diff options
Diffstat (limited to 'storage/xtradb/fts/fts0que.cc')
| -rw-r--r-- | storage/xtradb/fts/fts0que.cc | 57 |
1 files changed, 37 insertions, 20 deletions
diff --git a/storage/xtradb/fts/fts0que.cc b/storage/xtradb/fts/fts0que.cc index c5c5f954789..beeb31abb9e 100644 --- a/storage/xtradb/fts/fts0que.cc +++ b/storage/xtradb/fts/fts0que.cc @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 2007, 2013, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 2007, 2014, Oracle and/or its affiliates. All Rights Reserved. This program 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 Software @@ -2800,20 +2800,19 @@ fts_query_get_token( ulint str_len; byte* new_ptr = NULL; - str_len = ut_strlen((char*) node->term.ptr); + str_len = node->term.ptr->len; ut_a(node->type == FTS_AST_TERM); token->f_len = str_len; - token->f_str = node->term.ptr; + token->f_str = node->term.ptr->str; if (node->term.wildcard) { token->f_str = static_cast<byte*>(ut_malloc(str_len + 2)); token->f_len = str_len + 1; - /* Need to copy the NUL character too. */ - memcpy(token->f_str, node->term.ptr, str_len + 1); + memcpy(token->f_str, node->term.ptr->str, str_len); token->f_str[str_len] = '%'; token->f_str[token->f_len] = 0; @@ -2848,8 +2847,8 @@ fts_query_visitor( switch (node->type) { case FTS_AST_TEXT: - token.f_str = node->text.ptr; - token.f_len = ut_strlen((char*) token.f_str); + token.f_str = node->text.ptr->str; + token.f_len = node->text.ptr->len; if (query->oper == FTS_EXIST) { ut_ad(query->intersection == NULL); @@ -2878,8 +2877,8 @@ fts_query_visitor( break; case FTS_AST_TERM: - token.f_str = node->term.ptr; - token.f_len = ut_strlen(reinterpret_cast<char*>(token.f_str)); + token.f_str = node->term.ptr->str; + token.f_len = node->term.ptr->len; /* Add the word to our RB tree that will be used to calculate this terms per document frequency. */ @@ -3191,13 +3190,9 @@ fts_query_read_node( to assign the frequency on search string behalf. */ if (query->cur_node->type == FTS_AST_TERM && query->cur_node->term.wildcard) { - - /* These cast are safe since we only care about the - terminating NUL character as an end of string marker. */ - term.f_len = ut_strlen(reinterpret_cast<char*> - (query->cur_node->term.ptr)); + term.f_len = query->cur_node->term.ptr->len; ut_ad(FTS_MAX_WORD_LEN >= term.f_len); - memcpy(term.f_str, query->cur_node->term.ptr, term.f_len); + memcpy(term.f_str, query->cur_node->term.ptr->str, term.f_len); } else { term.f_len = word->f_len; ut_ad(FTS_MAX_WORD_LEN >= word->f_len); @@ -3507,14 +3502,15 @@ fts_query_prepare_result( doc_freq = rbt_value(fts_doc_freq_t, node); /* Don't put deleted docs into result */ - if (fts_bsearch(array, 0, static_cast<int>(size), doc_freq->doc_id) - >= 0) { + if (fts_bsearch(array, 0, static_cast<int>(size), + doc_freq->doc_id) >= 0) { + /* one less matching doc count */ + --word_freq->doc_count; continue; } ranking.doc_id = doc_freq->doc_id; - ranking.rank = static_cast<fts_rank_t>( - doc_freq->freq * word_freq->idf * word_freq->idf); + ranking.rank = static_cast<fts_rank_t>(doc_freq->freq); ranking.words = NULL; fts_query_add_ranking(query, result->rankings_by_id, @@ -3527,6 +3523,25 @@ fts_query_prepare_result( } } + /* Calculate IDF only after we exclude the deleted items */ + fts_query_calculate_idf(query); + + node = rbt_first(query->word_freqs); + word_freq = rbt_value(fts_word_freq_t, node); + + /* Calculate the ranking for each doc */ + for (node = rbt_first(result->rankings_by_id); + node != NULL; + node = rbt_next(result->rankings_by_id, node)) { + + fts_ranking_t* ranking; + + ranking = rbt_value(fts_ranking_t, node); + + ranking->rank = static_cast<fts_rank_t>( + ranking->rank * word_freq->idf * word_freq->idf); + } + return(result); } @@ -3898,6 +3913,7 @@ fts_query( /* Get the deleted doc ids that are in the cache. */ fts_cache_append_deleted_doc_ids( index->table->fts->cache, query.deleted->doc_ids); + DEBUG_SYNC_C("fts_deleted_doc_ids_append"); /* Sort the vector so that we can do a binary search over the ids. */ ib_vector_sort(query.deleted->doc_ids, fts_update_doc_id_cmp); @@ -3954,7 +3970,8 @@ fts_query( } /* Calculate the inverse document frequency of the terms. */ - if (query.error == DB_SUCCESS) { + if (query.error == DB_SUCCESS + && query.flags != FTS_OPT_RANKING) { fts_query_calculate_idf(&query); } |