-/* $Id: pazpar2.c,v 1.6 2006-12-27 21:11:10 quinn Exp $ */;
+/* $Id: pazpar2.c,v 1.51 2007-03-20 07:27:51 adam Exp $ */
#include <stdlib.h>
#include <stdio.h>
#include <ctype.h>
#include <assert.h>
+#include <yaz/marcdisp.h>
#include <yaz/comstack.h>
#include <yaz/tcpip.h>
#include <yaz/proto.h>
#include <yaz/yaz-util.h>
#include <yaz/nmem.h>
+#if HAVE_CONFIG_H
+#include "cconfig.h"
+#endif
+
+#define USE_TIMING 0
+#if USE_TIMING
+#include <yaz/timing.h>
+#endif
+
+#include <netinet/in.h>
+
#include "pazpar2.h"
#include "eventl.h"
-#include "command.h"
#include "http.h"
#include "termlists.h"
#include "reclists.h"
#include "relevance.h"
#include "config.h"
+#include "database.h"
-#define PAZPAR2_VERSION "0.1"
#define MAX_CHUNK 15
static void client_fatal(struct client *cl);
static void connection_destroy(struct connection *co);
static int client_prep_connection(struct client *cl);
static void ingest_records(struct client *cl, Z_Records *r);
+//static struct conf_retrievalprofile *database_retrieval_profile(struct database *db);
void session_alert_watch(struct session *s, int what);
IOCHAN channel_list = 0; // Master list of connections we're handling events to
static struct connection *connection_freelist = 0;
static struct client *client_freelist = 0;
-static struct host *hosts = 0; // The hosts we know about
-static struct database *databases = 0; // The databases we know about
-
static char *client_states[] = {
"Client_Connecting",
"Client_Connected",
"Client_Stopped"
};
+// Note: Some things in this structure will eventually move to configuration
struct parameters global_parameters =
{
+ "",
+ "",
+ 0,
+ 0,
30,
"81",
"Index Data PazPar2 (MasterKey)",
- PAZPAR2_VERSION,
+ VERSION,
600, // 10 minutes
60,
100,
0,
0,
0,
- 0,
0
};
-
static int send_apdu(struct client *c, Z_APDU *a)
{
struct connection *co = c->connection;
for (ndb = 0; db->databases[ndb]; ndb++)
databaselist[ndb] = db->databases[ndb];
- a->u.presentRequest->preferredRecordSyntax =
+ a->u.searchRequest->preferredRecordSyntax =
yaz_oidval_to_z3950oid(global_parameters.odr_out,
CLASS_RECSYN, VAL_USMARC);
a->u.searchRequest->smallSetUpperBound = &ssub;
int start = cl->records + 1;
toget = global_parameters.chunk;
+ if (toget > global_parameters.toget - cl->records)
+ toget = global_parameters.toget - cl->records;
if (toget > cl->hits - cl->records)
toget = cl->hits - cl->records;
if (r->presentStatus && !*r->presentStatus && r->records)
{
yaz_log(YLOG_DEBUG, "Records in search response");
- cl->records += *r->numberOfRecordsReturned;
ingest_records(cl, r->records);
}
cl->state = Client_Idle;
}
}
-const char *find_field(const char *rec, const char *field)
+char *normalize_mergekey(char *buf, int skiparticle)
{
- char lbuf[5];
- char *line;
+ char *p = buf, *pout = buf;
- lbuf[0] = '\n';
- strcpy(lbuf + 1, field);
-
- if ((line = strstr(rec, lbuf)))
- return ++line;
- else
- return 0;
-}
-
-const char *find_subfield(const char *field, char subfield)
-{
- const char *p = field;
-
- while (*p && *p != '\n')
+ if (skiparticle)
{
- while (*p != '\n' && *p != '\t')
- p++;
- if (*p == '\t' && *(++p) == subfield) {
- if (*(++p) == ' ')
- {
- while (isspace(*p))
- p++;
- return p;
- }
- }
- }
- return 0;
-}
+ char firstword[64];
+ char articles[] = "the den der die des an a "; // must end in space
-// Extract 245 $a $b 100 $a
-char *extract_title(struct session *s, const char *rec)
-{
- const char *field, *subfield;
- char *e, *ef;
- unsigned char *obuf, *p;
-
- wrbuf_rewind(s->wrbuf);
-
- if (!(field = find_field(rec, "245")))
- return 0;
- if (!(subfield = find_subfield(field, 'a')))
- return 0;
- ef = index(subfield, '\n');
- if ((e = index(subfield, '\t')) && e < ef)
- ef = e;
- if (ef)
- {
- wrbuf_write(s->wrbuf, subfield, ef - subfield);
- if ((subfield = find_subfield(field, 'b')))
- {
- ef = index(subfield, '\n');
- if ((e = index(subfield, '\t')) && e < ef)
- ef = e;
- if (ef)
- {
- wrbuf_putc(s->wrbuf, ' ');
- wrbuf_write(s->wrbuf, subfield, ef - subfield);
- }
- }
- }
- if ((field = find_field(rec, "100")))
- {
- if ((subfield = find_subfield(field, 'a')))
- {
- ef = index(subfield, '\n');
