X-Git-Url: http://git.indexdata.com/?p=idzebra-moved-to-github.git;a=blobdiff_plain;f=index%2Fzsets.c;h=4a8c74c39c8c26f6638ee1fd0e89069a0b05138a;hp=e77e5fbe744a1c1cf4a1b6f68ed4a452f26fc257;hb=d87f70faee1aeead5d32c388bfde8c41e859fb2f;hpb=c5bc0f71fe718b3c66d86b48b1b8669278e94c18 diff --git a/index/zsets.c b/index/zsets.c index e77e5fb..4a8c74c 100644 --- a/index/zsets.c +++ b/index/zsets.c @@ -1,8 +1,5 @@ -/* $Id: zsets.c,v 1.65 2004-10-21 12:43:09 heikki Exp $ - Copyright (C) 1995,1996,1997,1998,1999,2000,2001,2002,2003,2004 - Index Data Aps - -This file is part of the Zebra server. +/* This file is part of the Zebra server. + 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 @@ -15,9 +12,9 @@ FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License -along with Zebra; see the file LICENSE.zebra. If not, write to the -Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA -02111-1307, USA. +along with this program; if not, write to the Free Software +Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ @@ -30,18 +27,19 @@ Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA #endif #include "index.h" +#include "rank.h" +#include #include -#define SORT_IDX_ENTRYSIZE 64 -#define ZSET_SORT_MAX_LEVEL 3 +#define ZSET_SORT_MAX_LEVEL 10 struct zebra_set_term_entry { int reg_type; char *db; - int set; - int use; + char *index_name; char *term; }; + struct zebra_set { char *name; RSET rset; @@ -49,19 +47,25 @@ struct zebra_set { NMEM rset_nmem; /* for creating the rsets in */ zint hits; int num_bases; - char **basenames; + const char **basenames; Z_RPNQuery *rpn; + Z_SortKeySpecList *sortSpec; struct zset_sort_info *sort_info; struct zebra_set_term_entry *term_entries; int term_entries_max; struct zebra_set *next; int locked; + int estimated_hit_count; + + zint cache_position; /* last position */ + RSFD cache_rfd; /* rfd (NULL if not existing) */ + zint cache_psysno; /* sysno for last position */ + zint approx_limit; /* limit before we do approx */ }; struct zset_sort_entry { zint sysno; int score; - char buf[ZSET_SORT_MAX_LEVEL][SORT_IDX_ENTRYSIZE]; }; struct zset_sort_info { @@ -71,169 +75,189 @@ struct zset_sort_info { struct zset_sort_entry **entries; }; -ZebraSet resultSetAddRPN (ZebraHandle zh, NMEM m, - Z_RPNQuery *rpn, int num_bases, - char **basenames, - const char *setname) +static int log_level_set=0; +static int log_level_sort=0; +static int log_level_searchhits=0; +static int log_level_searchterms=0; +static int log_level_resultsets=0; + +static void loglevels(void) +{ + if (log_level_set) + return; + log_level_sort = yaz_log_module_level("sorting"); + log_level_searchhits = yaz_log_module_level("searchhits"); + log_level_searchterms = yaz_log_module_level("searchterms"); + log_level_resultsets = yaz_log_module_level("resultsets"); + log_level_set = 1; +} + + +static ZEBRA_RES resultSetSearch(ZebraHandle zh, NMEM nmem, NMEM rset_nmem, + Z_RPNQuery *rpn, ZebraSet sset) +{ + RSET rset = 0; + Z_SortKeySpecList *sort_sequence; + int sort_status, i; + ZEBRA_RES res = ZEBRA_OK; + + sort_sequence = (Z_SortKeySpecList *) + nmem_malloc(nmem, sizeof(*sort_sequence)); + sort_sequence->num_specs = 10; /* FIXME - Hard-coded number */ + sort_sequence->specs = (Z_SortKeySpec **) + nmem_malloc(nmem, sort_sequence->num_specs * + sizeof(*sort_sequence->specs)); + for (i = 0; inum_specs; i++) + sort_sequence->specs[i] = 0; + + 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, + &rset); + if (res != ZEBRA_OK) + { + sset->rset = 0; + return res; + } + for (i = 0; sort_sequence->specs[i]; i++) + ; + sort_sequence->num_specs = i; + rset_set_hits_limit(rset, sset->approx_limit); + + if (!i) + { + res = resultSetRank(zh, sset, rset, rset_nmem); + } + else + { + res = resultSetSortSingle(zh, nmem, sset, rset, + sort_sequence, &sort_status); + } + sset->rset = rset; + return res; +} + + +ZEBRA_RES resultSetAddRPN(ZebraHandle zh, NMEM m, Z_RPNQuery *rpn, + int num_bases, char **basenames, + const char *setname, + zint *hits, int *estimated_hit_count) { ZebraSet zebraSet; int i; + ZEBRA_RES res; - zh->errCode = 0; - zh->errString = NULL; - zh->hits = 0; + *hits = 0; + *estimated_hit_count = 0; - zebraSet = resultSetAdd (zh, setname, 1); + zebraSet = resultSetAdd(zh, setname, 1); if (!zebraSet) - return 0; + return ZEBRA_FAIL; zebraSet->locked = 1; zebraSet->rpn = 0; zebraSet->nmem = m; - zebraSet->rset_nmem=nmem_create(); + zebraSet->rset_nmem = nmem_create(); zebraSet->num_bases = num_bases; zebraSet->basenames = - nmem_malloc (zebraSet->nmem, num_bases * sizeof(*zebraSet->basenames)); + nmem_malloc(zebraSet->nmem, num_bases * sizeof(*zebraSet->basenames)); for (i = 0; ibasenames[i] = nmem_strdup (zebraSet->nmem, basenames[i]); + zebraSet->basenames[i] = nmem_strdup(zebraSet->nmem, basenames[i]); + res = resultSetSearch(zh, zebraSet->nmem, zebraSet->rset_nmem, + rpn, zebraSet); + *hits = zebraSet->hits; + if (zebraSet->estimated_hit_count) + *estimated_hit_count = 1; - zebraSet->rset = rpn_search (zh, zebraSet->nmem, zebraSet->rset_nmem, - rpn, zebraSet->num_bases, - zebraSet->basenames, zebraSet->name, - zebraSet); - zh->hits = zebraSet->hits; if (zebraSet->rset) zebraSet->rpn = rpn; zebraSet->locked = 0; - return zebraSet; + if (!zebraSet->rset) + return ZEBRA_FAIL; + return res; } -void resultSetAddTerm (ZebraHandle zh, ZebraSet s, int