cl, 1);
+ clients_research += client_parse_sort(cl, sp);
+ }
+ if (clients_research)
+ {
+ session_log(se, YLOG_DEBUG,
+ "session_sort: reset results due to %d clients researching",
+ clients_research);
+ session_clear_set(se, sp);
}
- yaz_log(YLOG_LOG, "search_sort: field=%s increasing=%d must fetch",
- field, increasing);
- sr = nmem_malloc(se->nmem, sizeof(*sr));
- sr->field = nmem_strdup(se->nmem, field);
- sr->increasing = increasing;
- sr->next = se->sorted_results;
- se->sorted_results = sr;
-
- for (l = se->clients; l; l = l->next)
+ else
+ {
+ // A new sorting based on same record set
+ struct reclist_sortparms *sr = nmem_malloc(se->nmem, sizeof(*sr));
+ sr->name = nmem_strdup(se->nmem, field);
+ sr->increasing = increasing;
+ sr->type = type;
+ sr->next = se->sorted_results;
+ se->sorted_results = sr;
+ session_log(se, YLOG_DEBUG, "session_sort: no research/ingesting done");
+ return ;
+ }
+ session_log(se, YLOG_DEBUG, "Re- search/ingesting for clients due to change in sort order");
+
+ for (l = se->clients_active; l; l = l->next)
{
struct client *cl = l->client;
- struct timeval tval;
- if (client_prep_connection(cl, se->service->z3950_operation_timeout,
- se->service->z3950_session_timeout,
- se->service->server->iochan_man,
- &tval))
+ if (client_get_state(cl) == Client_Connecting ||
+ client_get_state(cl) == Client_Idle ||
+ client_get_state(cl) == Client_Working) {
client_start_search(cl);
+ }
+ else
+ {
+ session_log(se, YLOG_DEBUG,
+ "session_sort: %s: No re-start/ingest in show. "
+ "Wrong client state: %d",
+ client_get_id(cl), client_get_state(cl));
+ }
}
- session_leave(se);
}
+void session_sort(struct session *se, struct reclist_sortparms *sp,
+ const char *mergekey, const char *rank)
+{
+ //session_enter(se, "session_sort");
+ session_sort_unlocked(se, sp, mergekey, rank);
+ //session_leave(se, "session_sort");
+}
+
+
enum pazpar2_error_code session_search(struct session *se,
const char *query,
const char *startrecs,
const char *filter,
const char *limit,
const char **addinfo,
- const char *sort_field, int increasing)
+ const char **addinfo2,
+ struct reclist_sortparms *sp,
+ const char *mergekey,
+ const char *rank)
{
int live_channels = 0;
int no_working = 0;
- int no_failed = 0;
- struct client_list *l;
- struct timeval tval;
- facet_limits_t facet_limits;
+ int no_failed_query = 0;
+ int no_failed_limit = 0;
+ struct client_list *l, *l0;
+
+ session_alert_watch(se, SESSION_WATCH_SHOW);
+ session_alert_watch(se, SESSION_WATCH_BYTARGET);
+ session_alert_watch(se, SESSION_WATCH_TERMLIST);
+ session_alert_watch(se, SESSION_WATCH_SHOW_PREF);
session_log(se, YLOG_DEBUG, "Search");
*addinfo = 0;
- session_remove_clients(se);
-
- session_enter(se);
- reclist_destroy(se->reclist);
- se->reclist = 0;
+ if (se->settings_modified) {
+ session_remove_cached_clients(se);
+ }
+ else
+ session_reset_active_clients(se, 0);
+
+ session_enter(se, "session_search");
+ se->settings_modified = 0;
+
+ if (mergekey)
+ {
+ xfree(se->mergekey);
+ se->mergekey = *mergekey ? xstrdup(mergekey) : 0;
+ }
+ if (rank)
+ {
+ xfree(se->rank);
+ se->rank = *rank ? xstrdup(rank) : 0;
+ }
+
+ session_clear_set(se, sp);
relevance_destroy(&se->relevance);
- nmem_reset(se->nmem);
- se->total_records = se->total_merged = 0;
- se->num_termlists = 0;
- /* reset list of sorted results and clear to relevance search */
- se->sorted_results = nmem_malloc(se->nmem, sizeof(*se->sorted_results));
- se->sorted_results->field = nmem_strdup(se->nmem, sort_field);
- se->sorted_results->increasing = increasing;
- se->sorted_results->next = 0;
-
live_channels = select_targets(se, filter);
if (!live_channels)
{
- session_leave(se);
+ session_leave(se, "session_search");
return PAZPAR2_NO_TARGETS;
}
- se->reclist = reclist_create(se->nmem);
-
- yaz_gettimeofday(&tval);
-
- tval.tv_sec += 5;
- facet_limits = facet_limits_create(limit);
- if (!facet_limits)
+ facet_limits_destroy(se->facet_limits);
+ se->facet_limits = facet_limits_create(limit);
+ if (!se->facet_limits)
{
*addinfo = "limit";
- session_leave(se);
+ session_leave(se, "session_search");
return PAZPAR2_MALFORMED_PARAMETER_VALUE;
}
- for (l = se->clients; l; l = l->next)
+
+ l0 = se->clients_active;
+ se->clients_active = 0;
+ session_leave(se, "session_search");
+
+ for (l = l0; l; l = l->next)
{
+ int parse_ret;
struct client *cl = l->client;
+ client_parse_init(cl, 1);
+ if (prepare_map(se, client_get_database(cl)) < 0)
+ continue;
- if (maxrecs)
- client_set_maxrecs(cl, atoi(maxrecs));
- if (startrecs)
- client_set_startrecs(cl, atoi(startrecs));
- if (prepare_session_database(se, client_get_database(cl)) < 0)
- ;
- else if (client_parse_query(cl, query, facet_limits) < 0)
- no_failed++;
+ parse_ret = client_parse_query(cl, query, se->facet_limits, addinfo2);
+ if (parse_ret == -1)
+ no_failed_query++;
+ else if (parse_ret == -2)
+ no_failed_limit++;
+ else if (parse_ret < 0)
+ no_working++; /* other error, such as bad CCL map */
else
{
+ client_parse_range(cl, startrecs, maxrecs);
+ client_parse_sort(cl, sp);
+ client_start_search(cl);
no_working++;
- if (client_prep_connection(cl, se->service->z3950_operation_timeout,
- se->service->z3950_session_timeout,
- se->service->server->iochan_man,
- &tval))
- client_start_search(cl);
}
}
- facet_limits_destroy(facet_limits);
- session_leave(se);
+ session_reset_active_clients(se, l0);
+
if (no_working == 0)
{
- if (no_failed > 0)
+ if (no_failed_query > 0)
{
*addinfo = "query";
return PAZPAR2_MALFORMED_PARAMETER_VALUE;
}
+ else if (no_failed_limit > 0)
+ {
+ *addinfo = "limit";
+ return PAZPAR2_MALFORMED_PARAMETER_VALUE;
+ }
else
return PAZPAR2_NO_TARGETS;
}
int i;
new->database = db;
-
+
new->map = 0;
assert(db->settings);
new->settings = nmem_malloc(se->session_nmem,
// Probably session_init_databases_fun should be refactored instead of
// called here.