- if ((e = index(subfield, '\t')) && e < ef)
- ef = e;
- if (ef)
- {
- wrbuf_puts(s->wrbuf, ", by ");
- wrbuf_write(s->wrbuf, subfield, ef - subfield);
- }
- }
- }
- wrbuf_putc(s->wrbuf, '\0');
- obuf = (unsigned char*) nmem_strdup(s->nmem, wrbuf_buf(s->wrbuf));
- for (p = obuf; *p; p++)
- if (*p == '&' || *p == '<' || *p > 122 || *p < ' ')
- *p = ' ';
- return (char*) obuf;
-}
-
-// Extract 245 $a $b 100 $a
-char *extract_mergekey(struct session *s, const char *rec)
-{
- const char *field, *subfield;
- char *e, *ef;
- char *out, *p, *pout;
-
- wrbuf_rewind(s->wrbuf);
-
- if (!(field = find_field(rec, "245")))
- return 0;
- if (!(subfield = find_subfield(field, 'a')))
- return 0;
- ef = index(subfield, '\n');
- if ((e = index(subfield, '\t')) && e < ef)
- ef = e;
- if (ef)
- {
- wrbuf_write(s->wrbuf, subfield, ef - subfield);
- if ((subfield = find_subfield(field, 'b')))
- {
- ef = index(subfield, '\n');
- if ((e = index(subfield, '\t')) && e < ef)
- ef = e;
- if (ef)
- {
- wrbuf_puts(s->wrbuf, " field ");
- wrbuf_write(s->wrbuf, subfield, ef - subfield);
- }
- }
- }
- if ((field = find_field(rec, "100")))
- {
- if ((subfield = find_subfield(field, 'a')))
- {
- ef = index(subfield, '\n');
- if ((e = index(subfield, '\t')) && e < ef)
- ef = e;
- if (ef)
- {
- wrbuf_puts(s->wrbuf, " field ");
- wrbuf_write(s->wrbuf, subfield, ef - subfield);
- }
- }
+ while (*p && !isalnum(*p))
+ p++;
+ pout = firstword;
+ while (*p && *p != ' ' && pout - firstword < 62)
+ *(pout++) = tolower(*(p++));
+ *(pout++) = ' ';
+ *(pout++) = '\0';
+ if (!strstr(articles, firstword))
+ p = buf;
+ pout = buf;
}
- wrbuf_putc(s->wrbuf, '\0');
- p = wrbuf_buf(s->wrbuf);
- out = pout = nmem_malloc(s->nmem, strlen(p) + 1);
while (*p)
{
+ while (*p && !isalnum(*p))
+ p++;
while (isalnum(*p))
*(pout++) = tolower(*(p++));
+ if (*p)
+ *(pout++) = ' ';
while (*p && !isalnum(*p))
p++;
- *(pout++) = ' ';
}
- if (out != pout)
- *(--pout) = '\0';
-
- return out;
-}
-
-#ifdef RECHEAP
-static void push_record(struct session *s, struct record *r)
-{
- int p;
- assert(s->recheap_max + 1 < s->recheap_size);
-
- s->recheap[p = ++s->recheap_max] = r;
- while (p > 0)
- {
- int parent = (p - 1) >> 1;
- if (strcmp(s->recheap[p]->merge_key, s->recheap[parent]->merge_key) < 0)
- {
- struct record *tmp;
- tmp = s->recheap[parent];
- s->recheap[parent] = s->recheap[p];
- s->recheap[p] = tmp;
- p = parent;
- }
- else
- break;
- }
-}
-
-static struct record *top_record(struct session *s)
-{
- return s-> recheap_max >= 0 ? s->recheap[0] : 0;
-}
-
-static struct record *pop_record(struct session *s)
-{
- struct record *res;
- int p = 0;
- int lastnonleaf = (s->recheap_max - 1) >> 1;
-
- if (s->recheap_max < 0)
- return 0;
-
- res = s->recheap[0];
-
- s->recheap[p] = s->recheap[s->recheap_max--];
-
- while (p <= lastnonleaf)
- {
- int right = (p + 1) << 1;
- int left = right - 1;
- int min = left;
-
- if (right < s->recheap_max &&
- strcmp(s->recheap[right]->merge_key, s->recheap[left]->merge_key) < 0)
- min = right;
- if (strcmp(s->recheap[min]->merge_key, s->recheap[p]->merge_key) < 0)
- {
- struct record *tmp = s->recheap[min];
- s->recheap[min] = s->recheap[p];
- s->recheap[p] = tmp;
- p = min;
+ if (buf != pout)
+ do {
+ *(pout--) = '\0';
}
- else
- break;
- }
- return res;
-}
-
-// Like pop_record but collapses identical (merge_key) records
-// The heap will contain multiple independent matching records and possibly
-// one cluster, created the last time the list was scanned
-static struct record *pop_mrecord(struct session *s)
-{
- struct record *this;
- struct record *next;
+ while (pout > buf && *pout == ' ');
- if (!(this = pop_record(s)))
- return 0;
-
- // Collapse identical records
- while ((next = top_record(s)))
- {
- struct record *p, *tmpnext;
- if (strcmp(this->merge_key, next->merge_key))
- break;
- // Absorb record (and clustersiblings) into a supercluster
- for (p = next; p; p = tmpnext) {
- tmpnext = p->next_cluster;
- p->next_cluster = this->next_cluster;
- this->next_cluster = p;
- }
-
- pop_record(s);
- }
- return this;
+ return buf;
}
-// Reads records in sort order. Store records in top of heapspace until rewind is called.