reg_type, - const char *db, int set, - int use, const char *term) +void resultSetAddTerm(ZebraHandle zh, ZebraSet s, int reg_type, + const char *db, const char *index_name, + const char *term) { assert(zh); /* compiler shut up */ if (!s->nmem) - s->nmem = nmem_create (); + s->nmem = nmem_create(); if (!s->term_entries) { int i; s->term_entries_max = 1000; s->term_entries = - nmem_malloc (s->nmem, s->term_entries_max * - sizeof(*s->term_entries)); + nmem_malloc(s->nmem, s->term_entries_max * + sizeof(*s->term_entries)); for (i = 0; i < s->term_entries_max; i++) s->term_entries[i].term = 0; } if (s->hits < s->term_entries_max) { s->term_entries[s->hits].reg_type = reg_type; - s->term_entries[s->hits].db = nmem_strdup (s->nmem, db); - s->term_entries[s->hits].set = set; - s->term_entries[s->hits].use = use; - s->term_entries[s->hits].term = nmem_strdup (s->nmem, term); + s->term_entries[s->hits].db = nmem_strdup(s->nmem, db); + s->term_entries[s->hits].index_name = nmem_strdup(s->nmem, index_name); + s->term_entries[s->hits].term = nmem_strdup(s->nmem, term); } (s->hits)++; } -#if 0 /* FIXME - Delete this, we don't count terms no more */ -int zebra_resultSetTerms (ZebraHandle zh, const char *setname, - int no, zint *count, - int *type, char *out, size_t *len) -{ - ZebraSet s = resultSetGet (zh, setname); - int no_max = 0; - - if (count) - *count = 0; - if (!s || !s->rset) - return 0; - no_max = s->rset->no_rset_terms; - if (no < 0 || no >= no_max) - return 0; - if (count) - *count = s->rset->rset_terms[no]->count; - if (type) - *type = s->rset->rset_terms[no]->type; - - if (out) - { - char *inbuf = s->rset->rset_terms[no]->name; - size_t inleft = strlen(inbuf); - size_t outleft = *len - 1; - int converted = 0; - - if (zh->iconv_from_utf8 != 0) - { - char *outbuf = out; - size_t ret; - - ret = yaz_iconv(zh->iconv_from_utf8, &inbuf, &inleft, - &outbuf, &outleft); - if (ret == (size_t)(-1)) - *len = 0; - else - *len = outbuf - out; - converted = 1; - } - if (!converted) - { - if (inleft > outleft) - inleft = outleft; - *len = inleft; - memcpy (out, inbuf, *len); - } - out[*len] = 0; - } - return no_max; -} - -#endif - -ZebraSet resultSetAdd (ZebraHandle zh, const char *name, int ov) +ZebraSet resultSetAdd(ZebraHandle zh, const char *name, int ov) { ZebraSet s; int i; for (s = zh->sets; s; s = s->next) - if (!strcmp (s->name, name)) + if (!strcmp(s->name, name)) break; + + if (!log_level_set) + loglevels(); if (s) { - yaz_log (LOG_DEBUG, "updating result set %s", name); + yaz_log(log_level_resultsets, "updating result set %s", name); if (!ov || s->locked) return NULL; if (s->rset) - rset_delete (s->rset); + { + if (s->cache_rfd) + rset_close(s->cache_rfd); + rset_delete(s->rset); + } if (s->rset_nmem) - nmem_destroy (s->rset_nmem); + nmem_destroy(s->rset_nmem); if (s->nmem) - nmem_destroy (s->nmem); + nmem_destroy(s->nmem); } else { const char *sort_max_str = zebra_get_resource(zh, "sortmax", "1000"); - yaz_log (LOG_DEBUG, "adding result set %s", name); - s = (ZebraSet) xmalloc (sizeof(*s)); + yaz_log(log_level_resultsets, "adding result set %s", name); + s = (ZebraSet) xmalloc(sizeof(*s)); s->next = zh->sets; zh->sets = s; - s->name = (char *) xmalloc (strlen(name)+1); - strcpy (s->name, name); + s->name = xstrdup(name); s->sort_info = (struct zset_sort_info *) - xmalloc (sizeof(*s->sort_info)); + xmalloc(sizeof(*s->sort_info)); s->sort_info->max_entries = atoi(sort_max_str); if (s->sort_info->max_entries < 2) s->sort_info->max_entries = 2; s->sort_info->entries = (struct zset_sort_entry **) - xmalloc (sizeof(*s->sort_info->entries) * - s->sort_info->max_entries); + xmalloc(sizeof(*s->sort_info->entries) * + s->sort_info->max_entries); s->sort_info->all_entries = (struct zset_sort_entry *) - xmalloc (sizeof(*s->sort_info->all_entries) * - s->sort_info->max_entries); + xmalloc(sizeof(*s->sort_info->all_entries) * + s->sort_info->max_entries); for (i = 0; i < s->sort_info->max_entries; i++) s->sort_info->entries[i] = s->sort_info->all_entries + i; } @@ -241,51 +265,80 @@ ZebraSet resultSetAdd (ZebraHandle zh, const char *name, int ov) s->term_entries = 0; s->hits = 0; s->rset = 0; - s->rset_nmem=0; + s->rset_nmem = 0; s->nmem = 0; s->rpn = 0; + s->sortSpec = 0; + s->cache_position = 0; + s->cache_rfd = 0; + s->approx_limit = zh->approx_limit; + s->estimated_hit_count = 0; return s; } -ZebraSet resultSetGet (ZebraHandle zh, const char *name) +ZebraSet resultSetGet(ZebraHandle zh, const char *name) { ZebraSet s; for (s = zh->sets; s; s = s->next) - if (!strcmp (s->name, name)) + if (!strcmp(s->name, name)) { if (!s->term_entries && !s->rset && s->rpn) { - NMEM nmem = nmem_create (); - yaz_log (LOG_LOG, "research %s", name); + NMEM nmem = nmem_create(); + yaz_log(log_level_resultsets, "research %s", name); if (!s->rset_nmem) - s->rset_nmem=nmem_create(); - s->rset = - rpn_search (zh, nmem, s->rset_nmem, s->rpn, s->num_bases, - s->basenames, s->name, s); - nmem_destroy (nmem); + s->rset_nmem = nmem_create(); + resultSetSearch(zh, nmem, s->rset_nmem, s->rpn, s); + if (s->rset && s->sortSpec) + { + int sort_status; + yaz_log(log_level_resultsets, "resort %s", name); + resultSetSortSingle(zh, nmem, s, s->rset, s->sortSpec, + &sort_status); + } + nmem_destroy(nmem); } return s; } return NULL; } -void resultSetInvalidate (ZebraHandle zh) +ZEBRA_RES