/* This file is part of the Zebra server.
- Copyright (C) 1995-2008 Index Data
+ Copyright (C) 1994-2011 Index Data
Zebra 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
*/
+#if HAVE_CONFIG_H
+#include <config.h>
+#endif
#include <stdio.h>
#include <assert.h>
#ifdef WIN32
rpn_get_top_approx_limit(zh, rpn->RPNStructure, &sset->approx_limit);
res = rpn_search_top(zh, rpn->RPNStructure, rpn->attributeSetId,
+ sset->approx_limit,
nmem, rset_nmem,
sort_sequence,
sset->num_bases, sset->basenames,
for (i = 0; sort_sequence->specs[i]; i++)
;
sort_sequence->num_specs = i;
- rset->hits_limit = sset->approx_limit;
+ rset_set_hits_limit(rset, sset->approx_limit);
+
if (!i)
{
res = resultSetRank(zh, sset, rset, rset_nmem);
};
void resultSetInsertSort(ZebraHandle zh, ZebraSet sset,
+ int database_no,
struct sortKeyInfo *criteria, int num_criteria,
zint sysno,
- char *cmp_buf[], char *tmp_cmp_buf[], int *cached_success_db)
+ char *cmp_buf[], char *tmp_cmp_buf[])
{
struct zset_sort_entry *new_entry = NULL;
struct zset_sort_info *sort_info = sset->sort_info;
int i, j;
- int scan_db,scan_count;
- int numbases = zh->num_basenames;
+ WRBUF w = wrbuf_alloc();
zebra_sort_sysno(zh->reg->sort_index, sysno);
for (i = 0; i<num_criteria; i++)
char *this_entry_buf = tmp_cmp_buf[i];
memset(this_entry_buf, '\0', SORT_IDX_ENTRYSIZE);
- /* if the first database doesn't have a sort index,
- we assume none of them will */
- if (criteria[i].ord[0] != -1)
+ if (criteria[i].ord[database_no] != -1)
{
- /* now make a best guess for the database in which we think
- the record is located if its not in our best guess, try the
- other databases one by one, till we had them all */
- scan_db = *cached_success_db;
- scan_count = 0;
-
- while (1)
+ yaz_log(log_level_sort, "pre zebra_sort_type ord is %d",
+ criteria[i].ord[database_no]);
+ zebra_sort_type(zh->reg->sort_index, criteria[i].ord[database_no]);
+ wrbuf_rewind(w);
+ if (zebra_sort_read(zh->reg->sort_index, 0, w))
{
- scan_count++;
- if (scan_count>numbases)
- {
- /* well...we scanned all databases and still nothing...give up */
- yaz_log(log_level_sort, "zebra_sort_read failed (record not found in indices)");
- break;
- }
-
- /* the criteria[i].ord is the file id of the sort index */
- yaz_log(log_level_sort, "pre zebra_sort_type ord is %d", criteria[i].ord[scan_db]);
- zebra_sort_type(zh->reg->sort_index, criteria[i].ord[scan_db]);
- if (zebra_sort_read(zh->reg->sort_index, this_entry_buf))
- {
- /* allright, found it */
- /* cache this db so we start trying from this db
- for next record */
- *cached_success_db=scan_db;
- break;
- }
- else
+ /* consider each sort entry and take lowest/highest one
+ of the one as sorting key depending on whether sort is
+ ascending/descending */
+ int off = 0;
+ while (off != wrbuf_len(w))
{
- yaz_log(log_level_sort, "record not found in database, trying next one");
- scan_db++;
- if (scan_db>=numbases)
- scan_db=0;
+ size_t l = strlen(wrbuf_buf(w)+off);
+ assert(off < wrbuf_len(w));
+
+ if (l >= SORT_IDX_ENTRYSIZE)
+ l = SORT_IDX_ENTRYSIZE-1;
+ if ( (off == 0)
+ || (criteria[i].relation == 'A'
+ && strcmp(wrbuf_buf(w)+off, this_entry_buf) < 0)
+ || (criteria[i].relation == 'D'
+ && strcmp(wrbuf_buf(w)+off, this_entry_buf) > 0)
+ )
+ {
+ memcpy(this_entry_buf, wrbuf_buf(w)+off, l);
