/* This file is part of Pazpar2.
- Copyright (C) 2006-2011 Index Data
+ Copyright (C) 2006-2012 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 "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
no_session_total += delta;
sessions = no_sessions;
yaz_mutex_leave(g_session_mutex);
- yaz_log(YLOG_DEBUG, "%s sesions=%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;
}
-int sessions_count(void) {
+int sessions_count(void)
+{
return session_use(0);
}
-int session_count_total(void) {
+int session_count_total(void)
+{
int total = 0;
if (!g_session_mutex)
return 0;
return total;
}
-
static void log_xml_doc(xmlDoc *doc)
{
FILE *lf = yaz_log_file();
yaz_mutex_leave(s->session_mutex);
}
-// Recursively traverse query structure to extract terms.
-void pull_terms(NMEM nmem, struct ccl_rpn_node *n, char **termlist, int *num)
+static void session_normalize_facet(struct session *s, const char *type,
+ const char *value,
+ WRBUF display_wrbuf,
+ WRBUF facet_wrbuf)
{
- char **words;
- int numwords;
+ struct conf_service *service = s->service;
+ pp2_charset_token_t prt;
+ const char *facet_component;
int i;
+ const char *icu_chain_id = 0;
- switch (n->kind)
+ for (i = 0; i < service->num_metadata; i++)
+ if (!strcmp((service->metadata + i)->name, type))
+ icu_chain_id = (service->metadata + i)->facetrule;
+ if (!icu_chain_id)
+ icu_chain_id = "facet";
+ prt = pp2_charset_token_create(service->charsets, icu_chain_id);
+ if (!prt)
{
- case CCL_RPN_AND:
- case CCL_RPN_OR:
- case CCL_RPN_NOT:
- case CCL_RPN_PROX:
- pull_terms(nmem, n->u.p[0], termlist, num);
- pull_terms(nmem, n->u.p[1], termlist, num);
- break;
- case CCL_RPN_TERM:
- nmem_strsplit(nmem, " ", n->u.t.term, &words, &numwords);
- for (i = 0; i < numwords; i++)
- termlist[(*num)++] = words[i];
- break;
- default: // NOOP
- break;
+ yaz_log(YLOG_FATAL, "Unknown ICU chain '%s' for facet of type '%s'",
+ icu_chain_id, type);
+ wrbuf_destroy(facet_wrbuf);
+ wrbuf_destroy(display_wrbuf);
+ return;
}
-}
-
-
-void add_facet(struct session *s, const char *type, const char *value, int count)
-{
- struct conf_service *service = s->service;
- pp2_relevance_token_t prt;
- const char *facet_component;
- WRBUF facet_wrbuf = wrbuf_alloc();
- prt = pp2_relevance_tokenize(service->facet_pct);
-
- pp2_relevance_first(prt, value, 0);
- while ((facet_component = pp2_relevance_token_next(prt)))
+ pp2_charset_token_first(prt, value, 0);
+ while ((facet_component = pp2_charset_token_next(prt)))
{
+ const char *display_component;
if (*facet_component)
{
if (wrbuf_len(facet_wrbuf))
wrbuf_puts(facet_wrbuf, " ");
wrbuf_puts(facet_wrbuf, facet_component);
}
+ display_component = pp2_get_display(prt);
+ if (display_component)
+ {
+ if (wrbuf_len(display_wrbuf))
+ wrbuf_puts(display_wrbuf, " ");
+ wrbuf_puts(display_wrbuf, display_component);
+ }
}
- pp2_relevance_token_destroy(prt);
-
+ 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;
{
session_log(s, YLOG_FATAL, "Too many termlists");
wrbuf_destroy(facet_wrbuf);
+ 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(facet_wrbuf),
- count);
+ termlist_insert(s->termlists[i].termlist, wrbuf_cstr(display_wrbuf),
+ wrbuf_cstr(facet_wrbuf), count);
}
wrbuf_destroy(facet_wrbuf);
+ wrbuf_destroy(display_wrbuf);
}
static xmlDoc *record_to_xml(struct session *se,
if (!rdoc)
{
session_log(se, YLOG_FATAL, "Non-wellformed XML received from %s",
- db->url);
+ db->id);
return 0;