resultSetGetBaseNames(ZebraHandle zh, const char *setname, + const char ***basenames, int *num_bases) +{ + ZebraSet sset = resultSetGet(zh, setname); + if (!sset) + return ZEBRA_FAIL; + *basenames = sset->basenames; + *num_bases = sset->num_bases; + return ZEBRA_OK; + +} + +void resultSetInvalidate(ZebraHandle zh) { ZebraSet s = zh->sets; + yaz_log(log_level_resultsets, "invalidating result sets"); for (; s; s = s->next) { if (s->rset) - rset_delete (s->rset); + { + if (s->cache_rfd) + rset_close(s->cache_rfd); + rset_delete(s->rset); + } s->rset = 0; + s->cache_rfd = 0; + s->cache_position = 0; if (s->rset_nmem) nmem_destroy(s->rset_nmem); s->rset_nmem=0; } } -void resultSetDestroy (ZebraHandle zh, int num, char **names,int *statuses) +void resultSetDestroy(ZebraHandle zh, int num, char **names,int *statuses) { ZebraSet * ss = &zh->sets; int i; @@ -293,8 +346,6 @@ void resultSetDestroy (ZebraHandle zh, int num, char **names,int *statuses) if (statuses) for (i = 0; ierrCode = 0; - zh->errString = NULL; while (*ss) { int i = -1; @@ -302,7 +353,7 @@ void resultSetDestroy (ZebraHandle zh, int num, char **names,int *statuses) if (num >= 0) { for (i = 0; iname, names[i])) + if (!strcmp(s->name, names[i])) { if (statuses) statuses[i] = Z_DeleteStatus_success; @@ -314,40 +365,75 @@ void resultSetDestroy (ZebraHandle zh, int num, char **names,int *statuses) { *ss = s->next; - xfree (s->sort_info->all_entries); - xfree (s->sort_info->entries); - xfree (s->sort_info); + xfree(s->sort_info->all_entries); + xfree(s->sort_info->entries); + xfree(s->sort_info); if (s->nmem) - nmem_destroy (s->nmem); + nmem_destroy(s->nmem); if (s->rset) - rset_delete (s->rset); + { + if (s->cache_rfd) + rset_close(s->cache_rfd); + rset_delete(s->rset); + } if (s->rset_nmem) nmem_destroy(s->rset_nmem); - xfree (s->name); - xfree (s); + xfree(s->name); + xfree(s); } else ss = &s->next; } } -ZebraPosSet zebraPosSetCreate (ZebraHandle zh, const char *name, - int num, int *positions) +ZebraMetaRecord *zebra_meta_records_create_range(ZebraHandle zh, + const char *name, + zint start, int num) +{ + zint pos_small[10]; + zint *pos = pos_small; + ZebraMetaRecord *mr; + int i; + + if (num > 10000 || num <= 0) + return 0; + + if (num > 10) + pos = xmalloc(sizeof(*pos) * num); + + for (i = 0; i 10) + xfree(pos); + return mr; +} + +ZebraMetaRecord *zebra_meta_records_create(ZebraHandle zh, const char *name, + int num, zint *positions) { ZebraSet sset; - ZebraPosSet sr = 0; + ZebraMetaRecord *sr = 0; RSET rset; int i; struct zset_sort_info *sort_info; + size_t sysno_mem_index = 0; - if (!(sset = resultSetGet (zh, name))) + if (zh->m_staticrank) + sysno_mem_index = 1; + + if (!log_level_set) + loglevels(); + if (!(sset = resultSetGet(zh, name))) return NULL; if (!(rset = sset->rset)) { if (!sset->term_entries) return 0; - sr = (ZebraPosSet) xmalloc (sizeof(*sr) * num); + sr = (ZebraMetaRecord *) xmalloc(sizeof(*sr) * num); for (i = 0; isort_info; if (sort_info) { - int position; + zint position; for (i = 0; i 0 && position <= sort_info->num_entries) { - yaz_log (LOG_DEBUG, "got pos=%d (sorted)", position); + yaz_log(log_level_sort, "got pos=" ZINT_FORMAT + " (sorted)", position); sr[i].sysno = sort_info->entries[position-1]->sysno; sr[i].score = sort_info->entries[position-1]->score; } @@ -396,7 +483,7 @@ ZebraPosSet zebraPosSetCreate (ZebraHandle zh, const char *name, } if (i < num) /* nope, get the rest, unsorted - sorry */ { - int position = 0; + zint position = 0; int num_i = 0; zint psysno = 0; RSFD rfd; @@ -404,12 +491,25 @@ ZebraPosSet zebraPosSetCreate (ZebraHandle zh, const char *name, if (sort_info) position = sort_info->num_entries; - while (num_i < num && positions[num_i] < position) + while (num_i < num && positions[num_i] <= position) num_i++; - rfd = rset_open (rset, RSETF_READ); - while (num_i < num && rset_read (rfd, &key, 0)) + + if (sset->cache_rfd && + num_i < num && positions[num_i] > sset->cache_position) + { + position = sset->cache_position; + rfd = sset->cache_rfd; + psysno = sset->cache_psysno; + } + else + { + if (sset->cache_rfd) + rset_close(sset->cache_rfd); + rfd = rset_open(rset, RSETF_READ); + } + while (num_i < num && rset_read(rfd, &key, 0)) { - zint this_sys = key.mem[0]; + zint this_sys = key.mem[sysno_mem_index]; if (this_sys != psysno) { psysno = this_sys; @@ -423,70 +523,129 @@ ZebraPosSet zebraPosSetCreate (ZebraHandle zh, const char *name, continue; } position++; - assert (num_i < num); + assert(num_i < num); if (position == positions[num_i]) { sr[num_i].sysno = psysno; - yaz_log (LOG_DEBUG, "got pos=%d (unsorted)", position); + yaz_log(log_level_sort, "got pos=" ZINT_FORMAT " (unsorted)", position); sr[num_i].score = -1; num_i++; } } } - rset_close (rfd); + sset->cache_position = position; + sset->cache_psysno = psysno; + sset->cache_rfd = rfd; } } return sr; } -void zebraPosSetDestroy (ZebraHandle zh, ZebraPosSet records, int num) +void zebra_meta_records_destroy(ZebraHandle zh, ZebraMetaRecord *records, + int num) { assert(zh); /* compiler shut up about unused arg */ - xfree (records); + xfree(records); } struct sortKeyInfo { int relation; - int attrUse; - int numerical; + int *ord; /* array of ord for each database searched */ + int *numerical; /* array of ord for each database searched */ + const char *index_type; }; -void resultSetInsertSort (ZebraHandle zh, ZebraSet sset, - struct sortKeyInfo *criteria, int num_criteria, - zint sysno) +void resultSetInsertSort(ZebraHandle zh, ZebraSet sset, + int database_no, + struct sortKeyInfo *criteria, int num_criteria, + zint sysno, + char *cmp_buf[], char *tmp_cmp_buf[]) { - struct zset_sort_entry this_entry; struct zset_sort_entry *new_entry = NULL; struct zset_sort_info *sort_info = sset->sort_info; int i, j; + WRBUF w = wrbuf_alloc(); - sortIdx_sysno (zh->reg->sortIdx, sysno); + zebra_sort_sysno(zh->reg->sort_index, sysno); for (i = 0; ireg->sortIdx, criteria[i].attrUse); - sortIdx_read (zh->reg->sortIdx, this_entry.buf[i]); + char *this_entry_buf = tmp_cmp_buf[i]; + memset(this_entry_buf, '\0', SORT_IDX_ENTRYSIZE); + + if (criteria[i].ord[database_no] != -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)) + { + /* 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)) + { + 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) { int rel = 0; for (j = 0; jentries[i]->buf[j]); - rel = 0; + char this_entry_org[1024]; + char other_entry_org[1024]; + double diff; + const char *index_type = criteria[j].index_type; + zebra_term_untrans(zh, index_type, this_entry_org, + this_entry_buf); + zebra_term_untrans(zh, index_type, other_entry_org, + other_entry_buf); + diff = atof(this_entry_org) - atof(other_entry_org); + if (diff > 0.0) rel = 1; else if (diff < 0.0) rel = -1; + else + rel = 0; } else { - rel = memcmp (this_entry.buf[j], sort_info->entries[i]->buf[j], - SORT_IDX_ENTRYSIZE); + rel = memcmp(this_entry_buf, other_entry_buf, + SORT_IDX_ENTRYSIZE); } + /* when the compare is equal, continue to next criteria, + else break out */ if (rel) break; } @@ -504,30 +663,47 @@ void resultSetInsertSort (ZebraHandle zh, ZebraSet sset, } } ++i; + yaz_log(log_level_sort, "ok, we want to insert record at position %d",i); j = sort_info->max_entries; - if (i == j) + if (i == j){ + yaz_log(log_level_sort, "sort_info->max_entries reached (%d) abort sort",j); return; + } if (sort_info->num_entries == j) --j; else j = (sort_info->num_entries)++; new_entry = sort_info->entries[j]; + /* move up all higher entries (to make room) */ while (j != i) { + int k; + for (k = 0; kentries[j] = sort_info->entries[j-1]; --j; } + /* and insert the new entry at the correct place */ sort_info->entries[i] = new_entry; - assert (new_entry); + assert(new_entry); + /* and add this to the compare buffer */ for (i = 0; ibuf[i], this_entry.buf[i], SORT_IDX_ENTRYSIZE); + { + char *new_entry_buf = cmp_buf[i] + j * SORT_IDX_ENTRYSIZE; + char *this_entry_buf = tmp_cmp_buf[i]; + memcpy(new_entry_buf, this_entry_buf, SORT_IDX_ENTRYSIZE); + } new_entry->sysno = sysno; new_entry->score = -1; } -void resultSetInsertRank (ZebraHandle zh, struct zset_sort_info *sort_info, - zint sysno, int score, int relation) +void resultSetInsertRank(ZebraHandle zh, struct zset_sort_info *sort_info, + zint sysno, int score, int relation) { struct zset_sort_entry *new_entry = NULL; int i, j; @@ -568,270 +744,453 @@ void resultSetInsertRank (ZebraHandle zh, struct zset_sort_info *sort_info, --j; } sort_info->entries[i] = new_entry; - assert (new_entry); + assert(new_entry); new_entry->sysno = sysno; new_entry->score = score; } -void resultSetSort (ZebraHandle zh, NMEM nmem, - int num_input_setnames, const char **input_setnames, - const char *output_setname, - Z_SortKeySpecList *sort_sequence, int *sort_status) +static Z_RPNQuery *copy_RPNQuery(Z_RPNQuery *src, NMEM nmem) +{ + Z_RPNQuery *dst = 0; + ODR encode = odr_createmem(ODR_ENCODE); + ODR decode = odr_createmem(ODR_DECODE); + + if (z_RPNQuery(encode, &src, 0, 0)) + { + int len; + char *buf = odr_getbuf(encode, &len, 0); + + if (buf) + { + odr_setbuf(decode, buf, len, 0); + z_RPNQuery(decode, &dst, 0, 0); + } + } + nmem_transfer(nmem, decode->mem); + odr_destroy(encode); + odr_destroy(decode); + return dst; +} + +static Z_SortKeySpecList *copy_SortKeySpecList(Z_SortKeySpecList *src, NMEM nmem) +{ + Z_SortKeySpecList *dst = 0; + ODR encode = odr_createmem(ODR_ENCODE); + ODR decode = odr_createmem(ODR_DECODE); + + if (z_SortKeySpecList(encode, &src, 0, 0)) + { + int len; + char *buf = odr_getbuf(encode, &len, 0); + + if (buf) + { + odr_setbuf(decode, buf, len, 0); + z_SortKeySpecList(decode, &dst, 0, 0); + } + } + nmem_transfer(nmem, decode->mem); + odr_destroy(encode); + odr_destroy(decode); + return dst; +} + +ZebraSet resultSetClone(ZebraHandle zh, const char *setname, + ZebraSet rset) +{ + ZebraSet nset; + int i; + + nset = resultSetAdd(zh, setname, 1); + if (!nset) + return 0; + + nset->nmem = nmem_create(); + + nset->num_bases = rset->num_bases; + nset->basenames = + nmem_malloc(nset->nmem, nset->num_bases * sizeof(*rset->basenames)); + for (i = 0; inum_bases; i++) + nset->basenames[i] = nmem_strdup(nset->nmem, rset->basenames[i]); + + if (rset->rset) + nset->rset = rset_dup(rset->rset); + if (rset->rpn) + nset->rpn = copy_RPNQuery(rset->rpn, nset->nmem); + return nset; +} + +ZEBRA_RES resultSetSort(ZebraHandle zh, NMEM nmem, + int num_input_setnames, const char **input_setnames, + const char *output_setname, + Z_SortKeySpecList *sort_sequence, int *sort_status) { ZebraSet sset; RSET rset; if (num_input_setnames == 0) { - zh->errCode = 208; - return ; + zebra_setError(zh, YAZ_BIB1_NO_RESULT_SET_NAME_SUPPLIED_ON_SORT, 0); + return ZEBRA_FAIL; } if (num_input_setnames > 1) { - zh->errCode = 230; - return; + zebra_setError(zh, YAZ_BIB1_SORT_TOO_MANY_INPUT_RESULTS, 0); + return ZEBRA_FAIL; } - yaz_log (LOG_DEBUG, "result set sort input=%s output=%s", - *input_setnames, output_setname); - sset = resultSetGet (zh, input_setnames[0]); + if (!log_level_set) + loglevels(); + yaz_log(log_level_sort, "result set sort input=%s output=%s", + *input_setnames, output_setname); + sset = resultSetGet(zh, input_setnames[0]); if (!sset) { - zh->errCode = 30; - zh->errString = nmem_strdup (nmem, input_setnames[0]); - return; + zebra_setError(zh, YAZ_BIB1_SPECIFIED_RESULT_SET_DOES_NOT_EXIST, + input_setnames[0]); + return ZEBRA_FAIL; } if (!