-static struct session_database *load_session_database(struct session *se,
+static struct session_database *load_session_database(struct session *se,
char *id)
{
- struct database *db = new_database(id, se->session_nmem);
-
+ struct database *db = new_database_inherit_settings(id, se->session_nmem, se->service->settings);
session_init_databases_fun((void*) se, db);
// New sdb is head of se->databases list
}
// Find an existing session database. If not found, load it
-static struct session_database *find_session_database(struct session *se,
+static struct session_database *find_session_database(struct session *se,
char *id)
{
struct session_database *sdb;
new->next = sdb->settings[offset];
sdb->settings[offset] = new;
+ se->settings_modified = 1;
+
// Force later recompute of settings-driven data structures
// (happens when a search starts and client connections are prepared)
switch (offset)
}
}
-void session_destroy(struct session *se) {
+void session_destroy(struct session *se)
+{
struct session_database *sdb;
- session_log(se, YLOG_DEBUG, "Destroying");
+ session_log(se, YLOG_LOG, "destroy");
session_use(-1);
- session_remove_clients(se);
+ session_remove_cached_clients(se);
for (sdb = se->databases; sdb; sdb = sdb->next)
session_database_destroy(sdb);
normalize_cache_destroy(se->normalize_cache);
relevance_destroy(&se->relevance);
reclist_destroy(se->reclist);
+ xfree(se->mergekey);
+ xfree(se->rank);
+ if (nmem_total(se->nmem))
+ session_log(se, YLOG_DEBUG, "NMEN operation usage %zd", nmem_total(se->nmem));
+ if (nmem_total(se->session_nmem))
+ session_log(se, YLOG_DEBUG, "NMEN session usage %zd", nmem_total(se->session_nmem));
+ facet_limits_destroy(se->facet_limits);
nmem_destroy(se->nmem);
service_destroy(se->service);
yaz_mutex_destroy(&se->session_mutex);
size_t session_nmem;
if (session == 0)
return 0;
- session_enter(session);
+ session_enter(session, "session_get_memory_status");
session_nmem = nmem_total(session->nmem);
- session_leave(session);
+ session_leave(session, "session_get_memory_status");
return session_nmem;
}
session->number_of_warnings_unknown_elements = 0;
session->number_of_warnings_unknown_metadata = 0;
session->num_termlists = 0;
- session->reclist = 0;
- session->clients = 0;
+ session->reclist = reclist_create(nmem);
+ session->clients_active = 0;
+ session->clients_cached = 0;
+ session->settings_modified = 0;
session->session_nmem = nmem;
session->nmem = nmem_create();
session->databases = 0;
+ session->sorted_results = 0;
+ session->facet_limits = 0;
+ session->mergekey = 0;
+ session->rank = 0;
+
for (i = 0; i <= SESSION_WATCH_MAX; i++)
{
session->watchlist[i].data = 0;
session->normalize_cache = normalize_cache_create();
session->session_mutex = 0;
pazpar2_mutex_create(&session->session_mutex, tmp_str);
+ session_log(session, YLOG_LOG, "create");
+
session_use(1);
return session;
}
struct client_list *l;
size_t sz = 0;
- for (l = se->clients; l; l = l->next)
+ for (l = se->clients_active; l; l = l->next)
sz++;
res = nmem_malloc(nmem, sizeof(*res) * sz);
*count = 0;
- for (l = se->clients; l; l = l->next)
+ for (l = se->clients_active; l; l = l->next)
{
struct client *cl = l->client;
WRBUF w = wrbuf_alloc();
res[*count].id = client_get_id(cl);
res[*count].name = *name ? name : "Unknown";
res[*count].hits = client_get_hits(cl);
+ res[*count].approximation = client_get_approximation(cl);
res[*count].records = client_get_num_records(cl);
- res[*count].diagnostic = client_get_diagnostic(cl);
+ res[*count].filtered = client_get_num_records_filtered(cl);
+ res[*count].diagnostic =
+ client_get_diagnostic(cl, &res[*count].message,
+ &res[*count].addinfo);
res[*count].state = client_get_state_str(cl);
res[*count].connected = client_get_connection(cl) ? 1 : 0;
session_settings_dump(se, client_get_database(cl), w);
struct hitsbytarget *get_hitsbytarget(struct session *se, int *count, NMEM nmem)
{
struct hitsbytarget *p;
- session_enter(se);
+ session_enter(se, "get_hitsbytarget");
p = hitsbytarget_nb(se, count, nmem);
- session_leave(se);
+ session_leave(se, "get_hitsbytarget");
return p;
}
-
-struct termlist_score **get_termlist_score(struct session *se,
- const char *name, int *num)
-{
- int i;
- struct termlist_score **tl = 0;
-
- session_enter(se);
- for (i = 0; i < se->num_termlists; i++)
- if (!strcmp((const char *) se->termlists[i].name, name))
- {
- tl = termlist_highscore(se->termlists[i].termlist, num);
- break;
- }
- session_leave(se);
- return tl;
-}
// Compares two hitsbytarget nodes by hitcount
static int cmp_ht(const void *p1, const void *p2)
return h2->hits - h1->hits;
}
+// Compares two hitsbytarget nodes by hitcount
+static int cmp_ht_approx(const void *p1, const void *p2)
+{
+ const struct hitsbytarget *h1 = p1;
+ const struct hitsbytarget *h2 = p2;
+ return h2->approximation - h1->approximation;
+}
+
static int targets_termlist_nb(WRBUF wrbuf, struct session *se, int num,
- NMEM nmem)
+ NMEM nmem, int version)
{
struct hitsbytarget *ht;
int count, i;
ht = hitsbytarget_nb(se, &count, nmem);
- qsort(ht, count, sizeof(struct hitsbytarget), cmp_ht);
+ if (version >= 2)
+ qsort(ht, count, sizeof(struct hitsbytarget), cmp_ht_approx);
+ else
+ qsort(ht, count, sizeof(struct hitsbytarget), cmp_ht);
for (i = 0; i < count && i < num && ht[i].hits > 0; i++)
{
// do only print terms which have display names
-
+
wrbuf_puts(wrbuf, "<term>\n");
wrbuf_puts(wrbuf, "<id>");
wrbuf_xmlputs(wrbuf, ht[i].id);
wrbuf_puts(wrbuf, "</id>\n");
-
+
wrbuf_puts(wrbuf, "<name>");
if (!ht[i].name || !ht[i].name[0])
wrbuf_xmlputs(wrbuf, "NO TARGET NAME");
else
wrbuf_xmlputs(wrbuf, ht[i].name);
wrbuf_puts(wrbuf, "</name>\n");
-
+
wrbuf_printf(wrbuf, "<frequency>" ODR_INT_PRINTF "</frequency>\n",
ht[i].hits);
-
+
+ if (version >= 2) {
+ // Should not print if we know it isn't a approximation.