}
if (global_parameters.dump_records)
{
- session_log(se, YLOG_LOG, "Un-normalized record from %s", db->url);
+ session_log(se, YLOG_LOG, "Un-normalized record from %s", db->id);
log_xml_doc(rdoc);
}
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->url);
+ 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->url);
+ session_log(se, YLOG_LOG, "Normalized record from %s",
+ sdb->database->id);
log_xml_doc(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->url);
- 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->server->config, 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->url);
- 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)
{
ret = -1;
else
{
-
+
s->watchlist[what].fun = fun;
s->watchlist[what].data = data;
s->watchlist[what].obs = http_add_observer(chan, &s->watchlist[what],
}
//callback for grep_databases
-static void select_targets_callback(void *context, struct session_database *db)
+static void select_targets_callback(struct session *se,
+ struct session_database *db)
{
- struct session *se = (struct session*) context;
- struct client *cl = client_create();
+ 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;
+ l = se->clients_active;
+ se->clients_active = new_list;
+ session_leave(se);
+
+ 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);
+ l = se->clients_cached;
+ se->clients_cached = 0;
session_leave(se);
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;
}
-enum pazpar2_error_code search(struct session *se,
- const char *query,
- const char *startrecs, const char *maxrecs,
- const char *filter,
- const char *limit,
- const char **addinfo)
+static void session_clear_set(struct session *se,
+ const char *sort_field, int increasing, int position)
+{
+ reclist_destroy(se->reclist);
+ se->reclist = 0;
+ 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;
+
+ /* 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->position = position;
+ se->sorted_results->next = 0;
+
+ session_log(se, YLOG_DEBUG, "clear_set search_sort: field=%s increasing=%d position=%d configured",
+ sort_field, increasing, position);
+
+ se->reclist = reclist_create(se->nmem);
+}
+
+void session_sort(struct session *se, const char *field, int increasing,
+ int position)
+{
+ struct session_sorted_results *sr;
+ struct client_list *l;
+
+ session_enter(se);
+
+ yaz_log(YLOG_LOG, "session_sort field=%s increasing=%d position=%d", field, increasing, position);
+ /* see if we already have sorted for this critieria */
+ for (sr = se->sorted_results; sr; sr = sr->next)
+ {
+ if (!strcmp(field, sr->field) && increasing == sr->increasing && sr->position == position)
+ break;
+ }
+ if (sr)
+ {
+ session_log(se, YLOG_DEBUG, "search_sort: field=%s increasing=%d position=%d already fetched",
+ field, increasing, position);
+ session_leave(se);
+ return;
+ }
+ session_log(se, YLOG_DEBUG, "search_sort: field=%s increasing=%d position=%d must fetch",
+ field, increasing, position);
+ if (position)
+ {
+ yaz_log(YLOG_DEBUG, "Reset results due to position");
+ session_clear_set(se, field, increasing, position);
+ }
+ else {
+ sr = nmem_malloc(se->nmem, sizeof(*sr));
+ sr->field = nmem_strdup(se->nmem, field);
+ sr->increasing = increasing;
+ sr->position = position;
+ sr->next = se->sorted_results;
+ se->sorted_results = sr;
+ }
+ yaz_log(YLOG_DEBUG, "Restarting search for clients due to change in sort order");
+