(rset = sset->rset)) { - zh->errCode = 30; - zh->errString = nmem_strdup (nmem, input_setnames[0]); - return; + zebra_setError(zh, YAZ_BIB1_SPECIFIED_RESULT_SET_DOES_NOT_EXIST, + input_setnames[0]); + return ZEBRA_FAIL; } - if (strcmp (output_setname, input_setnames[0])) - { - rset = rset_dup (rset); - sset = resultSetAdd (zh, output_setname, 1); - sset->rset = rset; - } - resultSetSortSingle (zh, nmem, sset, rset, sort_sequence, sort_status); + if (strcmp(output_setname, input_setnames[0])) + sset = resultSetClone(zh, output_setname, sset); + sset->sortSpec = copy_SortKeySpecList(sort_sequence, sset->nmem); + return resultSetSortSingle(zh, nmem, sset, rset, sort_sequence, + sort_status); } -void resultSetSortSingle (ZebraHandle zh, NMEM nmem, - ZebraSet sset, RSET rset, - Z_SortKeySpecList *sort_sequence, int *sort_status) +ZEBRA_RES resultSetSortSingle(ZebraHandle zh, NMEM nmem, + ZebraSet sset, RSET rset, + Z_SortKeySpecList *sort_sequence, + int *sort_status) { int i; + int ib; + int n = 0; + zint kno = 0; zint psysno = 0; struct it_key key; - struct sortKeyInfo sort_criteria[3]; + struct sortKeyInfo sort_criteria[ZSET_SORT_MAX_LEVEL]; + char *cmp_buf[ZSET_SORT_MAX_LEVEL]; + char *tmp_cmp_buf[ZSET_SORT_MAX_LEVEL]; int num_criteria; RSFD rfd; TERMID termid; + TERMID *terms; + int numTerms = 0; + size_t sysno_mem_index = 0; + + int numbases = zh->num_basenames; + yaz_log(log_level_sort, "searching %d databases",numbases); + + if (zh->m_staticrank) + sysno_mem_index = 1; - yaz_log (LOG_LOG, "resultSetSortSingle start"); assert(nmem); /* compiler shut up about unused param */ sset->sort_info->num_entries = 0; + rset_getterms(rset, 0, 0, &n); + terms = (TERMID *) nmem_malloc(nmem, sizeof(*terms)*n); + rset_getterms(rset, terms, n, &numTerms); + sset->hits = 0; num_criteria = sort_sequence->num_specs; - if (num_criteria > 3) - num_criteria = 3; + if (num_criteria > ZSET_SORT_MAX_LEVEL) + num_criteria = ZSET_SORT_MAX_LEVEL; + /* set up the search criteria */ for (i = 0; i < num_criteria; i++) { Z_SortKeySpec *sks = sort_sequence->specs[i]; Z_SortKey *sk; + + sort_criteria[i].ord = (int *) + nmem_malloc(nmem, sizeof(int)*numbases); + sort_criteria[i].numerical = (int *) + nmem_malloc(nmem, sizeof(int)*numbases); + + /* initialize ord and numerical for each database */ + for (ib = 0; ib < numbases; ib++) + { + sort_criteria[i].ord[ib] = -1; + sort_criteria[i].numerical[ib] = 0; + } + if (sks->which == Z_SortKeySpec_missingValueData) + { + zebra_setError(zh, YAZ_BIB1_UNSUPP_MISSING_DATA_ACTION, 0); + return ZEBRA_FAIL; + } if (*sks->sortRelation == Z_SortKeySpec_ascending) sort_criteria[i].relation = 'A'; else if (*sks->sortRelation == Z_SortKeySpec_descending) sort_criteria[i].relation = 'D'; else { - zh->errCode = 214; - return; + zebra_setError(zh, YAZ_BIB1_ILLEGAL_SORT_RELATION, 0); + return ZEBRA_FAIL; } if (sks->sortElement->which == Z_SortElement_databaseSpecific) { - zh->errCode = 210; - return; + zebra_setError(zh, YAZ_BIB1_DATABASE_SPECIFIC_SORT_UNSUPP, 0); + return ZEBRA_FAIL; } else if (sks->sortElement->which != Z_SortElement_generic) { - zh->errCode = 237; - return; + zebra_setError(zh, YAZ_BIB1_SORT_ILLEGAL_SORT, 0); + return ZEBRA_FAIL; } sk = sks->sortElement->u.generic; switch (sk->which) { case Z_SortKey_sortField: - yaz_log (LOG_DEBUG, "Sort: key %d is of type sortField", i+1); - zh->errCode = 207; - return; - case Z_SortKey_elementSpec: - yaz_log (LOG_DEBUG, "Sort: key %d is of type elementSpec", i+1); - zh->errCode = 207; - return; - case Z_SortKey_sortAttributes: - yaz_log (LOG_DEBUG, "Sort: key %d is of type sortAttributes", i+1); - sort_criteria[i].attrUse = - zebra_maps_sort (zh->reg->zebra_maps, - sk->u.sortAttributes, - &sort_criteria[i].numerical); - yaz_log (LOG_DEBUG, "use value = %d", sort_criteria[i].attrUse); - if (sort_criteria[i].attrUse == -1) + yaz_log(log_level_sort, "key %d is of type sortField", i+1); + for (ib = 0; ib < numbases; ib++) { - zh->errCode = 116; - return; + zebraExplain_curDatabase(zh->reg->zei, zh->basenames[ib]); + sort_criteria[i].numerical[ib] = 0; + sort_criteria[i].ord[ib] = + zebraExplain_lookup_attr_str(zh->reg->zei, + zinfo_index_category_sort, + 0, sk->u.sortField); + if (sks->which != Z_SortKeySpec_null + && sort_criteria[i].ord[ib] == -1) + { + zebra_setError(zh, + YAZ_BIB1_CANNOT_SORT_ACCORDING_TO_SEQUENCE, 0); + return ZEBRA_FAIL; + } } - if (sortIdx_type (zh->reg->sortIdx, sort_criteria[i].attrUse)) + break; + case Z_SortKey_elementSpec: + yaz_log(log_level_sort, "key %d is of type elementSpec", i+1); + zebra_setError(zh, YAZ_BIB1_CANNOT_SORT_ACCORDING_TO_SEQUENCE, 0); + return ZEBRA_FAIL; + case