+ wrbuf_printf(wrbuf, "<approximation>" ODR_INT_PRINTF "</approximation>\n", ht[i].approximation);
+ wrbuf_printf(wrbuf, "<records>%d</records>\n", ht[i].records - ht[i].filtered);
+ wrbuf_printf(wrbuf, "<filtered>%d</filtered>\n", ht[i].filtered);
+ }
+
wrbuf_puts(wrbuf, "<state>");
wrbuf_xmlputs(wrbuf, ht[i].state);
wrbuf_puts(wrbuf, "</state>\n");
-
- wrbuf_printf(wrbuf, "<diagnostic>%d</diagnostic>\n",
+
+ wrbuf_printf(wrbuf, "<diagnostic>%d</diagnostic>\n",
ht[i].diagnostic);
wrbuf_puts(wrbuf, "</term>\n");
}
}
void perform_termlist(struct http_channel *c, struct session *se,
- const char *name, int num)
+ const char *name, int num, int version)
{
int i, j;
NMEM nmem_tmp = nmem_create();
char **names;
int num_names = 0;
- if (name)
- nmem_strsplit(nmem_tmp, ",", name, &names, &num_names);
+ if (!name)
+ name = "*";
+
+ nmem_strsplit(nmem_tmp, ",", name, &names, &num_names);
- session_enter(se);
+ session_enter(se, "perform_termlist");
for (j = 0; j < num_names; j++)
{
const char *tname;
-
- wrbuf_puts(c->wrbuf, "<list name=\"");
- wrbuf_xmlputs(c->wrbuf, names[j]);
- wrbuf_puts(c->wrbuf, "\">\n");
+ int must_generate_empty = 1; /* bug 5350 */
for (i = 0; i < se->num_termlists; i++)
{
tname = se->termlists[i].name;
- if (num_names > 0 && !strcmp(names[j], tname))
+ if (!strcmp(names[j], tname) || !strcmp(names[j], "*"))
{
struct termlist_score **p = 0;
int len;
- p = termlist_highscore(se->termlists[i].termlist, &len);
+
+ wrbuf_puts(c->wrbuf, "<list name=\"");
+ wrbuf_xmlputs(c->wrbuf, tname);
+ wrbuf_puts(c->wrbuf, "\">\n");
+ must_generate_empty = 0;
+
+ p = termlist_highscore(se->termlists[i].termlist, &len,
+ nmem_tmp);
if (p)
{
int i;
// prevent sending empty term elements
if (!p[i]->display_term || !p[i]->display_term[0])
continue;
-
+
wrbuf_puts(c->wrbuf, "<term>");
wrbuf_puts(c->wrbuf, "<name>");
wrbuf_xmlputs(c->wrbuf, p[i]->display_term);
wrbuf_puts(c->wrbuf, "</name>");
-
- wrbuf_printf(c->wrbuf,
- "<frequency>%d</frequency>",
+
+ wrbuf_printf(c->wrbuf,
+ "<frequency>%d</frequency>",
p[i]->frequency);
wrbuf_puts(c->wrbuf, "</term>\n");
}
}
+ wrbuf_puts(c->wrbuf, "</list>\n");
}
}
tname = "xtargets";
- if (num_names > 0 && !strcmp(names[j], tname))
+ if (!strcmp(names[j], tname) || !strcmp(names[j], "*"))
+ {
+ wrbuf_puts(c->wrbuf, "<list name=\"");
+ wrbuf_xmlputs(c->wrbuf, tname);
+ wrbuf_puts(c->wrbuf, "\">\n");
+
+ targets_termlist_nb(c->wrbuf, se, num, c->nmem, version);
+ wrbuf_puts(c->wrbuf, "</list>\n");
+ must_generate_empty = 0;
+ }
+ if (must_generate_empty)
{
- targets_termlist_nb(c->wrbuf, se, num, c->nmem);
+ wrbuf_puts(c->wrbuf, "<list name=\"");
+ wrbuf_xmlputs(c->wrbuf, names[j]);
+ wrbuf_puts(c->wrbuf, "\"/>\n");
}
- wrbuf_puts(c->wrbuf, "</list>\n");
}
- session_leave(se);
+ session_leave(se, "perform_termlist");
nmem_destroy(nmem_tmp);
}
nmem_get_memory_in_use(&in_use);
nmem_get_memory_free(&is_free);
- yaz_log(YLOG_LOG, "nmem stat: use=%ld free=%ld",
+ yaz_log(YLOG_LOG, "nmem stat: use=%ld free=%ld",
(long) in_use, (long) is_free);
}
#endif
{
struct record_cluster *r = 0;
- session_enter(se);
+ session_enter(se, "show_single_start");
*prev_r = 0;
*next_r = 0;
- if (se->reclist)
+ reclist_limit(se->reclist, se, 1);
+
+ reclist_enter(se->reclist);
+ while ((r = reclist_read_record(se->reclist)))
{
- reclist_enter(se->reclist);
- while ((r = reclist_read_record(se->reclist)))
+ if (!strcmp(r->recid, id))
{
- if (!strcmp(r->recid, id))
- {
- *next_r = reclist_read_record(se->reclist);
- break;
- }
- *prev_r = r;
+ *next_r = reclist_read_record(se->reclist);
+ break;
}
- reclist_leave(se->reclist);
+ *prev_r = r;
}
+ reclist_leave(se->reclist);
if (!r)
- session_leave(se);
+ session_leave(se, "show_single_start");
return r;
}
void show_single_stop(struct session *se, struct record_cluster *rec)
{
- session_leave(se);
+ session_leave(se, "show_single_stop");
+}
+
+
+int session_fetch_more(struct session *se)
+{
+ struct client_list *l;
+ int ret = 0;
+
+ for (l = se->clients_active; l; l = l->next)
+ {
+ struct client *cl = l->client;
+ if (client_get_state(cl) == Client_Idle)
+ {
+ if (client_fetch_more(cl))
+ {
+ session_log(se, YLOG_LOG, "%s: more to fetch",
+ client_get_id(cl));
+ ret = 1;
+ }
+ else
+ {
+ session_log(se, YLOG_LOG, "%s: no more to fetch",
+ client_get_id(cl));
+ }
+ }
+ else