-static struct record *read_recheap(struct session *s)
-{
- struct record *r = pop_mrecord(s);
-
- if (r)
- {
- if (s->recheap_scratch < 0)
- s->recheap_scratch = s->recheap_size;
- s->recheap[--s->recheap_scratch] = r;
- }
-
- return r;
-}
-
-// Return records to heap after read
-static void rewind_recheap(struct session *s)
-{
- while (s->recheap_scratch >= 0) {
- push_record(s, s->recheap[s->recheap_scratch++]);
- if (s->recheap_scratch >= s->recheap_size)
- s->recheap_scratch = -1;
- }
-}
-
-#endif
+#ifdef GAGA
// FIXME needs to be generalized. Should flexibly generate X lists per search
static void extract_subject(struct session *s, const char *rec)
{
assert(len < 1023);
memcpy(buf, subfield, len);
buf[len] = '\0';
+#ifdef FIXME
if (*buf)
termlist_insert(s->termlist, buf);
+#endif
}
}
}
+#endif
-static void pull_relevance_field(struct session *s, struct record *head, const char *rec,
- char *field, int mult)
+static void add_facet(struct session *s, const char *type, const char *value)
{
- const char *fb;
- while ((fb = find_field(rec, field)))
+ int i;
+
+ if (!*value)
+ return;
+ for (i = 0; i < s->num_termlists; i++)
+ if (!strcmp(s->termlists[i].name, type))
+ break;
+ if (i == s->num_termlists)
+ {
+ if (i == SESSION_MAX_TERMLISTS)
+ {
+ yaz_log(YLOG_FATAL, "Too many termlists");
+ exit(1);
+ }
+ s->termlists[i].name = nmem_strdup(s->nmem, type);
+ s->termlists[i].termlist = termlist_create(s->nmem, s->expected_maxrecs, 15);
+ s->num_termlists = i + 1;
+ }
+ termlist_insert(s->termlists[i].termlist, value);
+}
+
+static xmlDoc *normalize_record(struct client *cl, Z_External *rec)
+{
+ struct conf_retrievalprofile *rprofile = cl->database->rprofile;
+ struct conf_retrievalmap *m;
+ xmlNode *res;
+ xmlDoc *rdoc;
+
+ // First normalize to XML
+ if (rprofile->native_syntax == Nativesyn_iso2709)
+ {
+ char *buf;
+ int len;
+ if (rec->which != Z_External_octet)
+ {
+ yaz_log(YLOG_WARN, "Unexpected external branch, probably BER");
+ return 0;
+ }
+ buf = (char*) rec->u.octet_aligned->buf;
+ len = rec->u.octet_aligned->len;
+ if (yaz_marc_read_iso2709(rprofile->yaz_marc, buf, len) < 0)
+ {
+ yaz_log(YLOG_WARN, "Failed to decode MARC");
+ return 0;
+ }
+ if (yaz_marc_write_xml(rprofile->yaz_marc, &res,
+ "http://www.loc.gov/MARC21/slim", 0, 0) < 0)
+ {
+ yaz_log(YLOG_WARN, "Failed to encode as XML");
+ return 0;
+ }
+ rdoc = xmlNewDoc("1.0");
+ xmlDocSetRootElement(rdoc, res);
+ }
+ else
+ {
+ yaz_log(YLOG_FATAL, "Unknown native_syntax in normalize_record");
+ exit(1);
+ }
+
+ if (global_parameters.dump_records)
{
- char *ffield = strchr(fb, '\t');
- if (!ffield)
- return;
- char *eol = strchr(ffield, '\n');
- if (!eol)
- return;
- relevance_countwords(s->relevance, head, ffield, eol - ffield, mult);
- rec = field + 1; // Crude way to cause a loop through repeating fields
+ fprintf(stderr, "Input Record (normalized):\n----------------\n");
+#if LIBXML_VERSION >= 20600
+ xmlDocFormatDump(stderr, rdoc, 1);
+#else
+ xmlDocDump(stderr, rdoc);
+#endif
}
+
+ for (m = rprofile->maplist; m; m = m->next)
+ {
+ xmlDoc *new;
+ if (m->type != Map_xslt)
+ {
+ yaz_log(YLOG_WARN, "Unknown map type");
+ return 0;
+ }
+ if (!(new = xsltApplyStylesheet(m->stylesheet, rdoc, 0)))
+ {
+ yaz_log(YLOG_WARN, "XSLT transformation failed");
+ return 0;
+ }
+ xmlFreeDoc(rdoc);
+ rdoc = new;
+ }
+ if (global_parameters.dump_records)
+ {
+ fprintf(stderr, "Record:\n----------------\n");
+#if LIBXML_VERSION >= 20600
+ xmlDocFormatDump(stderr, rdoc, 1);
+#else
+ xmlDocDump(stderr, rdoc);
+#endif
+ }
+ return rdoc;
}
-static void pull_relevance_keys(struct session *s, struct record *head, struct record *rec)
+// Extract what appears to be years from buf, storing highest and
+// lowest values.