+ for (l = se->clients_active; l; l = l->next)
+ {
+ struct client *cl = l->client;
+ if (client_get_state(cl) == Client_Connecting ||
+ client_get_state(cl) == Client_Idle ||
+ client_get_state(cl) == Client_Working) {
+ yaz_log(YLOG_DEBUG, "Client %s: Restarting search due to change in sort order", client_get_id(cl));
+ client_start_search(cl);
+ }
+ }
+ session_leave(se);
+}
+
+enum pazpar2_error_code session_search(struct session *se,
+ const char *query,
+ const char *startrecs,
+ const char *maxrecs,
+ const char *filter,
+ const char *limit,
+ const char **addinfo,
+ struct reclist_sortparms *sp)
{
int live_channels = 0;
int no_working = 0;
- int no_failed = 0;
- struct client_list *l;
+ int no_failed_query = 0;
+ int no_failed_limit = 0;
+ struct client_list *l, *l0;
struct timeval tval;
facet_limits_t facet_limits;
*addinfo = 0;
- session_remove_clients(se);
-
+ if (se->settings_modified)
+ session_remove_cached_clients(se);
+ else
+ session_reset_active_clients(se, 0);
+
session_enter(se);
- reclist_destroy(se->reclist);
- se->reclist = 0;
+ se->settings_modified = 0;
+ session_clear_set(se, sp->name, sp->increasing, sp->type == Metadata_sortkey_position);
relevance_destroy(&se->relevance);
- nmem_reset(se->nmem);
- se->total_records = se->total_hits = se->total_merged = 0;
- se->num_termlists = 0;
+
live_channels = select_targets(se, filter);
if (!live_channels)
{
session_leave(se);
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)
{
*addinfo = "limit";
+ session_leave(se);
return PAZPAR2_MALFORMED_PARAMETER_VALUE;
}
- for (l = se->clients; l; l = l->next)
+
+ l0 = se->clients_active;
+ se->clients_active = 0;
+ session_leave(se);
+
+ for (l = l0; l; l = l->next)
{
+ int parse_ret;
struct client *cl = l->client;
- 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++;
+ if (prepare_map(se, client_get_database(cl)) < 0)
+ continue;
+
+ parse_ret = client_parse_query(cl, query, facet_limits, startrecs,
+ maxrecs, se->service->ccl_bibset);
+ if (parse_ret == -1)
+ no_failed_query++;
+ else if (parse_ret == -2)
+ no_failed_limit++;
else
{
- no_working++;
- if (client_prep_connection(cl, se->service->z3950_operation_timeout,
+ int r =
+ client_prep_connection(cl, se->service->z3950_operation_timeout,
se->service->z3950_session_timeout,
se->service->server->iochan_man,
- &tval))
+ &tval);
+ if (parse_ret == 1 && r == 2)
+ {
+ session_log(se, YLOG_LOG, "client %s REUSE result", client_get_id(cl));
+ client_reingest(cl);
+ }
+ else if (r)
+ {
+ session_log(se, YLOG_LOG, "client %s NEW search", client_get_id(cl));
client_start_search(cl);
+ }
+ no_working++;
}
}
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;
}
+ session_log(se, YLOG_LOG, "session_start_search done");
return PAZPAR2_NO_ERROR;
}
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);
-
- resolve_database(se->service, db);
-
+ 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;
for (sdb = se->databases; sdb; sdb = sdb->next)
- if (!strcmp(sdb->database->url, id))
+ if (!strcmp(sdb->database->id, id))
return sdb;
return load_session_database(se, id);
}
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_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);
+ 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));
nmem_destroy(se->nmem);
service_destroy(se->service);
yaz_mutex_destroy(&se->session_mutex);
- wrbuf_destroy(se->wrbuf);
-}
-
-/* Depreciated: use session_destroy */
