X-Git-Url: http://git.indexdata.com/?p=idzebra-moved-to-github.git;a=blobdiff_plain;f=index%2Fzsets.c;h=a1687304db170f5f44215eaa81372f267a2de2bc;hp=e2d3da037dada1a00ff81aef319b60996e91cd32;hb=96e4c5479e111511f5df3531b6648931251b9e5d;hpb=125946893fceb92f36727a7c91c1674a2668571f diff --git a/index/zsets.c b/index/zsets.c index e2d3da0..a168730 100644 --- a/index/zsets.c +++ b/index/zsets.c @@ -1,5 +1,5 @@ /* 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 @@ -18,6 +18,9 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA */ +#if HAVE_CONFIG_H +#include +#endif #include #include #ifdef WIN32 @@ -113,6 +116,7 @@ static ZEBRA_RES resultSetSearch(ZebraHandle zh, NMEM nmem, NMEM rset_nmem, 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, @@ -125,7 +129,8 @@ static ZEBRA_RES resultSetSearch(ZebraHandle zh, NMEM nmem, NMEM rset_nmem, 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); @@ -554,15 +559,15 @@ struct sortKeyInfo { }; 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; ireg->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) { @@ -625,17 +623,11 @@ void resultSetInsertSort(ZebraHandle zh, ZebraSet sset, 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); @@ -880,7 +872,6 @@ ZEBRA_RES resultSetSortSingle(ZebraHandle zh, NMEM nmem, { int i; int ib; - int cached_success_db = 0; int n = 0; zint kno = 0; zint psysno = 0; @@ -917,7 +908,6 @@ ZEBRA_RES resultSetSortSingle(ZebraHandle zh, NMEM nmem, { Z_SortKeySpec *sks = sort_sequence->specs[i]; Z_SortKey *sk; - ZEBRA_RES res; sort_criteria[i].ord = (int *) nmem_malloc(nmem, sizeof(int)*numbases); @@ -989,13 +979,12 @@ ZEBRA_RES resultSetSortSingle(ZebraHandle zh, NMEM nmem, 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 @@ -1030,6 +1019,7 @@ ZEBRA_RES resultSetSortSingle(ZebraHandle zh, NMEM nmem, 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)) @@ -1040,9 +1030,28 @@ ZEBRA_RES resultSetSortSingle(ZebraHandle zh, NMEM nmem, } (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); @@ -1138,6 +1147,11 @@ ZEBRA_RES resultSetRank(ZebraHandle zh, ZebraSet zebraSet, } 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, @@ -1146,12 +1160,6 @@ ZEBRA_RES resultSetRank(ZebraHandle zh, ZebraSet zebraSet, 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]; @@ -1517,6 +1525,7 @@ void zebra_count_set(ZebraHandle zh, RSET rset, zint *count, /* * Local variables: * c-basic-offset: 4 + * c-file-style: "Stroustrup" * indent-tabs-mode: nil * End: * vim: shiftwidth=4 tabstop=8 expandtab