+ {
+ session_log(se, YLOG_LOG, "%s: no fetch due to state=%s",
+ client_get_id(cl), client_get_state_str(cl));
+ }
+
+ }
+ return ret;
}
struct record_cluster **show_range_start(struct session *se,
- struct reclist_sortparms *sp,
- int start, int *num, int *total, Odr_int *sumhits)
+ struct reclist_sortparms *sp,
+ int start, int *num, int *total,
+ Odr_int *sumhits, Odr_int *approx_hits,
+ void (*show_records_ready)(void *data),
+ struct http_channel *chan)
{
- struct record_cluster **recs;
+ struct record_cluster **recs = 0;
struct reclist_sortparms *spp;
+ struct client_list *l;
int i;
-#if USE_TIMING
+#if USE_TIMING
yaz_timing_t t = yaz_timing_create();
#endif
- session_enter(se);
- recs = nmem_malloc(se->nmem, *num * sizeof(struct record_cluster *));
- if (!se->relevance)
- {
- *num = 0;
- *total = 0;
- *sumhits = 0;
- recs = 0;
- }
- else
+ session_enter(se, "show_range_start");
+ *sumhits = 0;
+ *approx_hits = 0;
+ *total = 0;
+ reclist_limit(se->reclist, se, 0);
+ if (se->relevance)
{
- struct client_list *l;
-
for (spp = sp; spp; spp = spp->next)
- if (spp->type == Metadata_sortkey_relevance)
+ if (spp->type == Metadata_type_relevance)
{
relevance_prepare_read(se->relevance, se->reclist);
break;
}
- reclist_sort(se->reclist, sp);
-
- reclist_enter(se->reclist);
- *total = reclist_get_num_records(se->reclist);
-
- *sumhits = 0;
- for (l = se->clients; l; l = l->next)
+ for (l = se->clients_active; l; l = l->next) {
*sumhits += client_get_hits(l->client);
-
- for (i = 0; i < start; i++)
- if (!reclist_read_record(se->reclist))
- {
- *num = 0;
- recs = 0;
- break;
- }
-
- for (i = 0; i < *num; i++)
+ *approx_hits += client_get_approximation(l->client);
+ }
+ }
+ reclist_sort(se->reclist, sp);
+
+ reclist_enter(se->reclist);
+ *total = reclist_get_num_records(se->reclist);
+
+ for (l = se->clients_active; l; l = l->next)
+ client_update_show_stat(l->client, 0);
+
+ for (i = 0; i < start; i++)
+ {
+ struct record_cluster *r = reclist_read_record(se->reclist);
+ if (!r)
{
- struct record_cluster *r = reclist_read_record(se->reclist);
- if (!r)
- {
- *num = i;
- break;
- }
+ *num = 0;
+ break;
+ }
+ else
+ {
+ struct record *rec = r->records;
+ for (;rec; rec = rec->next)
+ client_update_show_stat(rec->client, 1);
+ }
+ }
+ recs = nmem_malloc(se->nmem, (*num > 0 ? *num : 1) * sizeof(*recs));
+ for (i = 0; i < *num; i++)
+ {
+ struct record_cluster *r = reclist_read_record(se->reclist);
+ if (!r)
+ {
+ *num = i;
+ break;
+ }
+ else
+ {
+ struct record *rec = r->records;
+ for (;rec; rec = rec->next)
+ client_update_show_stat(rec->client, 1);
recs[i] = r;
}
- reclist_leave(se->reclist);
}
+ reclist_leave(se->reclist);
#if USE_TIMING
yaz_timing_stop(t);
- yaz_log(YLOG_LOG, "show %6.5f %3.2f %3.2f",
+ session_log(se, YLOG_LOG, "show %6.5f %3.2f %3.2f",
yaz_timing_get_real(t), yaz_timing_get_user(t),
yaz_timing_get_sys(t));
yaz_timing_destroy(&t);
#endif
+
+ if (!session_fetch_more(se))
+ session_log(se, YLOG_LOG, "can not fetch more");
+ else
+ {
+ show_range_stop(se, recs);
+ session_log(se, YLOG_LOG, "fetching more in progress");
+ if (session_set_watch(se, SESSION_WATCH_SHOW,
+ show_records_ready, chan, chan))
+ {
+ session_log(se, YLOG_WARN, "Ignoring show block");
+ session_enter(se, "show_range_start");
+ }
+ else
+ {
+ session_log(se, YLOG_LOG, "session watch OK");
+ return 0;
+ }
+ }
return recs;
}
void show_range_stop(struct session *se, struct record_cluster **recs)
{
- session_leave(se);
+ session_leave(se, "show_range_stop");
}
void statistics(struct session *se, struct statistics *stat)
memset(stat, 0, sizeof(*stat));
stat->num_hits = 0;
- for (l = se->clients; l; l = l->next)
+ for (l = se->clients_active; l; l = l->next)
{
struct client *cl = l->client;
if (!client_get_connection(cl))
{
struct record_metadata *rec_md = record_metadata_create(nmem);
struct record_metadata_attr **attrp = &rec_md->attributes;
-
+
for (; attr; attr = attr->next)
{
if (attr->children && attr->children->content)
{
- if (strcmp((const char *) attr->name, "type"))
- { /* skip the "type" attribute.. Its value is already part of
- the element in output (md-%s) and so repeating it here
- is redundant */
+ if (strcmp((const char *) attr->name, "type")
+ && strcmp((const char *) attr->name, "empty"))
+ { /* skip the "type" + "empty" attribute..