-void destroy_session(struct session *se)
-{
- session_destroy(se);
}
size_t session_get_memory_status(struct session *session) {
session_log(session, YLOG_DEBUG, "New");
session->service = service;
session->relevance = 0;
- session->total_hits = 0;
session->total_records = 0;
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->clients_active = 0;
+ session->clients_cached = 0;
+ session->settings_modified = 0;
session->session_nmem = nmem;
session->nmem = nmem_create();
- session->wrbuf = wrbuf_alloc();
session->databases = 0;
for (i = 0; i <= SESSION_WATCH_MAX; i++)
{
return session;
}
-struct hitsbytarget *hitsbytarget(struct session *se, int *count, NMEM nmem)
+const char * client_get_suggestions_xml(struct client *cl, WRBUF wrbuf);
+
+static struct hitsbytarget *hitsbytarget_nb(struct session *se,
+ int *count, NMEM nmem)
{
struct hitsbytarget *res = 0;
struct client_list *l;
size_t sz = 0;
- session_enter(se);
- 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();
const char *name = session_setting_oneval(client_get_database(cl),
PZ_NAME);
- res[*count].id = client_get_database(cl)->database->url;
+ 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].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);
res[*count].settings_xml = nmem_strdup(nmem, wrbuf_cstr(w));
+ wrbuf_rewind(w);
+ wrbuf_puts(w, "");
+ res[*count].suggestions_xml = nmem_strdup(nmem, client_get_suggestions_xml(cl, w));
wrbuf_destroy(w);
(*count)++;
}
- session_leave(se);
return res;
}
-struct termlist_score **get_termlist_score(struct session *se,
- const char *name, int *num)
+struct hitsbytarget *get_hitsbytarget(struct session *se, int *count, NMEM nmem)
{
- int i;
- struct termlist_score **tl = 0;
+ struct hitsbytarget *p;
+ session_enter(se);
+ p = hitsbytarget_nb(se, count, nmem);
+ session_leave(se);
+ return p;
+}
+
+// Compares two hitsbytarget nodes by hitcount
+static int cmp_ht(const void *p1, const void *p2)
+{
+ const struct hitsbytarget *h1 = p1;
+ const struct hitsbytarget *h2 = 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, int version)
+{
+ struct hitsbytarget *ht;
+ int count, i;
+
+ ht = hitsbytarget_nb(se, &count, nmem);
+ 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",
+ ht[i].diagnostic);
+ wrbuf_puts(wrbuf, "</term>\n");
+ }
+ return count;
+}
+
+void perform_termlist(struct http_channel *c, struct session *se,
+ const char *name, int num, int version)
+{
+ int i, j;
+ NMEM nmem_tmp = nmem_create();
+ char **names;
+ int num_names = 0;
+
+ if (!name)
+ name = "*";
+
+ nmem_strsplit(nmem_tmp, ",", name, &names, &num_names);
session_enter(se);
- for (i = 0; i < se->num_termlists; i++)
- if (!strcmp((const char *) se->termlists[i].name, name))
+
+ for (j = 0; j < num_names; j++)
+ {
+ const char *tname;
+ int must_generate_empty = 1; /* bug 5350 */
+
+ for (i = 0; i < se->num_termlists; i++)
{
- tl = termlist_highscore(se->termlists[i].termlist, num);
- break;
+ tname = se->termlists[i].name;
+ if (!strcmp(names[j], tname) || !strcmp(names[j], "*"))
+ {
+ struct termlist_score **p = 0;
+ int 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;
+ for (i = 0; i < len && i < num; 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>",
+ p[i]->frequency);
+ wrbuf_puts(c->wrbuf, "</term>\n");
+ }
+ }
+ wrbuf_puts(c->wrbuf, "</list>\n");
+ }
}
+ tname = "xtargets";
+ 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)