Z_SortKey_sortAttributes: + yaz_log(log_level_sort, "key %d is of type sortAttributes", i+1); + /* for every database we searched, get the sort index file + id (ord) and its numerical indication and store them in + the sort_criteria */ + for (ib = 0; ib < numbases; ib++) { - zh->errCode = 207; - return; + zebraExplain_curDatabase(zh->reg->zei, zh->basenames[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; } break; } + /* right now we look up the index type based on the first database + if the index_type's can differ between the indexes of different + databases (which i guess they can?) then we have to store the + index types for each database, just like the ord and numerical */ + if (zebraExplain_lookup_ord(zh->reg->zei, sort_criteria[i].ord[0], + &sort_criteria[i].index_type, + 0, 0)) + { + zebra_setError(zh, YAZ_BIB1_CANNOT_SORT_ACCORDING_TO_SEQUENCE, 0); + return ZEBRA_FAIL; + } + } + /* allocate space for each cmpare buf + one extra for tmp comparison */ + /* cmp_buf is an array of array, the first dimension is the criteria and the second dimension are + all other result entries to compare against. This is slowly filled when records are processed. + tmp_cmp_buf is an array with a value of the current record for each criteria + */ + for (i = 0; isort_info->max_entries + * SORT_IDX_ENTRYSIZE); + tmp_cmp_buf[i] = xmalloc(SORT_IDX_ENTRYSIZE); } - rfd = rset_open (rset, RSETF_READ); - while (rset_read (rfd, &key,&termid)) - /* FIXME - pass a TERMID *, and use it for something below !! */ + rfd = rset_open(rset, RSETF_READ); + while (rset_read(rfd, &key, &termid)) { - zint this_sys = key.mem[0]; + zint this_sys = key.mem[sysno_mem_index]; + if (log_level_searchhits) + key_logdump_txt(log_level_searchhits, &key, termid->name); + 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)) + { + rset_set_hits_limit(rset, 0); + break; + } + } (sset->hits)++; psysno = this_sys; - resultSetInsertSort (zh, sset, - sort_criteria, num_criteria, psysno); - } - } - rset_close (rfd); - + + /* 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 - for (i = 0; i < rset->no_rset_terms; i++) - yaz_log (LOG_LOG, "term=\"%s\" nn=" ZINT_FORMAT - " type=%s count=" ZINT_FORMAT, - rset->rset_terms[i]->name, - rset->rset_terms[i]->nn, - rset->rset_terms[i]->flags, - rset->rset_terms[i]->count); + 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); + } + } + rset_close(rfd); + + /* free the compare buffers */ + for (i = 0; ihits); + for (i = 0; i < numTerms; i++) + yaz_log(log_level_sort, "term=\"%s\" type=%s count=" ZINT_FORMAT, + terms[i]->name, terms[i]->flags, terms[i]->rset->hits_count); *sort_status = Z_SortResponse_success; - yaz_log (LOG_LOG, "resultSetSortSingle end"); + return ZEBRA_OK; } -RSET resultSetRef (ZebraHandle zh, const char *resultSetId) +RSET resultSetRef(ZebraHandle zh, const char *resultSetId) { ZebraSet s; - if ((s = resultSetGet (zh, resultSetId))) + if ((s = resultSetGet(zh, resultSetId))) return s->rset; return NULL; } -void resultSetRank (ZebraHandle zh, ZebraSet zebraSet, RSET rset) +ZEBRA_RES resultSetRank(ZebraHandle zh, ZebraSet zebraSet, + RSET rset, NMEM nmem) { - zint kno = 0; struct it_key key; - RSFD rfd; TERMID termid; + TERMID *terms; + zint kno = 0; + int numTerms = 0; + int n = 0; int i; ZebraRankClass rank_class; - struct rank_control *rc; struct zset_sort_info *sort_info; const char *rank_handler_name = res_get_def(zh->res, "rank", "rank-1"); - double cur,tot; - zint est=-2; /* -2 not done, -1 can't do, >0 actual estimate*/ - zint esthits; - double ratio; + size_t sysno_mem_index = 0; + + if (zh->m_staticrank) + sysno_mem_index = 1; + if (!log_level_set) + loglevels(); sort_info = zebraSet->sort_info; sort_info->num_entries = 0; zebraSet->hits = 0; - rfd = rset_open (rset, RSETF_READ); + zebraSet->estimated_hit_count = 0; + rset_getterms(rset, 0, 0, &n); + terms = (TERMID *) nmem_malloc(nmem, sizeof(*terms)*n); + rset_getterms(rset, terms, n, &numTerms); - rank_class = zebraRankLookup (zh, rank_handler_name); + rank_class = zebraRankLookup(zh, rank_handler_name); if (!rank_class) { - yaz_log (LOG_WARN, "No such rank handler: %s", rank_handler_name); - return; + yaz_log(YLOG_WARN, "No such rank handler: %s", rank_handler_name); + zebra_setError(zh, YAZ_BIB1_UNSUPP_SEARCH, "Cannot find rank handler"); + return ZEBRA_FAIL; } - rc = rank_class->control; - - if (rset_read (rfd, &key, &termid)) + else { - zint psysno = key.mem[0]; - int score; - void *handle = - (*rc->begin) (zh->reg, rank_class->class_handle, rset); - (zebraSet->hits)++; - esthits=atoi(res_get_def(zh->res,"estimatehits","0")); - if (!esthits) - est=-1; /* can not do */ - do - { - zint this_sys = key.mem[0]; - kno++; - if (this_sys != psysno) - { - score = (*rc->calc) (handle, psysno); + RSFD rfd = rset_open(rset, RSETF_READ); + struct rank_control *rc = rank_class->control; + int score; + zint count = 0; + void *handle = (*rc->begin) (zh->reg, rank_class->class_handle, rset, + nmem, terms, numTerms); + zint psysno = 0; /* previous doc id / sys no */ + zint pstaticrank = 0; /* previous static rank */ + int stop_flag = 0; + while (rset_read(rfd, &key, &termid)) + { + zint this_sys = key.mem[sysno_mem_index]; - resultSetInsertRank (zh, sort_info, psysno, score, 'A'); - (zebraSet->hits)++; - psysno = this_sys; - } - /* FIXME - Ranking is broken, since rsets no longer have */ - /* term lists! */ - /* (*rc->add) (handle, this_sys, term_index); */ - - if ( (est==-2) && (zebraSet->hits==esthits)) - { /* time to estimate the hits */ - rset_pos(rfd,&cur,&tot); - if (tot>0) { - ratio=cur/tot; - est=(zint)(0.5+zebraSet->hits/ratio); - logf(LOG_LOG, "Estimating hits (%s) " - "%0.1f->"ZINT_FORMAT - "; %0.1f->"ZINT_FORMAT, - rset->control->desc, - cur, zebraSet->hits, - tot,est); - i=0; /* round to 3 significant digits */ - while (est>1000) { - est/=10; - i++; + zint seqno = key.mem[key.len-1]; + kno++; + if (log_level_searchhits) + key_logdump_txt(log_level_searchhits, &key, termid->name); + if (this_sys != psysno) + { /* new record .. */ + if (!