+ The "Type" is already part of the element in output
+ (md-%s) and so repeating it here is redundant */
*attrp = nmem_malloc(nmem, sizeof(**attrp));
(*attrp)->name =
nmem_strdup(nmem, (const char *) attr->name);
}
*attrp = 0;
- if (type == Metadata_type_generic)
+ switch (type)
{
- char *p = nmem_strdup(nmem, value);
-
- p = normalize7bit_generic(p, " ,/.:([");
-
- rec_md->data.text.disp = p;
+ case Metadata_type_generic:
+ case Metadata_type_skiparticle:
+ if (strstr(value, "://")) /* looks like a URL */
+ rec_md->data.text.disp = nmem_strdup(nmem, value);
+ else
+ rec_md->data.text.disp =
+ normalize7bit_generic(nmem_strdup(nmem, value), " ,/.:([");
rec_md->data.text.sort = 0;
- }
- else if (type == Metadata_type_year || type == Metadata_type_date)
+ rec_md->data.text.snippet = 0;
+ break;
+ case Metadata_type_year:
+ case Metadata_type_date:
{
int first, last;
int longdate = 0;
rec_md->data.number.min = first;
rec_md->data.number.max = last;
}
- else
+ break;
+ case Metadata_type_float:
+ rec_md->data.fnumber = atof(value);
+ break;
+ case Metadata_type_relevance:
+ case Metadata_type_position:
return 0;
+ }
return rec_md;
}
+static void mergekey_norm_wr(pp2_charset_fact_t charsets,
+ WRBUF norm_wr, const char *value)
+{
+ const char *norm_str;
+ pp2_charset_token_t prt =
+ pp2_charset_token_create(charsets, "mergekey");
+
+ pp2_charset_token_first(prt, value, 0);
+ while ((norm_str = pp2_charset_token_next(prt)))
+ {
+ if (*norm_str)
+ {
+ if (wrbuf_len(norm_wr))
+ wrbuf_puts(norm_wr, " ");
+ wrbuf_puts(norm_wr, norm_str);
+ }
+ }
+ pp2_charset_token_destroy(prt);
+}
+
static int get_mergekey_from_doc(xmlDoc *doc, xmlNode *root, const char *name,
struct conf_service *service, WRBUF norm_wr)
{
else if (!strcmp(name, (const char *) type))
{
xmlChar *value = xmlNodeListGetString(doc, n->children, 1);
- if (value)
+ if (value && *value)
{
- const char *norm_str;
- pp2_charset_token_t prt =
- pp2_charset_token_create(service->charsets, "mergekey");
-
- pp2_charset_token_first(prt, (const char *) value, 0);
if (wrbuf_len(norm_wr) > 0)
wrbuf_puts(norm_wr, " ");
wrbuf_puts(norm_wr, name);
- while ((norm_str =
- pp2_charset_token_next(prt)))
- {
- if (*norm_str)
- {
- wrbuf_puts(norm_wr, " ");
- wrbuf_puts(norm_wr, norm_str);
- }
- }
- xmlFree(value);
- pp2_charset_token_destroy(prt);
+ mergekey_norm_wr(service->charsets, norm_wr,
+ (const char *) value);
no_found++;
}
+ if (value)
+ xmlFree(value);
}
xmlFree(type);
}
return no_found;
}
-static const char *get_mergekey(xmlDoc *doc, struct client *cl, int record_no,
- struct conf_service *service, NMEM nmem)
+static const char *get_mergekey(xmlDoc *doc, xmlNode *root,
+ struct client *cl, int record_no,
+ struct conf_service *service, NMEM nmem,
+ const char *session_mergekey)
{
char *mergekey_norm = 0;
- xmlNode *root = xmlDocGetRootElement(doc);
WRBUF norm_wr = wrbuf_alloc();
+ xmlChar *mergekey;
- /* consider mergekey from XSL first */
- xmlChar *mergekey = xmlGetProp(root, (xmlChar *) "mergekey");
- if (mergekey)
+ if (session_mergekey)
{
- const char *norm_str;
- pp2_charset_token_t prt =
- pp2_charset_token_create(service->charsets, "mergekey");
+ int i, num = 0;
+ char **values = 0;
+ nmem_strsplit_escape2(nmem, ",", session_mergekey, &values,
+ &num, 1, '\\', 1);
- pp2_charset_token_first(prt, (const char *) mergekey, 0);
- while ((norm_str = pp2_charset_token_next(prt)))
- {
- if (*norm_str)
- {
- if (wrbuf_len(norm_wr))
- wrbuf_puts(norm_wr, " ");
- wrbuf_puts(norm_wr, norm_str);
- }
- }
- pp2_charset_token_destroy(prt);
+ for (i = 0; i < num; i++)
+ get_mergekey_from_doc(doc, root, values[i], service, norm_wr);
+ }
+ else if ((mergekey = xmlGetProp(root, (xmlChar *) "mergekey")))
+ {
+ mergekey_norm_wr(service->charsets, norm_wr, (const char *) mergekey);
xmlFree(mergekey);
}
else
service, norm_wr);
if (r == 0 && ser_md->mergekey == Metadata_mergekey_required)
{
- /* no mergekey on this one and it is required..
+ /* no mergekey on this one and it is required..
Generate unique key instead */
wrbuf_rewind(norm_wr);
break;
return mergekey_norm;
}
-/** \brief see if metadata for pz:recordfilter exists
+/** \brief see if metadata for pz:recordfilter exists
\param root xml root element of normalized record
\param sdb session database for client
\retval 0 if there is no metadata for pz:recordfilter
If there is no pz:recordfilter defined, this function returns 1
as well.