+static int extract_years(const char *buf, int *first, int *last)
{
- relevance_newrec(s->relevance, head);
- pull_relevance_field(s, head, rec->buf, "100", 2);
- pull_relevance_field(s, head, rec->buf, "245", 4);
- //pull_relevance_field(s, head, rec->buf, "530", 1);
- pull_relevance_field(s, head, rec->buf, "630", 1);
- pull_relevance_field(s, head, rec->buf, "650", 1);
- pull_relevance_field(s, head, rec->buf, "700", 1);
- relevance_donerecord(s->relevance, head);
+ *first = -1;
+ *last = -1;
+ while (*buf)
+ {
+ const char *e;
+ int len;
+
+ while (*buf && !isdigit(*buf))
+ buf++;
+ len = 0;
+ for (e = buf; *e && isdigit(*e); e++)
+ len++;
+ if (len == 4)
+ {
+ int value = atoi(buf);
+ if (*first < 0 || value < *first)
+ *first = value;
+ if (*last < 0 || value > *last)
+ *last = value;
+ }
+ buf = e;
+ }
+ return *first;
}
-static struct record *ingest_record(struct client *cl, char *buf, int len)
+static struct record *ingest_record(struct client *cl, Z_External *rec)
{
- struct session *se = cl->session;
+ xmlDoc *xdoc = normalize_record(cl, rec);
+ xmlNode *root, *n;
struct record *res;
- struct record *head;
- const char *recbuf;
+ struct record_cluster *cluster;
+ struct session *se = cl->session;
+ xmlChar *mergekey, *mergekey_norm;
+ xmlChar *type = 0;
+ xmlChar *value = 0;
+ struct conf_service *service = global_parameters.server->service;
+
+ if (!xdoc)
+ return 0;
- wrbuf_rewind(se->wrbuf);
- yaz_marc_xml(global_parameters.yaz_marc, YAZ_MARC_LINE);
- if (yaz_marc_decode_wrbuf(global_parameters.yaz_marc, buf, len, se->wrbuf) < 0)
+ root = xmlDocGetRootElement(xdoc);
+ if (!(mergekey = xmlGetProp(root, "mergekey")))
{
- yaz_log(YLOG_WARN, "Failed to decode MARC record");
+ yaz_log(YLOG_WARN, "No mergekey found in record");
+ xmlFreeDoc(xdoc);
return 0;
}
- wrbuf_putc(se->wrbuf, '\0');
- recbuf = wrbuf_buf(se->wrbuf);
res = nmem_malloc(se->nmem, sizeof(struct record));
- res->buf = nmem_strdup(se->nmem, recbuf);
+ res->next = 0;
+ res->client = cl;
+ res->metadata = nmem_malloc(se->nmem,
+ sizeof(struct record_metadata*) * service->num_metadata);
+ memset(res->metadata, 0, sizeof(struct record_metadata*) * service->num_metadata);
+
+ mergekey_norm = nmem_strdup(se->nmem, (char*) mergekey);
+ xmlFree(mergekey);
+ normalize_mergekey(mergekey_norm, 0);
+
+ cluster = reclist_insert(se->reclist, res, mergekey_norm, &se->total_merged);
+ if (global_parameters.dump_records)
+ yaz_log(YLOG_LOG, "Cluster id %d from %s (#%d)", cluster->recid,
+ cl->database->url, cl->records);
+ if (!cluster)
+ {
+ /* no room for record */
+ xmlFreeDoc(xdoc);
+ return 0;
+ }
+ relevance_newrec(se->relevance, cluster);
- extract_subject(se, res->buf);
+ for (n = root->children; n; n = n->next)
+ {
+ if (type)
+ xmlFree(type);
+ if (value)
+ xmlFree(value);
+ type = value = 0;
- res->title = extract_title(se, res->buf);
- res->merge_key = extract_mergekey(se, res->buf);
- if (!res->merge_key)
- return 0;
- res->client = cl;
- res->next_cluster = 0;
- res->target_offset = -1;
- res->term_frequency_vec = 0;
+ if (n->type != XML_ELEMENT_NODE)
+ continue;
+ if (!strcmp(n->name, "metadata"))
+ {
+ struct conf_metadata *md = 0;
+ struct conf_sortkey *sk = 0;
+ struct record_metadata **wheretoput, *newm;
+ int imeta;
+ int first, last;
+
+ type = xmlGetProp(n, "type");
+ value = xmlNodeListGetString(xdoc, n->children, 0);
+
+ if (!type || !value)
+ continue;
- head = reclist_insert(se->reclist, res);
+ // First, find out what field we're looking at
+ for (imeta = 0; imeta < service->num_metadata; imeta++)
+ if (!strcmp(type, service->metadata[imeta].name))
+ {
+ md = &service->metadata[imeta];
+ if (md->sortkey_offset >= 0)
+ sk = &service->sortkeys[md->sortkey_offset];
+ break;
+ }
+ if (!md)
+ {
+ yaz_log(YLOG_WARN, "Ignoring unknown metadata element: %s", type);
+ continue;
+ }
+
+ // Find out where we are putting it
+ if (md->merge == Metadata_merge_no)
+ wheretoput = &res->metadata[imeta];
+ else
+ wheretoput = &cluster->metadata[imeta];