(rfd->counted_items & 255) && zh->break_handler_func) + { + if (zh->break_handler_func(zh->break_handler_data)) + { + yaz_log(YLOG_LOG, "Aborted search"); + stop_flag = 1; } - while (i--) est*=10; - zebraSet->hits=est; } - } - } - while (rset_read (rfd, &key,&termid) && (est<0) ); - score = (*rc->calc) (handle, psysno); - resultSetInsertRank (zh, sort_info, psysno, score, 'A'); - (*rc->end) (zh->reg, handle); + 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, + &stop_flag); + /* insert the hit. A=Ascending */ + resultSetInsertRank(zh, sort_info, psysno, score, 'A'); + count++; + } + psysno = this_sys; + if (zh->m_staticrank) + pstaticrank = key.mem[0]; + } + (*rc->add)(handle, CAST_ZINT_TO_INT(seqno), termid); + } + /* no more items */ + if (psysno) + { /* we had - at least - one record */ + score = (*rc->calc)(handle, psysno, pstaticrank, &stop_flag); + /* insert the hit. A=Ascending */ + resultSetInsertRank(zh, sort_info, psysno, score, 'A'); + count++; + } + (*rc->end)(zh->reg, handle); + rset_close(rfd); } - rset_close (rfd); -/* - for (i = 0; i < rset->no_rset_terms; i++) - { - if (est>0) - rset->rset_terms[i]->count = - est=(zint)(rset->rset_terms[i]->count/ratio); - yaz_log (LOG_LOG, "term=\"%s\" nn=" ZINT_FORMAT - " type=%s count=" ZINT_FORMAT, - rset->rset_terms[i]->name, - rset->rset_terms[i]->nn, - rset->rset_terms[i]->flags, - rset->rset_terms[i]->count); + zebraSet->hits = rset->hits_count; + + yaz_log(log_level_searchterms, ZINT_FORMAT " keys, " + ZINT_FORMAT " sysnos, rank", kno, zebraSet->hits); + for (i = 0; i < numTerms; i++) + { + yaz_log(log_level_searchterms, "term=\"%s\" type=%s count=" + ZINT_FORMAT, + terms[i]->name, terms[i]->flags, terms[i]->rset->hits_count); } -*/ - yaz_log (LOG_DEBUG, ZINT_FORMAT " keys, "ZINT_FORMAT" distinct sysnos", - kno, zebraSet->hits); + return ZEBRA_OK; } -ZebraRankClass zebraRankLookup (ZebraHandle zh, const char *name) +ZebraRankClass zebraRankLookup(ZebraHandle zh, const char *name) { ZebraRankClass p = zh->reg->rank_classes; - while (p && strcmp (p->control->name, name)) + while (p && strcmp(p->control->name, name)) p = p->next; if (p && !p->init_flag) { @@ -842,18 +1201,18 @@ ZebraRankClass zebraRankLookup (ZebraHandle zh, const char *name) return p; } -void zebraRankInstall (struct zebra_register *reg, struct rank_control *ctrl) +void zebraRankInstall(struct zebra_register *reg, struct rank_control *ctrl) { - ZebraRankClass p = (ZebraRankClass) xmalloc (sizeof(*p)); - p->control = (struct rank_control *) xmalloc (sizeof(*p->control)); - memcpy (p->control, ctrl, sizeof(*p->control)); - p->control->name = xstrdup (ctrl->name); + ZebraRankClass p = (ZebraRankClass) xmalloc(sizeof(*p)); + p->control = (struct rank_control *) xmalloc(sizeof(*p->control)); + memcpy(p->control, ctrl, sizeof(*p->control)); + p->control->name = xstrdup(ctrl->name); p->init_flag = 0; p->next = reg->rank_classes; reg->rank_classes = p; } -void zebraRankDestroy (struct zebra_register *reg) +void zebraRankDestroy(struct zebra_register *reg) { ZebraRankClass p = reg->rank_classes; while (p) @@ -861,10 +1220,311 @@ void zebraRankDestroy (struct zebra_register *reg) ZebraRankClass p_next = p->next; if (p->init_flag && p->control->destroy) (*p->control->destroy)(reg, p->class_handle); - xfree (p->control->name); - xfree (p->control); - xfree (p); + xfree(p->control->name); + xfree(p->control); + xfree(p); p = p_next; } reg->rank_classes = NULL; } + +static int trav_rset_for_termids(RSET rset, TERMID *termid_array, + zint *hits_array, int *approx_array) +{ + int no = 0; + int i; + for (i = 0; ino_children; i++) + no += trav_rset_for_termids(rset->children[i], + (termid_array ? termid_array + no : 0), + (hits_array ? hits_array + no : 0), + (approx_array ? approx_array + no : 0)); + if (rset->term) + { + if (termid_array) + termid_array[no] = rset->term; + if (hits_array) + hits_array[no] = rset->hits_count; + if (approx_array) + approx_array[no] = rset->hits_approx; +#if 0 + yaz_log(YLOG_LOG, "rset=%p term=%s limit=" ZINT_FORMAT + " count=" ZINT_FORMAT, + rset, rset->term->name, rset->hits_limit, rset->hits_count); +#endif + no++; + } + return no; +} + +ZEBRA_RES zebra_result_set_term_no(ZebraHandle zh, const char *setname, + int *num_terms) +{ + ZebraSet sset = resultSetGet(zh, setname); + *num_terms = 0; + if (sset) + { + *num_terms = trav_rset_for_termids(sset->rset, 0, 0, 0); + return ZEBRA_OK; + } + return ZEBRA_FAIL; +} + +ZEBRA_RES zebra_result_set_term_info(ZebraHandle zh, const char *setname, + int no, zint *count, int *approx, + char *termbuf, size_t *termlen, + const char **term_ref_id) +{ + ZebraSet sset = resultSetGet(zh, setname); + if (sset) + { + int num_terms = trav_rset_for_termids(sset->rset, 0, 0, 0); + if (no >= 0 && no < num_terms) + { + TERMID *term_array = xmalloc(num_terms * sizeof(*term_array)); + zint *hits_array = xmalloc(num_terms * sizeof(*hits_array)); + int *approx_array = xmalloc(num_terms * sizeof(*approx_array)); + + trav_rset_for_termids(sset->rset, term_array, + hits_array, approx_array); + + if (count) + *count = hits_array[no]; + if (approx) + *approx = approx_array[no]; + if (termbuf) + { + char *inbuf = term_array[no]->name; + size_t inleft = strlen(inbuf); + size_t outleft = *termlen - 1; + + if (zh->iconv_from_utf8 != 0) + { + char *outbuf = termbuf; + size_t ret; + + ret = yaz_iconv(zh->iconv_from_utf8, &inbuf, &inleft, + &outbuf, &outleft); + if (ret == (size_t)(-1)) + *termlen = 0; + else + { + yaz_iconv(zh->iconv_from_utf8, 0, 0, + &outbuf, &outleft); + *termlen = outbuf - termbuf; + } + } + else + { + if (inleft > outleft) + inleft = outleft; + *termlen = inleft; + memcpy(termbuf, inbuf, *termlen); + } + termbuf[*termlen] = '\0'; + } + if (term_ref_id) + *term_ref_id = term_array[no]->ref_id; + + xfree(term_array); + xfree(hits_array); + xfree(approx_array); + return ZEBRA_OK; + } + } + return ZEBRA_FAIL; +} + +ZEBRA_RES zebra_snippets_hit_vector(ZebraHandle zh, const char *setname, + zint sysno, zebra_snippets *snippets) +{ + ZebraSet sset = resultSetGet(zh, setname); + yaz_log(YLOG_DEBUG, "zebra_get_hit_vector setname=%s zysno=" ZINT_FORMAT, + setname, sysno); + if (!sset) + return ZEBRA_FAIL; + else + { + struct rset_key_control *kc = zebra_key_control_create(zh); + NMEM nmem = nmem_create(); + struct it_key key; + RSET rsets[2], rset_comb; + RSET rset_temp = rset_create_temp(nmem, kc, kc->scope, + res_get(zh->res, "setTmpDir"),0 ); + + TERMID termid; + RSFD rsfd = rset_open(rset_temp, RSETF_WRITE); + + key.mem[0] = sysno; + key.mem[1] = 0; + key.mem[2] = 0; + key.mem[3] = 0; + key.len = 2; + rset_write(rsfd, &key); + rset_close(rsfd); + + rsets[0] = rset_temp; + rsets[1] = rset_dup(sset->rset); + + rset_comb = rset_create_and(nmem, kc, kc->scope, 2, rsets); + + rsfd = rset_open(rset_comb, RSETF_READ); + + while (rset_read(rsfd, &key, &termid)) + { + if (termid) + { + struct ord_list *ol; + for (ol = termid->ol; ol; ol = ol->next) + { + zebra_snippets_append(snippets, key.mem[key.len-1], 0, + ol->ord, termid->name); + } + } + } + rset_close(rsfd); + + rset_delete(rset_comb); + nmem_destroy(nmem); + kc->dec(kc); + } + return ZEBRA_OK; +} + +static ZEBRA_RES zebra_recid_to_sysno(ZebraHandle zh, + const char **basenames, int num_bases, + zint recid, + zint *sysnos, int *no_sysnos) +{ + ZEBRA_RES res = ZEBRA_OK; + int sysnos_offset = 0; + int i; + + if (!zh->reg->isamb || !zh->m_segment_indexing) + { + if (sysnos_offset < *no_sysnos) + *sysnos = recid; + sysnos_offset++; + } + else + { + for (i = 0; res == ZEBRA_OK && i < num_bases; i++) + { + const char *database = basenames[i]; + if (zebraExplain_curDatabase(zh->reg->zei, database) == 0) + { + const char *index_type = "w"; + const char *use_string = "_ALLRECORDS"; + int ord; + zinfo_index_category_t cat = zinfo_index_category_alwaysmatches; + ord = zebraExplain_lookup_attr_str(zh->reg->zei, cat, + index_type, use_string); + if (ord != -1) + { + char ord_buf[32]; + int ord_len = key_SU_encode(ord, ord_buf); + char *info; + + ord_buf[ord_len] = '\0'; + + info = dict_lookup(zh->reg->dict, ord_buf); + if (info) + { + if (*info != sizeof(ISAM_P)) + { + res = ZEBRA_FAIL; + } + else + { + ISAM_P isam_p; + ISAMB_PP pt; + struct it_key key_until, key_found; + int i = 0; + int r; + + memcpy(&isam_p, info+1, sizeof(ISAM_P)); + + pt = isamb_pp_open(zh->reg->isamb, isam_p, 2); + if (!pt) + res = ZEBRA_FAIL; + else + { + key_until.mem[i++] = recid; + key_until.mem[i++] = 0; /* section_id */ + if (zh->m_segment_indexing) + key_until.mem[i++] = 0; /* segment */ + key_until.mem[i++] = 0; + key_until.len = i; + + r = isamb_pp_forward(pt, &key_found, &key_until); + while (r && key_found.mem[0] == recid) + { + if (sysnos_offset < *no_sysnos) + sysnos[sysnos_offset++] = + key_found.mem[key_found.len-1]; + r = isamb_pp_read(pt, &key_found); + } + isamb_pp_close(pt); + } + } + } + } + } + } + } + *no_sysnos = sysnos_offset; + return res; +} + +ZEBRA_RES zebra_result_recid_to_sysno(ZebraHandle zh, + const char *setname, + zint recid, + zint *sysnos, int *no_sysnos) +{ + const char **basenames; + int num_bases; + ZEBRA_RES res; + + res = resultSetGetBaseNames(zh, setname, &basenames, &num_bases); + if (res != ZEBRA_OK) + return ZEBRA_FAIL; + + return zebra_recid_to_sysno(zh, basenames, num_bases, + recid, sysnos, no_sysnos); +} + +void zebra_count_set(ZebraHandle zh, RSET rset, zint *count, + zint approx_limit) +{ + zint psysno = 0; + struct it_key key; + RSFD rfd; + + yaz_log(YLOG_DEBUG, "count_set"); + + rset->hits_limit = approx_limit; + + *count = 0; + rfd = rset_open(rset, RSETF_READ); + while (rset_read(rfd, &key,0 /* never mind terms */)) + { + if (key.mem[0] != psysno) + { + psysno = key.mem[0]; + if (rfd->counted_items >= rset->hits_limit) + break; + } + } + rset_close(rfd); + *count = rset->hits_count; +} + + +/* + * Local variables: + * c-basic-offset: 4 + * c-file-style: "Stroustrup" + * indent-tabs-mode: nil + * End: + * vim: shiftwidth=4 tabstop=8 expandtab + */ +