*/
-
+
static int check_record_filter(xmlNode *root, struct session_database *sdb)
{
int match = 0;
return match;
}
-
static int ingest_to_cluster(struct client *cl,
xmlDoc *xdoc,
xmlNode *root,
int record_no,
- const char *mergekey_norm);
+ struct record_metadata_attr *mergekey);
+
+static int ingest_sub_record(struct client *cl, xmlDoc *xdoc, xmlNode *root,
+ int record_no, NMEM nmem,
+ struct session_database *sdb,
+ struct record_metadata_attr *mergekeys)
+{
+ int ret = 0;
+ struct session *se = client_get_session(cl);
+ struct conf_service *service = se->service;
+
+ insert_settings_values(sdb, xdoc, root, service);
+
+ if (!check_record_filter(root, sdb))
+ {
+ session_log(se, YLOG_LOG,
+ "Filtered out record no %d from %s",
+ record_no, sdb->database->id);
+ return 0;
+ }
+ session_enter(se, "ingest_sub_record");
+ if (client_get_session(cl) == se && se->relevance)
+ ret = ingest_to_cluster(cl, xdoc, root, record_no, mergekeys);
+ session_leave(se, "ingest_sub_record");
+
+ return ret;
+}
/** \brief ingest XML record
\param cl client holds the result set for record
int record_no, NMEM nmem)
{
struct session *se = client_get_session(cl);
- int ret = 0;
struct session_database *sdb = client_get_database(cl);
struct conf_service *service = se->service;
xmlDoc *xdoc = normalize_record(se, sdb, service, rec, nmem);
+ int r = 0;
xmlNode *root;
- const char *mergekey_norm;
-
+
if (!xdoc)
return -1;
-
+
+ if (global_parameters.dump_records)
+ {
+ session_log(se, YLOG_LOG, "Normalized record from %s",
+ sdb->database->id);
+ log_xml_doc(xdoc);
+ }
+
root = xmlDocGetRootElement(xdoc);
-
- if (!check_record_filter(root, sdb))
+
+ if (!strcmp((const char *) root->name, "cluster"))
{
- session_log(se, YLOG_LOG, "Filtered out record no %d from %s",
- record_no, sdb->database->id);
- xmlFreeDoc(xdoc);
- return -2;
+ int no_merge_keys = 0;
+ int no_merge_dups = 0;
+ xmlNode *sroot;
+ struct record_metadata_attr *mk = 0;
+
+ for (sroot = root->children; sroot; sroot = sroot->next)
+ if (sroot->type == XML_ELEMENT_NODE &&
+ !strcmp((const char *) sroot->name, "record"))
+ {
+ struct record_metadata_attr **mkp;
+ const char *mergekey_norm =
+ get_mergekey(xdoc, sroot, cl, record_no, service, nmem,
+ se->mergekey);
+ if (!mergekey_norm)
+ {
+ r = -1;
+ break;
+ }
+ for (mkp = &mk; *mkp; mkp = &(*mkp)->next)
+ if (!strcmp((*mkp)->value, mergekey_norm))
+ break;
+ if (!*mkp)
+ {
+ *mkp = (struct record_metadata_attr*)
+ nmem_malloc(nmem, sizeof(**mkp));
+ (*mkp)->name = 0;
+ (*mkp)->value = nmem_strdup(nmem, mergekey_norm);
+ (*mkp)->next = 0;
+ no_merge_keys++;
+ }
+ else
+ no_merge_dups++;
+ }
+ if (no_merge_keys > 1 || no_merge_dups > 0)
+ {
+ yaz_log(YLOG_LOG, "Got %d mergekeys, %d dups for position %d",
+ no_merge_keys, no_merge_dups, record_no);
+ }
+ for (sroot = root->children; !r && sroot; sroot = sroot->next)
+ if (sroot->type == XML_ELEMENT_NODE &&
+ !strcmp((const char *) sroot->name, "record"))
+ {
+ r = ingest_sub_record(cl, xdoc, sroot, record_no, nmem, sdb,
+ mk);
+ }
}
-
- mergekey_norm = get_mergekey(xdoc, cl, record_no, service, nmem);
- if (!mergekey_norm)
+ else if (!strcmp((const char *) root->name, "record"))
{
- session_log(se, YLOG_WARN, "Got no mergekey");
- xmlFreeDoc(xdoc);
- return -1;
+ const char *mergekey_norm =
+ get_mergekey(xdoc, root, cl, record_no, service, nmem,
+ se->mergekey);
+ if (mergekey_norm)
+ {
+ struct record_metadata_attr *mk = (struct record_metadata_attr*)
+ nmem_malloc(nmem, sizeof(*mk));
+ mk->name = 0;
+ mk->value = nmem_strdup(nmem, mergekey_norm);
+ mk->next = 0;
+
+ r = ingest_sub_record(cl, xdoc, root, record_no, nmem, sdb, mk);
+ }
+ }
+ else
+ {
+ session_log(se, YLOG_WARN, "Bad pz root element: %s",
+ (const char *) root->name);
+ r = -1;
}
- session_enter(se);
- if (client_get_session(cl) == se)
- ingest_to_cluster(cl, xdoc, root, record_no, mergekey_norm);
- session_leave(se);
-
xmlFreeDoc(xdoc);
+ return r;
+}
+
+
+// struct conf_metadata *ser_md = &service->metadata[md_field_id];
+// struct record_metadata *rec_md = record->metadata[md_field_id];
+static int match_metadata_local(struct conf_service *service,
+ struct conf_metadata *ser_md,
+ struct record_metadata *rec_md0,
+ char **values, int num_v)
+{
+ int i;
+ struct record_metadata *rec_md = rec_md0;
+ WRBUF val_wr = 0;
+ WRBUF text_wr = wrbuf_alloc();
+ for (i = 0; i < num_v; )
+ {
+ if (rec_md)
+ {
+ if (ser_md->type == Metadata_type_year
+ || ser_md->type == Metadata_type_date)
+ {
+ int y = atoi(values[i]);
+ if (y >= rec_md->data.number.min
+ && y <= rec_md->data.number.max)
+ break;
+ }
+ else
+ {
+ if (!val_wr)
+ {
+ val_wr = wrbuf_alloc();
+ mergekey_norm_wr(service->charsets, val_wr, values[i]);
+ }
+ wrbuf_rewind(text_wr);
+ mergekey_norm_wr(service->charsets, text_wr,
+ rec_md->data.text.disp);
+ if (!strcmp(wrbuf_cstr(val_wr), wrbuf_cstr(text_wr)))
+ break;
+ }
+ rec_md = rec_md->next;
+ }
+ else
+ {
+ rec_md = rec_md0;
+ wrbuf_destroy(val_wr);