+
+ // Put it there
+ newm = nmem_malloc(se->nmem, sizeof(struct record_metadata));
+ newm->next = 0;
+ if (md->type == Metadata_type_generic)
+ {
+ char *p, *pe;
+ for (p = value; *p && isspace(*p); p++)
+ ;
+ for (pe = p + strlen(p) - 1;
+ pe > p && strchr(" ,/.:([", *pe); pe--)
+ *pe = '\0';
+ newm->data.text = nmem_strdup(se->nmem, p);
+
+ }
+ else if (md->type == Metadata_type_year)
+ {
+ if (extract_years(value, &first, &last) < 0)
+ continue;
+ }
+ else
+ {
+ yaz_log(YLOG_WARN, "Unknown type in metadata element %s", type);
+ continue;
+ }
+ if (md->type == Metadata_type_year && md->merge != Metadata_merge_range)
+ {
+ yaz_log(YLOG_WARN, "Only range merging supported for years");
+ continue;
+ }
+ if (md->merge == Metadata_merge_unique)
+ {
+ struct record_metadata *mnode;
+ for (mnode = *wheretoput; mnode; mnode = mnode->next)
+ if (!strcmp(mnode->data.text, newm->data.text))
+ break;
+ if (!mnode)
+ {
+ newm->next = *wheretoput;
+ *wheretoput = newm;
+ }
+ }
+ else if (md->merge == Metadata_merge_longest)
+ {
+ if (!*wheretoput ||
+ strlen(newm->data.text) > strlen((*wheretoput)->data.text))
+ {
+ *wheretoput = newm;
+ if (sk)
+ {
+ char *s = nmem_strdup(se->nmem, newm->data.text);
+ if (!cluster->sortkeys[md->sortkey_offset])
+ cluster->sortkeys[md->sortkey_offset] =
+ nmem_malloc(se->nmem, sizeof(union data_types));
+ normalize_mergekey(s,
+ (sk->type == Metadata_sortkey_skiparticle));
+ cluster->sortkeys[md->sortkey_offset]->text = s;
+ }
+ }
+ }
+ else if (md->merge == Metadata_merge_all || md->merge == Metadata_merge_no)
+ {
+ newm->next = *wheretoput;
+ *wheretoput = newm;
+ }
+ else if (md->merge == Metadata_merge_range)
+ {
+ assert(md->type == Metadata_type_year);
+ if (!*wheretoput)
+ {
+ *wheretoput = newm;
+ (*wheretoput)->data.number.min = first;
+ (*wheretoput)->data.number.max = last;
+ if (sk)
+ cluster->sortkeys[md->sortkey_offset] = &newm->data;
+ }
+ else
+ {
+ if (first < (*wheretoput)->data.number.min)
+ (*wheretoput)->data.number.min = first;
+ if (last > (*wheretoput)->data.number.max)
+ (*wheretoput)->data.number.max = last;
+ }
+#ifdef GAGA
+ if (sk)
+ {
+ union data_types *sdata = cluster->sortkeys[md->sortkey_offset];
+ yaz_log(YLOG_LOG, "SK range: %d-%d", sdata->number.min, sdata->number.max);
+ }
+#endif
+ }
+ else
+ yaz_log(YLOG_WARN, "Don't know how to merge on element name %s", md->name);
+
+ if (md->rank)
+ relevance_countwords(se->relevance, cluster, value, md->rank);
+ if (md->termlist)
+ {
+ if (md->type == Metadata_type_year)
+ {
+ char year[64];
+ sprintf(year, "%d", last);
+ add_facet(se, type, year);
+ if (first != last)
+ {
+ sprintf(year, "%d", first);
+ add_facet(se, type, year);
+ }
+ }
+ else
+ add_facet(se, type, value);
+ }
+ xmlFree(type);
+ xmlFree(value);
+ type = value = 0;
+ }
+ else
+ yaz_log(YLOG_WARN, "Unexpected element %s in internal record", n->name);
+ }
+ if (type)
+ xmlFree(type);
+ if (value)
+ xmlFree(value);
- pull_relevance_keys(se, head, res);
+ xmlFreeDoc(xdoc);
+ relevance_donerecord(se->relevance, cluster);
se->total_records++;
return res;
static void ingest_records(struct client *cl, Z_Records *r)
{
+#if USE_TIMING
+ yaz_timing_t t = yaz_timing_create();
+#endif
struct record *rec;
struct session *s = cl->session;
Z_NamePlusRecordList *rlist;
for (i = 0; i < rlist->num_records; i++)
{
Z_NamePlusRecord *npr = rlist->records[i];
- Z_External *e;
- char *buf;
- int len;
+ cl->records++;
if (npr->which != Z_NamePlusRecord_databaseRecord)
{
yaz_log(YLOG_WARN, "Unexpected record type, probably diagnostic");
continue;
}
- e = npr->u.databaseRecord;
- if (e->which != Z_External_octet)
- {
- yaz_log(YLOG_WARN, "Unexpected external branch, probably BER");
- continue;
- }
- buf = (char*) e->u.octet_aligned->buf;
- len = e->u.octet_aligned->len;
- rec = ingest_record(cl, buf, len);
+ rec = ingest_record(cl, npr->u.databaseRecord);
if (!rec)
continue;
}
if (s->watchlist[SESSION_WATCH_RECORDS].fun && rlist->num_records)
session_alert_watch(s, SESSION_WATCH_RECORDS);