+ {
+ wrbuf_puts(c->wrbuf, "<list name=\"");
+ wrbuf_xmlputs(c->wrbuf, names[j]);
+ wrbuf_puts(c->wrbuf, "\"/>\n");
+ }
+ }
session_leave(se);
- return tl;
+ nmem_destroy(nmem_tmp);
}
#ifdef MISSING_HEADERS
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 **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)
{
struct record_cluster **recs;
struct reclist_sortparms *spp;
int i;
-#if USE_TIMING
+#if USE_TIMING
yaz_timing_t t = yaz_timing_create();
#endif
session_enter(se);
*num = 0;
*total = 0;
*sumhits = 0;
+ *approx_hits = 0;
recs = 0;
}
else
{
+ struct client_list *l;
+
for (spp = sp; spp; spp = spp->next)
if (spp->type == Metadata_sortkey_relevance)
{
break;
}
reclist_sort(se->reclist, sp);
-
+
reclist_enter(se->reclist);
*total = reclist_get_num_records(se->reclist);
- *sumhits = se->total_hits;
-
+
+ *sumhits = 0;
+ *approx_hits = 0;
+ for (l = se->clients_active; l; l = l->next) {
+ *sumhits += client_get_hits(l->client);
+ *approx_hits += client_get_approximation(l->client);
+ }
for (i = 0; i < start; i++)
if (!reclist_read_record(se->reclist))
{
recs = 0;
break;
}
-
+
for (i = 0; i < *num; i++)
{
struct record_cluster *r = reclist_read_record(se->reclist);
}
#if USE_TIMING
yaz_timing_stop(t);
- yaz_log(YLOG_LOG, "show %6.5f %3.2f %3.2f",
+ yaz_log(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);
int count = 0;
memset(stat, 0, sizeof(*stat));
- for (l = se->clients; l; l = l->next)
+ stat->num_hits = 0;
+ for (l = se->clients_active; l; l = l->next)
{
struct client *cl = l->client;
if (!client_get_connection(cl))
stat->num_no_connection++;
+ stat->num_hits += client_get_hits(cl);
switch (client_get_state(cl))
{
case Client_Connecting: stat->num_connecting++; break;
}
count++;
}
- stat->num_hits = se->total_hits;
stat->num_records = se->total_records;
stat->num_clients = count;
{
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)
char *p = nmem_strdup(nmem, value);
p = normalize7bit_generic(p, " ,/.:([");
-
+
rec_md->data.text.disp = p;
rec_md->data.text.sort = 0;
}
if (value)
{
const char *norm_str;
- pp2_relevance_token_t prt =
- pp2_relevance_tokenize(service->mergekey_pct);
-
- pp2_relevance_first(prt, (const char *) value, 0);
+ 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_relevance_token_next(prt)))
+ pp2_charset_token_next(prt)))
{
if (*norm_str)
{
}
}
xmlFree(value);
- pp2_relevance_token_destroy(prt);
+ pp2_charset_token_destroy(prt);
no_found++;
}
}
if (mergekey)
{
const char *norm_str;
- pp2_relevance_token_t prt =
- pp2_relevance_tokenize(service->mergekey_pct);
+ pp2_charset_token_t prt =
+ pp2_charset_token_create(service->charsets, "mergekey");
- pp2_relevance_first(prt, (const char *) mergekey, 0);
- while ((norm_str = pp2_relevance_token_next(prt)))
+ pp2_charset_token_first(prt, (const char *) mergekey, 0);
+ while ((norm_str = pp2_charset_token_next(prt)))
{
if (*norm_str)
{
wrbuf_puts(norm_wr, norm_str);
}
}
- pp2_relevance_token_destroy(prt);
+ pp2_charset_token_destroy(prt);
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;
/* generate unique key if none is not generated already or is empty */
if (wrbuf_len(norm_wr) == 0)
{
- wrbuf_printf(norm_wr, "%s-%d",
- client_get_database(cl)->database->url, record_no);
+ wrbuf_printf(norm_wr, "position: %s-%d",
+ client_get_id(cl), record_no);
+ }
+ else
+ {
+ const char *lead = "content: ";
+ wrbuf_insert(norm_wr, 0, lead, strlen(lead));
}
