/* 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
struct client_list *next;
};
-struct session_sorted_results {
- const char *field;
- int increasing;
- struct session_sorted_results *next;
-};
-
/* session counting (1) , disable client counting (0) */
static YAZ_MUTEX g_session_mutex = 0;
static int no_sessions = 0;
yaz_mutex_leave(s->session_mutex);
}
-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;
}
}
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))
{
}
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);
// 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 && sdb->settings[PZ_XSLT] && !sdb->map &&
- (s = session_setting_oneval(sdb, PZ_XSLT)))
+ if (sdb->settings && !sdb->map)
{
- 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;
}
{
cl = client_create(db->database->id);
client_set_database(cl, db);
- client_set_session(cl, se);
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;
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;
}
return res == 0;
}
-static const char *get_strategy_plus_sort(struct client *l, const char *field)
+static void session_clear_set(struct session *se,
+ const char *sort_field, int increasing, int position)
{
- struct session_database *sdb = client_get_database(l);
- struct setting *s;
-
- const char *strategy_plus_sort = 0;
+ 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;
- for (s = sdb->settings[PZ_SORTMAP]; s; s = s->next)
- {
- char *p = strchr(s->name + 3, ':');
- if (!p)
- {
- yaz_log(YLOG_WARN, "Malformed sortmap name: %s", s->name);
- continue;
- }
- p++;
- if (!strcmp(p, field))
- {
- strategy_plus_sort = s->value;
- break;
- }
- }
- return strategy_plus_sort;
+ /* 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;
+
+ se->reclist = reclist_create(se->nmem);
}
-void session_sort(struct session *se, const char *field, int increasing)
+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 */
+ /* TODO I do not see the point in saving all previous sorts. Dont we re-sort anyway ? */
for (sr = se->sorted_results; sr; sr = sr->next)
{
- if (!strcmp(field, sr->field) && increasing == sr->increasing)
+ if (!strcmp(field, sr->field) && increasing == sr->increasing && sr->position == position)
break;
}
if (sr)
{
- yaz_log(YLOG_LOG, "search_sort: field=%s increasing=%d already fetched",
- field, increasing);
+ session_log(se, YLOG_DEBUG, "search_sort: field=%s increasing=%d position=%d already fetched",
+ field, increasing, position);
session_leave(se);
return;
}
- 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;
+ 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;
- const char *strategy_plus_sort = get_strategy_plus_sort(cl, field);
- if (strategy_plus_sort)
- {
- struct timeval tval;
- 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, strategy_plus_sort, increasing);
+ 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);
const char *filter,
const char *limit,
const char **addinfo,
- const char *sort_field, int increasing)
+ const char *sort_field,
+ int increasing)
{
int live_channels = 0;
int no_working = 0;
session_reset_active_clients(se, 0);
session_enter(se);
- reclist_destroy(se->reclist);
- se->reclist = 0;
se->settings_modified = 0;
+ session_clear_set(se, sort_field, increasing, 0); /* hardcoded position */
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);
return PAZPAR2_NO_TARGETS;
}
- se->reclist = reclist_create(se->nmem);
yaz_gettimeofday(&tval);
{
int parse_ret;
struct client *cl = l->client;
- const char *strategy_plus_sort = get_strategy_plus_sort(cl, sort_field);
if (prepare_map(se, client_get_database(cl)) < 0)
continue;
parse_ret = client_parse_query(cl, query, facet_limits, startrecs,
- maxrecs);
+ maxrecs, se->service->ccl_bibset);
if (parse_ret == -1)
no_failed_query++;
else if (parse_ret == -2)
no_failed_limit++;
- else if (parse_ret == 0)
- {
- session_log(se, YLOG_LOG, "client NEW %s", client_get_id(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, strategy_plus_sort, increasing);
- }
else
{
- session_log(se, YLOG_LOG, "client REUSE %s", client_get_id(cl));
- 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);
else
return PAZPAR2_NO_TARGETS;
}
- yaz_log(YLOG_LOG, "session_start_search done");
+ session_log(se, YLOG_LOG, "session_start_search done");
return PAZPAR2_NO_ERROR;
}
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
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);
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);
session_leave(se);
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++)
{
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");
}
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);
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;
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);
nmem_destroy(nmem_tmp);
struct record_cluster **show_range_start(struct session *se,
struct reclist_sortparms *sp,
- int start, int *num, int *total, Odr_int *sumhits)
+ int start, int *num, int *total, Odr_int *sumhits, Odr_int *approx_hits)
{
struct record_cluster **recs;
struct reclist_sortparms *spp;
{
*num = 0;
*total = 0;
- *sumhits = 0;
+ *sumhits = 0;
+ *approx_hits = 0;
recs = 0;
}
else
*total = reclist_get_num_records(se->reclist);
*sumhits = 0;
- for (l = se->clients_active; l; l = l->next)
+ *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))
{
if (!check_record_filter(root, sdb))
{
- session_log(se, YLOG_LOG, "Filtered out record no %d from %s",
- record_no, sdb->database->id);
+ session_log(se, YLOG_LOG, "Filtered out record no %d from %s", record_no, sdb->database->id);
xmlFreeDoc(xdoc);
return -2;
}
}
session_enter(se);
if (client_get_session(cl) == se)
- ingest_to_cluster(cl, xdoc, root, record_no, mergekey_norm);
+ 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,
}
}
+ 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)
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);
continue;
ser_md = &service->metadata[md_field_id];
-
+
if (ser_md->sortkey_offset >= 0)
{
sk_field_id = ser_md->sortkey_offset;
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:
}
}
-
// ranking of _all_ fields enabled ...
- if (ser_md->rank)
+ if (rank)
+ {
relevance_countwords(se->relevance, cluster,
- (char *) value, ser_md->rank,
- ser_md->name);
+ (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);
}