-}
-xsltStylesheetPtr load_stylesheet(const char *fname)
-{
- xsltStylesheetPtr ret;
- if (!(ret = xsltParseStylesheetFile((const xmlChar *) fname)))
- {
- yaz_log(YLOG_FATAL|YLOG_ERRNO, "Failed to load stylesheet %s", fname);
- exit(1);
- }
- return ret;
+#if USE_TIMING
+ yaz_timing_stop(t);
+ yaz_log(YLOG_LOG, "ingest_records %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
}
static void do_presentResponse(IOCHAN i, Z_APDU *a)
if (!*r->presentStatus && cl->state != Client_Error)
{
yaz_log(YLOG_DEBUG, "Good Present response");
- cl->records += *r->numberOfRecordsReturned;
ingest_records(cl, r->records);
cl->state = Client_Idle;
}
if (!(addr = cs_straddr(link, cl->database->host->ipport)))
{
- yaz_log(YLOG_WARN|YLOG_ERRNO, "Lookup of IP address failed?");
+ yaz_log(YLOG_WARN|YLOG_ERRNO, "Lookup of IP address %s failed?",
+ cl->database->host->ipport);
return 0;
}
if (co)
{
if (co->state == Conn_Connecting)
+ {
cl->state = Client_Connecting;
+ iochan_setflag(co->iochan, EVENT_OUTPUT);
+ }
else if (co->state == Conn_Open)
{
if (cl->state == Client_Error || cl->state == Client_Disconnected)
cl->state = Client_Idle;
+ iochan_setflag(co->iochan, EVENT_OUTPUT);
}
- iochan_setflag(co->iochan, EVENT_OUTPUT);
return 1;
}
else
return 0;
}
+#ifdef GAGA // Moved to database.c
+
+// This function will most likely vanish when a proper target profile mechanism is
+// introduced.
void load_simpletargets(const char *fn)
{
FILE *f = fopen(fn, "r");
while (fgets(line, 255, f))
{
char *url, *db;
+ char *name;
struct host *host;
struct database *database;
if (strncmp(line, "target ", 7))
continue;
+ line[strlen(line) - 1] = '\0';
+
+ if ((name = strchr(line, ';')))
+ *(name++) = '\0';
+
url = line + 7;
- url[strlen(url) - 1] = '\0';
- yaz_log(YLOG_DEBUG, "Target: %s", url);
if ((db = strchr(url, '/')))
*(db++) = '\0';
else
db = "Default";
+ yaz_log(YLOG_LOG, "Target: %s, '%s'", url, db);
for (host = hosts; host; host = host->next)
if (!strcmp(url, host->hostport))
break;
}
assert(addrinfo->ai_family == PF_INET);
memcpy(addrbuf, &((struct sockaddr_in*)addrinfo->ai_addr)->sin_addr.s_addr, 4);
- sprintf(ipport, "%hhd.%hhd.%hhd.%hhd:%s",
+ sprintf(ipport, "%u.%u.%u.%u:%s",
addrbuf[0], addrbuf[1], addrbuf[2], addrbuf[3], port);
host->ipport = xstrdup(ipport);
freeaddrinfo(addrinfo);
strcpy(database->url, url);
strcat(database->url, "/");
strcat(database->url, db);
+ if (name)
+ database->name = xstrdup(name);
+ else
+ database->name = 0;
database->databases = xmalloc(2 * sizeof(char *));
database->databases[0] = xstrdup(db);
database->databases[1] = 0;
database->errors = 0;
+ database->qprofile = 0;
+ database->rprofile = database_retrieval_profile(database);
database->next = databases;
databases = database;
fclose(f);
}
+#endif
+
static void pull_terms(NMEM nmem, struct ccl_rpn_node *n, char **termlist, int *num)
{
switch (n->kind)
s->watchlist[what].data = 0;
}
+//callback for grep_databases
+static void select_targets_callback(void *context, struct database *db)
+{
+ struct session *se = (struct session*) context;
+ struct client *cl = client_create();
+ cl->database = db;
+ cl->session = se;
+ cl->next = se->clients;
+ se->clients = cl;
+}
+
// This should be extended with parameters to control selection criteria
// Associates a set of clients with a session;
-int select_targets(struct session *se)
+int select_targets(struct session *se, struct database_criterion *crit)
{
- struct database *db;
- int c = 0;
-
while (se->clients)
client_destroy(se->clients);
- for (db = databases; db; db = db->next)
+
+ return grep_databases(se, crit, select_targets_callback);
+}
+
+int session_active_clients(struct session *s)
+{
+ struct client *c;
+ int res = 0;
+
+ for (c = s->clients; c; c = c->next)
+ if (c->connection && (c->state == Client_Connecting ||
+ c->state == Client_Initializing ||
+ c->state == Client_Searching ||
+ c->state == Client_Presenting))
+ res++;
+
+ return res;
+}
+
+// parses crit1=val1,crit2=val2,...