+ val_wr = 0;
+ i++;
+ }
+ }
+ wrbuf_destroy(val_wr);
+ wrbuf_destroy(text_wr);
+ return i < num_v ? 1 : 0;
+}
+
+int session_check_cluster_limit(struct session *se, struct record_cluster *rec)
+{
+ int i;
+ struct conf_service *service = se->service;
+ int ret = 1;
+ const char *name;
+ const char *value;
+ NMEM nmem_tmp = nmem_create();
+
+ for (i = 0; (name = facet_limits_get(se->facet_limits, i, &value)); i++)
+ {
+ int j;
+ for (j = 0; j < service->num_metadata; j++)
+ {
+ struct conf_metadata *md = service->metadata + j;
+ if (!strcmp(md->name, name) && md->limitcluster)
+ {
+ char **values = 0;
+ int num = 0;
+ int md_field_id =
+ conf_service_metadata_field_id(service,
+ md->limitcluster);
+
+ if (md_field_id < 0)
+ {
+ ret = 0;
+ break;
+ }
+
+ nmem_strsplit_escape2(nmem_tmp, "|", value, &values,
+ &num, 1, '\\', 1);
+
+ if (!match_metadata_local(service,
+ &service->metadata[md_field_id],
+ rec->metadata[md_field_id],
+ values, num))
+ {
+ ret = 0;
+ break;
+ }
+ }
+ }
+ }
+ nmem_destroy(nmem_tmp);
return ret;
}
+// Skip record on non-zero
+static int check_limit_local(struct client *cl,
+ struct record *record,
+ int record_no)
+{
+ int skip_record = 0;
+ struct session *se = client_get_session(cl);
+ struct conf_service *service = se->service;
+ NMEM nmem_tmp = nmem_create();
+ struct session_database *sdb = client_get_database(cl);
+ int l = 0;
+ while (!skip_record)
+ {
+ int md_field_id;
+ char **values = 0;
+ int num_v = 0;
+ const char *name =
+ client_get_facet_limit_local(cl, sdb, &l, nmem_tmp,
+ &num_v, &values);
+ if (!name)
+ break;
+
+ if (!strcmp(name, "*"))
+ {
+ for (md_field_id = 0; md_field_id < service->num_metadata;
+ md_field_id++)
+ {
+ if (match_metadata_local(
+ service,
+ &service->metadata[md_field_id],
+ record->metadata[md_field_id],
+ values, num_v))
+ break;
+ }
+ if (md_field_id == service->num_metadata)
+ skip_record = 1;
+ }
+ else
+ {
+ md_field_id = conf_service_metadata_field_id(service, name);
+ if (md_field_id < 0)
+ {
+ skip_record = 1;
+ break;
+ }
+ if (!match_metadata_local(
+ service,
+ &service->metadata[md_field_id],
+ record->metadata[md_field_id],
+ values, num_v))
+ {
+ skip_record = 1;
+ }
+ }
+ }
+ nmem_destroy(nmem_tmp);
+ return skip_record;
+}
+
static int ingest_to_cluster(struct client *cl,
xmlDoc *xdoc,
xmlNode *root,
int record_no,
- const char *mergekey_norm)
+ struct record_metadata_attr *merge_keys)
{
xmlNode *n;
xmlChar *type = 0;
struct conf_service *service = se->service;
int term_factor = 1;
struct record_cluster *cluster;
+ struct record_metadata **metadata0;
struct session_database *sdb = client_get_database(cl);
- struct record *record = record_create(se->nmem,
+ NMEM ingest_nmem = 0;
+ char **rank_values = 0;
+ int rank_num = 0;
+ struct record *record = record_create(se->nmem,
service->num_metadata,
service->num_sortkeys, cl,
record_no);
if (value)
xmlFree(value);
type = value = 0;
-
+
if (n->type != XML_ELEMENT_NODE)
continue;
if (!strcmp((const char *) n->name, "metadata"))
struct record_metadata **wheretoput = 0;
struct record_metadata *rec_md = 0;
int md_field_id = -1;
-
+
type = xmlGetProp(n, (xmlChar *) "type");
value = xmlNodeListGetString(xdoc, n->children, 1);
-
- if (!type || !value || !*value)
+ if (!type)
continue;
-
- md_field_id
+ if (!value || !*value)
+ {
+ xmlChar *empty = xmlGetProp(n, (xmlChar *) "empty");
+ if (!empty)
+ continue;
+ if (value)
+ xmlFree(value);
+ value = empty;
+ }
+ md_field_id
= conf_service_metadata_field_id(service, (const char *) type);
if (md_field_id < 0)
{
if (se->number_of_warnings_unknown_metadata == 0)
{
- session_log(se, YLOG_WARN,
+ session_log(se, YLOG_WARN,
"Ignoring unknown metadata element: %s", type);
}
se->number_of_warnings_unknown_metadata++;
continue;
}
-
+
ser_md = &service->metadata[md_field_id];
// non-merged metadata
"for element '%s'", value, type);
continue;
}
+
+ if (ser_md->type == Metadata_type_generic)
+ {
+ WRBUF w = wrbuf_alloc();
+ if (relevance_snippet(se->relevance,
+ (char*) value, ser_md->name, w))
+ rec_md->data.text.snippet = nmem_strdup(se->nmem,
+ wrbuf_cstr(w));
+ wrbuf_destroy(w);
+ }
+
+
wheretoput = &record->metadata[md_field_id];
while (*wheretoput)
wheretoput = &(*wheretoput)->next;
}
}
- cluster = reclist_insert(se->reclist, service, record,
- mergekey_norm, &se->total_merged);
+ if (check_limit_local(cl, record, record_no))
+ {
+ session_log(se, YLOG_LOG, "Facet filtered out record no %d from %s",
+ record_no, sdb->database->id);
+ if (type)
+ xmlFree(type);
+ if (value)
+ xmlFree(value);
+ return -2;
+ }
+ cluster = reclist_insert(se->reclist, se->relevance, service, record,
+ merge_keys, &se->total_merged);
if (!cluster)
- return -1;
+ {
+ if (type)
+ xmlFree(type);
+ if (value)
+ xmlFree(value);
+ return 0; // complete match with existing record
+ }
{
const char *use_term_factor_str =
int hits = (int) client_get_hits(cl);
term_factor = MAX(hits, maxrecs) / MAX(1, maxrecs);
assert(term_factor >= 1);