if (wrbuf_len(norm_wr) > 0)
mergekey_norm = nmem_strdup(nmem, wrbuf_cstr(norm_wr));
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;
xmlDoc *xdoc = normalize_record(se, sdb, service, rec, nmem);
xmlNode *root;
const char *mergekey_norm;
-
+
if (!xdoc)
return -1;
-
+
root = xmlDocGetRootElement(xdoc);
-
+
if (!check_record_filter(root, sdb))
{
- session_log(se, YLOG_LOG, "Filtered out record no %d from %s",
- record_no, sdb->database->url);
+ session_log(se, YLOG_LOG, "Filtered out record no %d from %s", record_no, sdb->database->id);
xmlFreeDoc(xdoc);
return -2;
}
-
+
mergekey_norm = get_mergekey(xdoc, cl, record_no, service, nmem);
if (!mergekey_norm)
{
if (client_get_session(cl) == se)
ret = ingest_to_cluster(cl, xdoc, root, record_no, mergekey_norm);
session_leave(se);
-
+
xmlFreeDoc(xdoc);
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)
+ {
+ struct conf_metadata *ser_md = 0;
+ struct record_metadata *rec_md = 0;
+ int md_field_id;
+ char **values = 0;
+ int i, num_v = 0;
+
+ const char *name = client_get_facet_limit_local(cl, sdb, &l, nmem_tmp, &num_v, &values);
+ if (!name)
+ break;
+
+ md_field_id = conf_service_metadata_field_id(service, name);
+ if (md_field_id < 0)
+ {
+ skip_record = 1;
+ break;
+ }
+ ser_md = &service->metadata[md_field_id];
+ rec_md = record->metadata[md_field_id];
+ yaz_log(YLOG_DEBUG, "check limit local %s", name);
+ 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
+ {
+ yaz_log(YLOG_DEBUG, "cmp: '%s' '%s'", rec_md->data.text.disp, values[i]);
+ if (!strcmp(rec_md->data.text.disp, values[i]))
+ {
+ // Value equals, should not be filtered.
+ break;
+ }
+ }
+ rec_md = rec_md->next;
+ }
+ else
+ {
+ rec_md = record->metadata[md_field_id];
+ i++;
+ }
+ }
+ // At end , not match
+ if (i == num_v)
+ {
+ skip_record = 1;
+ break;
+ }
+ }
+ nmem_destroy(nmem_tmp);
+ return skip_record;
+}
+
static int ingest_to_cluster(struct client *cl,
xmlDoc *xdoc,
xmlNode *root,
xmlNode *n;
xmlChar *type = 0;
xmlChar *value = 0;
- struct session_database *sdb = client_get_database(cl);
struct session *se = client_get_session(cl);
struct conf_service *service = se->service;
- struct record *record = record_create(se->nmem,
+ int term_factor = 1;
+ struct record_cluster *cluster;
+ struct session_database *sdb = client_get_database(cl);
+ struct record *record = record_create(se->nmem,
service->num_metadata,
service->num_sortkeys, cl,
record_no);
- struct record_cluster *cluster = reclist_insert(se->reclist,
- service,
- record,
- mergekey_norm,
- &se->total_merged);
-
- const char *use_term_factor_str = session_setting_oneval(sdb, PZ_TERMLIST_TERM_FACTOR);
- int use_term_factor = 0;
- int term_factor = 1;
- if (use_term_factor_str && use_term_factor_str[0] != 0)
- use_term_factor = atoi(use_term_factor_str);
- if (use_term_factor) {
- int maxrecs = client_get_maxrecs(cl);
- 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));
- }
- if (!cluster)
- return -1;
- if (global_parameters.dump_records)
- session_log(se, YLOG_LOG, "Cluster id %s from %s (#%d)", cluster->recid,
- sdb->database->url, record_no);
- 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)
{
- pp2_relevance_token_t prt;
if (type)
xmlFree(type);
if (value)
xmlFree(value);
type = value = 0;
-
+
if (n->type != XML_ELEMENT_NODE)
continue;
if (!strcmp((const char *) n->name, "metadata"))
{
struct conf_metadata *ser_md = 0;
- struct conf_sortkey *ser_sk = 0;