+static struct database_criterion *parse_filter(NMEM m, const char *buf)
+{
+ struct database_criterion *res = 0;
+ char **values;
+ int num;
+ int i;
+
+ if (!buf || !*buf)
+ return 0;
+ nmem_strsplit(m, ",", buf, &values, &num);
+ for (i = 0; i < num; i++)
{
- struct client *cl = client_create();
- cl->database = db;
- cl->session = se;
- cl->next = se->clients;
- se->clients = cl;
- c++;
+ struct database_criterion *new = nmem_malloc(m, sizeof(*new));
+ char *eq = strchr(values[i], '=');
+ if (!eq)
+ {
+ yaz_log(YLOG_WARN, "Missing equal-sign in filter");
+ return 0;
+ }
+ *(eq++) = '\0';
+ new->name = values[i];
+ new->value = eq;
+ new->next = res;
+ res = new;
}
- return c;
+ return res;
}
-char *search(struct session *se, char *query)
+char *search(struct session *se, char *query, char *filter)
{
int live_channels = 0;
struct client *cl;
+ struct database_criterion *criteria;
yaz_log(YLOG_DEBUG, "Search");
+ nmem_reset(se->nmem);
+ criteria = parse_filter(se->nmem, filter);
strcpy(se->query, query);
se->requestid++;
- nmem_reset(se->nmem);
+ // Release any existing clients
+ select_targets(se, criteria);
for (cl = se->clients; cl; cl = cl->next)
{
- cl->hits = -1;
- cl->records = 0;
- cl->diagnostic = 0;
-
if (client_prep_connection(cl))
live_channels++;
}
{
char *p[512];
int maxrecs = live_channels * global_parameters.toget;
- se->termlist = termlist_create(se->nmem, maxrecs, 15);
+ se->num_termlists = 0;
se->reclist = reclist_create(se->nmem, maxrecs);
extract_terms(se->nmem, query, p);
se->relevance = relevance_create(se->nmem, (const char **) p, maxrecs);
- se->total_records = se->total_hits = 0;
+ se->total_records = se->total_hits = se->total_merged = 0;
+ se->expected_maxrecs = maxrecs;
}
else
return "NOTARGETS";
while (s->clients)
client_destroy(s->clients);
nmem_destroy(s->nmem);
- wrbuf_free(s->wrbuf, 1);
+ wrbuf_destroy(s->wrbuf);
}
struct session *new_session()
session->total_hits = 0;
session->total_records = 0;
- session->termlist = 0;
+ session->num_termlists = 0;
session->reclist = 0;
session->requestid = -1;
session->clients = 0;
+ session->expected_maxrecs = 0;
session->query[0] = '\0';
session->nmem = nmem_create();
session->wrbuf = wrbuf_alloc();
session->watchlist[i].fun = 0;
}
- select_targets(session);
-
return session;
}
*count = 0;
for (cl = se->clients; cl; cl = cl->next)
{
- strcpy(res[*count].id, cl->database->host->hostport);
+ res[*count].id = cl->database->url;
+ res[*count].name = cl->database->name;
res[*count].hits = cl->hits;
res[*count].records = cl->records;
res[*count].diagnostic = cl->diagnostic;
return res;
}
-struct termlist_score **termlist(struct session *s, int *num)
+struct termlist_score **termlist(struct session *s, const char *name, int *num)
{
- return termlist_highscore(s->termlist, num);
+ int i;
+
+ for (i = 0; i < s->num_termlists; i++)
+ if (!strcmp(s->termlists[i].name, name))
+ return termlist_highscore(s->termlists[i].termlist, num);
+ return 0;
}
-#ifdef REPORT_NMEM
-// conditional compilation by SH: This lead to a warning with currently installed
-// YAZ header files on us1
+#ifdef MISSING_HEADERS
void report_nmem_stats(void)
{
size_t in_use, is_free;
}
#endif
-struct record **show(struct session *s, int start, int *num, int *total,
- int *sumhits, NMEM nmem_show)
+struct record_cluster *show_single(struct session *s, int id)
{
- struct record **recs = nmem_malloc(nmem_show, *num
- * sizeof(struct record *));
+ struct record_cluster *r;
+
+ reclist_rewind(s->reclist);
+ while ((r = reclist_read_record(s->reclist)))
+ if (r->recid == id)
+ return r;
+ return 0;
+}
+
+struct record_cluster **show(struct session *s, struct reclist_sortparms *sp, int start,
+ int *num, int *total, int *sumhits, NMEM nmem_show)
+{
+ struct record_cluster **recs = nmem_malloc(nmem_show, *num
+ * sizeof(struct record_cluster *));