- yaz_log(YLOG_DEBUG, "Using term factor: %d (%d / %d)", term_factor, MAX(hits, maxrecs), MAX(1, maxrecs));
+ session_log(se, YLOG_DEBUG, "Using term factor: %d (%d / %d)",
+ term_factor, MAX(hits, maxrecs), MAX(1, maxrecs));
}
}
session_log(se, YLOG_LOG, "Cluster id %s from %s (#%d)", cluster->recid,
sdb->database->id, record_no);
+ // original metadata, to check if first existence of a field
+ metadata0 = xmalloc(sizeof(*metadata0) * service->num_metadata);
+ memcpy(metadata0, cluster->metadata,
+ sizeof(*metadata0) * service->num_metadata);
+
+ ingest_nmem = nmem_create();
+ if (se->rank)
+ {
+ yaz_log(YLOG_LOG, "local in sort : %s", se->rank);
+ nmem_strsplit_escape2(ingest_nmem, ",", se->rank, &rank_values,
+ &rank_num, 1, '\\', 1);
+ }
- relevance_newrec(se->relevance, cluster);
-
// now parsing XML record and adding data to cluster or record metadata
for (n = root->children; n; n = n->next)
{
if (value)
xmlFree(value);
type = value = 0;
-
+
if (n->type != XML_ELEMENT_NODE)
continue;
if (!strcmp((const char *) n->name, "metadata"))
struct record_metadata *rec_md = 0;
int md_field_id = -1;
int sk_field_id = -1;
-
+ const char *rank = 0;
+ xmlChar *xml_rank = 0;
+
type = xmlGetProp(n, (xmlChar *) "type");
value = xmlNodeListGetString(xdoc, n->children, 1);
-
+
if (!type || !value || !*value)
continue;
-
- md_field_id
+
+ md_field_id
= conf_service_metadata_field_id(service, (const char *) type);
if (md_field_id < 0)
continue;
-
+
ser_md = &service->metadata[md_field_id];
-
+
if (ser_md->sortkey_offset >= 0)
{
sk_field_id = ser_md->sortkey_offset;
// merged metadata
rec_md = record_metadata_init(se->nmem, (const char *) value,
ser_md->type, 0);
+
+ // see if the field was not in cluster already (from beginning)
+
if (!rec_md)
continue;
+ if (rank_num)
+ {
+ int i;
+ for (i = 0; i < rank_num; i++)
+ {
+ const char *val = rank_values[i];
+ const char *cp = strchr(val, '=');
+ if (!cp)
+ continue;
+ if ((cp - val) == strlen((const char *) type)
+ && !memcmp(val, type, cp - val))
+ {
+ rank = cp + 1;
+ break;
+ }
+ }
+ }
+ else
+ {
+ xml_rank = xmlGetProp(n, (xmlChar *) "rank");
+ rank = xml_rank ? (const char *) xml_rank : ser_md->rank;
+ }
+
wheretoput = &cluster->metadata[md_field_id];
- // and polulate with data:
- // assign cluster or record based on merge action
- if (ser_md->merge == Metadata_merge_unique)
+ if (ser_md->merge == Metadata_merge_first)
+ {
+ if (!metadata0[md_field_id])
+ {
+ while (*wheretoput)
+ wheretoput = &(*wheretoput)->next;
+ *wheretoput = rec_md;
+ }
+ }
+ else if (ser_md->merge == Metadata_merge_unique)
{
while (*wheretoput)
{
- if (!strcmp((const char *) (*wheretoput)->data.text.disp,
+ if (!strcmp((const char *) (*wheretoput)->data.text.disp,
rec_md->data.text.disp))
break;
wheretoput = &(*wheretoput)->next;
}
else if (ser_md->merge == Metadata_merge_longest)
{
- if (!*wheretoput
- || strlen(rec_md->data.text.disp)
+ if (!*wheretoput
+ || strlen(rec_md->data.text.disp)
> strlen((*wheretoput)->data.text.disp))
{
*wheretoput = rec_md;
if (ser_sk)
{
const char *sort_str = 0;
- int skip_article =
- ser_sk->type == Metadata_sortkey_skiparticle;
+ int skip_article =
+ ser_sk->type == Metadata_type_skiparticle;
if (!cluster->sortkeys[sk_field_id])
- cluster->sortkeys[sk_field_id] =
- nmem_malloc(se->nmem,
+ cluster->sortkeys[sk_field_id] =
+ nmem_malloc(se->nmem,
sizeof(union data_types));
-
+
prt =
pp2_charset_token_create(service->charsets, "sort");
skip_article);
pp2_charset_token_next(prt);
-
+
sort_str = pp2_get_sort(prt);
-
- cluster->sortkeys[sk_field_id]->text.disp =
+
+ cluster->sortkeys[sk_field_id]->text.disp =
rec_md->data.text.disp;
if (!sort_str)
{
sort_str = rec_md->data.text.disp;
- session_log(se, YLOG_WARN,
+ session_log(se, YLOG_WARN,
"Could not make sortkey. Bug #1858");
}
- cluster->sortkeys[sk_field_id]->text.sort =
+ cluster->sortkeys[sk_field_id]->text.sort =
nmem_strdup(se->nmem, sort_str);
pp2_charset_token_destroy(prt);
}
{
*wheretoput = rec_md;
if (ser_sk)
- cluster->sortkeys[sk_field_id]
+ cluster->sortkeys[sk_field_id]
= &rec_md->data;
}
else
}
}
-
- // ranking of _all_ fields enabled ...
- if (ser_md->rank)
- relevance_countwords(se->relevance, cluster,
- (char *) value, ser_md->rank,
- ser_md->name);
-
+ // ranking of _all_ fields enabled ...
+ if (rank)
+ {
+ relevance_countwords(se->relevance, cluster,
+ (char *) value, rank, ser_md->name);
+ }
// construct facets ... unless the client already has reported them
if (ser_md->termlist && !client_has_facet(cl, (char *) type))
{
}
// cleaning up
+ if (xml_rank)
+ xmlFree(xml_rank);
xmlFree(type);
xmlFree(value);
type = value = 0;
if (value)
xmlFree(value);
+ nmem_destroy(ingest_nmem);
+ xfree(metadata0);
relevance_donerecord(se->relevance, cluster);
se->total_records++;
va_start(ap, fmt);
yaz_vsnprintf(buf, sizeof(buf)-30, fmt, ap);
- yaz_log(level, "Session (%u): %s", s->session_id, buf);
+ yaz_log(level, "Session %u: %s", s ? s->session_id : 0, buf);
va_end(ap);
}