struct record_metadata **wheretoput = 0;
struct record_metadata *rec_md = 0;
int md_field_id = -1;
- int sk_field_id = -1;
-
+
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)
{
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];
-
- if (ser_md->sortkey_offset >= 0){
- sk_field_id = ser_md->sortkey_offset;
- ser_sk = &service->sortkeys[sk_field_id];
- }
// non-merged metadata
rec_md = record_metadata_init(se->nmem, (const char *) value,
while (*wheretoput)
wheretoput = &(*wheretoput)->next;
*wheretoput = rec_md;
+ }
+ }
+
+ 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, service, record,
+ mergekey_norm, &se->total_merged);
+ if (!cluster)
+ return -1;
+
+ {
+ const char *use_term_factor_str =
+ session_setting_oneval(sdb, PZ_TERMLIST_TERM_FACTOR);
+ if (use_term_factor_str && use_term_factor_str[0] == '1')
+ {
+ int maxrecs = client_get_maxrecs(cl);
+ 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));
+ }
+ }
+
+ if (global_parameters.dump_records)
+ session_log(se, YLOG_LOG, "Cluster id %s from %s (#%d)", cluster->recid,
+ sdb->database->id, record_no);
+
+
+ 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)
+ {
+ pp2_charset_token_t prt;
+ if (type)
+ xmlFree(type);
+ if (value)
+ xmlFree(value);
+ type = value = 0;
+
+ if (n->type != XML_ELEMENT_NODE)
+ continue;
+ if (!strcmp((const char *) n->name, "metadata"))
+ {
+ struct conf_metadata *ser_md = 0;
+ struct conf_sortkey *ser_sk = 0;
+ struct record_metadata **wheretoput = 0;
+ struct record_metadata *rec_md = 0;
+ int md_field_id = -1;
+ int sk_field_id = -1;
+ const char *rank;
+ xmlChar *xml_rank;
+
+ type = xmlGetProp(n, (xmlChar *) "type");
+ value = xmlNodeListGetString(xdoc, n->children, 1);
+
+ if (!type || !value || !*value)
+ continue;
+
+ 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;
+ ser_sk = &service->sortkeys[sk_field_id];
+ }
// merged metadata
rec_md = record_metadata_init(se->nmem, (const char *) value,
ser_md->type, 0);
+ if (!rec_md)
+ continue;
+
+ 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)
{
- struct record_metadata *mnode;
- for (mnode = *wheretoput; mnode; mnode = mnode->next)
- if (!strcmp((const char *) mnode->data.text.disp,
+ while (*wheretoput)
+ {
+ if (!strcmp((const char *) (*wheretoput)->data.text.disp,
rec_md->data.text.disp))
break;
- if (!mnode)
- {
- rec_md->next = *wheretoput;
- *wheretoput = rec_md;
+ wheretoput = &(*wheretoput)->next;
}
+ if (!*wheretoput)
+ *wheretoput = rec_md;
}
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 =
+ int skip_article =
ser_sk->type == Metadata_sortkey_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_relevance_tokenize(service->sort_pct);
- pp2_relevance_first(prt, rec_md->data.text.disp,
- skip_article);
+ prt =
+ pp2_charset_token_create(service->charsets, "sort");
+
+ pp2_charset_token_first(prt, rec_md->data.text.disp,
+ skip_article);
+
+ pp2_charset_token_next(prt);
- pp2_relevance_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_relevance_token_destroy(prt);
+ pp2_charset_token_destroy(prt);
}
}
}
else if (ser_md->merge == Metadata_merge_all)
{
- rec_md->next = *wheretoput;
+ while (*wheretoput)
+ wheretoput = &(*wheretoput)->next;
*wheretoput = rec_md;
}
else if (ser_md->merge == Metadata_merge_range)
{
*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;
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->session_id, buf);
va_end(ap);
}