+ struct reclist_sortparms *spp;
int i;
+#if USE_TIMING
+ yaz_timing_t t = yaz_timing_create();
+#endif
- relevance_prepare_read(s->relevance, s->reclist);
+ for (spp = sp; spp; spp = spp->next)
+ if (spp->type == Metadata_sortkey_relevance)
+ {
+ relevance_prepare_read(s->relevance, s->reclist);
+ break;
+ }
+ reclist_sort(s->reclist, sp);
*total = s->reclist->num_records;
*sumhits = s->total_hits;
if (!reclist_read_record(s->reclist))
{
*num = 0;
- return 0;
+ recs = 0;
+ break;
}
for (i = 0; i < *num; i++)
{
- struct record *r = reclist_read_record(s->reclist);
+ struct record_cluster *r = reclist_read_record(s->reclist);
if (!r)
{
*num = i;
}
recs[i] = r;
}
+#if USE_TIMING
+ yaz_timing_stop(t);
+ 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);
+#endif
return recs;
}
struct client *cl;
int count = 0;
- bzero(stat, sizeof(*stat));
+ memset(stat, 0, sizeof(*stat));
for (cl = se->clients; cl; cl = cl->next)
{
if (!cl->connection)
return res;
}
+static void start_http_listener(void)
+{
+ char hp[128] = "";
+ struct conf_server *ser = global_parameters.server;
+
+ if (*global_parameters.listener_override)
+ strcpy(hp, global_parameters.listener_override);
+ else
+ {
+ strcpy(hp, ser->host ? ser->host : "");
+ if (ser->port)
+ {
+ if (*hp)
+ strcat(hp, ":");
+ sprintf(hp + strlen(hp), "%d", ser->port);
+ }
+ }
+ http_init(hp);
+}
+
+static void start_proxy(void)
+{
+ char hp[128] = "";
+ struct conf_server *ser = global_parameters.server;
+
+ if (*global_parameters.proxy_override)
+ strcpy(hp, global_parameters.proxy_override);
+ else if (ser->proxy_host || ser->proxy_port)
+ {
+ strcpy(hp, ser->proxy_host ? ser->proxy_host : "");
+ if (ser->proxy_port)
+ {
+ if (*hp)
+ strcat(hp, ":");
+ sprintf(hp + strlen(hp), "%d", ser->proxy_port);
+ }
+ }
+ else
+ return;
+
+ http_set_proxyaddr(hp, ser->myurl ? ser->myurl : "");
+}
+
int main(int argc, char **argv)
{
int ret;
char *arg;
- int setport = 0;
- if (signal(SIGPIPE, SIG_IGN) < 0)
+ if (signal(SIGPIPE, SIG_IGN) == SIG_ERR)
yaz_log(YLOG_WARN|YLOG_ERRNO, "signal");
yaz_log_init(YLOG_DEFAULT_LEVEL, "pazpar2", 0);
- while ((ret = options("f:x:c:h:p:C:s:", argv, argc, &arg)) != -2)
+ while ((ret = options("f:x:h:p:C:s:d", argv, argc, &arg)) != -2)
{
switch (ret) {
case 'f':
if (!read_config(arg))
exit(1);
break;
- case 'c':
- command_init(atoi(arg));
- setport++;
- break;
case 'h':
- http_init(arg);
- setport++;
+ strcpy(global_parameters.listener_override, arg);
break;
case 'C':
global_parameters.ccl_filter = load_cclfile(arg);
break;
case 'p':
- http_set_proxyaddr(arg);
+ strcpy(global_parameters.proxy_override, arg);
break;
case 's':
load_simpletargets(arg);
break;
- case 'x':
- global_parameters.xsl = load_stylesheet(arg);
+ case 'd':
+ global_parameters.dump_records = 1;
break;
default:
fprintf(stderr, "Usage: pazpar2\n"
" -f configfile\n"
" -h [host:]port (REST protocol listener)\n"
- " -c cmdport (telnet-style)\n"
" -C cclconfig\n"
" -s simpletargetfile\n"
- " -p hostname[:portno] (HTTP proxy)\n");
+ " -p hostname[:portno] (HTTP proxy)\n"
+ " -d (show internal records)\n");
exit(1);
}
}
- if (!setport)
+ if (!config)
{
- fprintf(stderr, "Set command port with -h or -c\n");
+ yaz_log(YLOG_FATAL, "Load config with -f");
exit(1);
}
+ global_parameters.server = config->servers;
- if (!global_parameters.xsl)
- global_parameters.xsl = load_stylesheet("../etc/default.xsl");
- global_parameters.ccl_filter = load_cclfile("../etc/default.bib");
+ start_http_listener();
+ start_proxy();
+ if (!global_parameters.ccl_filter)
+ global_parameters.ccl_filter = load_cclfile("../etc/default.bib");
global_parameters.yaz_marc = yaz_marc_create();
yaz_marc_subfield_str(global_parameters.yaz_marc, "\t");
global_parameters.odr_in = odr_createmem(ODR_DECODE);
projects / pazpar2-moved-to-github.git / blobdiff