+ this_entry_buf[l] = '\0';
+ }
+ off += 1 + strlen(wrbuf_buf(w)+off);
}
}
-
}
else
{
yaz_log(log_level_sort, "criteria[i].ord is -1 so not reading from sort index");
}
}
+ wrbuf_destroy(w);
i = sort_info->num_entries;
while (--i >= 0)
{
char *this_entry_buf = tmp_cmp_buf[j];
char *other_entry_buf =
cmp_buf[j] + i * SORT_IDX_ENTRYSIZE;
- if (criteria[j].numerical[*cached_success_db])
+ if (criteria[j].numerical[database_no])
{
char this_entry_org[1024];
char other_entry_org[1024];
double diff;
- /* when searching multiple databases, we use the index
- type of the first one. So if they differ between
- databases, we have a problem here we could store the
- index_type for each database, but if we didn't find the
- record in any sort index, then we still don't know to
- which database it belongs. */
const char *index_type = criteria[j].index_type;
zebra_term_untrans(zh, index_type, this_entry_org,
this_entry_buf);
{
int i;
int ib;
- int cached_success_db = 0;
int n = 0;
zint kno = 0;
zint psysno = 0;
{
Z_SortKeySpec *sks = sort_sequence->specs[i];
Z_SortKey *sk;
- ZEBRA_RES res;
sort_criteria[i].ord = (int *)
nmem_malloc(nmem, sizeof(int)*numbases);
for (ib = 0; ib < numbases; ib++)
{
zebraExplain_curDatabase(zh->reg->zei, zh->basenames[ib]);
- res = zebra_sort_get_ord(zh, sk->u.sortAttributes,
- &sort_criteria[i].ord[ib],
- &sort_criteria[i].numerical[ib]);
+ if (zebra_sort_get_ord(zh, sk->u.sortAttributes,
+ &sort_criteria[i].ord[ib],
+ &sort_criteria[i].numerical[ib]) !=
+ ZEBRA_OK && sks->which != Z_SortKeySpec_null)
+ return ZEBRA_FAIL;
}
-
- if (sks->which != Z_SortKeySpec_null && res != ZEBRA_OK)
- return ZEBRA_FAIL;
break;
}
/* right now we look up the index type based on the first database
kno++;
if (this_sys != psysno)
{
+ int database_no = 0;
if ((sset->hits & 255) == 0 && zh->break_handler_func)
{
if (zh->break_handler_func(zh->break_handler_data))
}
(sset->hits)++;
psysno = this_sys;
- resultSetInsertSort(zh, sset,
+
+ /* determine database from the term, but only bother if more than
+ one database is in use*/
+ if (numbases > 1 && termid->ol)
+ {
+ const char *this_db = 0;
+ if (zebraExplain_lookup_ord(zh->reg->zei, termid->ol->ord, 0, &this_db, 0)
+ == 0 && this_db)
+ {
+ for (ib = 0; ib < numbases; ib++)
+ if (!strcmp(this_db, zh->basenames[ib]))
+ database_no = ib;
+ }
+ }
+#if 0
+ yaz_log(YLOG_LOG, "sysno=" ZINT_FORMAT " database_no=%d", this_sys,
+ database_no);
+ ord_list_print(termid->ol);
+#endif
+ resultSetInsertSort(zh, sset, database_no,
sort_criteria, num_criteria, psysno, cmp_buf,
- tmp_cmp_buf, &cached_success_db);
+ tmp_cmp_buf);
}
}
rset_close(rfd);
}
if (rfd->counted_items > rset->hits_limit)
stop_flag = 1;
+ if (stop_flag)
+ {
+ zebraSet->estimated_hit_count = 1;
+ break;
+ }
if (psysno)
{ /* only if we did have a previous record */
score = (*rc->calc)(handle, psysno, pstaticrank,
resultSetInsertRank(zh, sort_info, psysno, score, 'A');
count++;
}
- if (stop_flag)
- {
- zebraSet->estimated_hit_count = 1;
- rset_set_hits_limit(rset, 0);
- break;
- }
psysno = this_sys;
if (zh->m_staticrank)
pstaticrank = key.mem[0];
/*
* Local variables:
* c-basic-offset: 4
+ * c-file-style: "Stroustrup"
* indent-tabs-mode: nil
* End:
* vim: shiftwidth=4 tabstop=8 expandtab
projects / idzebra-moved-to-github.git / blobdiff