/* Copyright (c) 2000, 2024, Oracle and/or its affiliates. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License, version 2.0, as published by the Free Software Foundation. This program is designed to work with certain software (including but not limited to OpenSSL) that is licensed under separate terms, as designated in a particular file or component or in included license documentation. The authors of MySQL hereby grant you an additional permission to link the program and your derivative works with the separately licensed software that they have either included with the program or referenced in the documentation. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License, version 2.0, for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA */ /** @file Optimising of MIN(), MAX() and COUNT(*) queries without 'group by' clause by replacing the aggregate expression with a constant. Given a table with a compound key on columns (a,b,c), the following types of queries are optimised (assuming the table handler supports the required methods) @verbatim SELECT COUNT(*) FROM t1[,t2,t3,...] SELECT MIN(b) FROM t1 WHERE a=const SELECT MAX(c) FROM t1 WHERE a=const AND b=const SELECT MAX(b) FROM t1 WHERE a=const AND bconst SELECT MIN(b) FROM t1 WHERE a=const AND b BETWEEN const AND const SELECT MAX(b) FROM t1 WHERE a=const AND b BETWEEN const AND const @endverbatim Instead of '<' one can use '<=', '>', '>=' and '=' as well. Instead of 'a=const' the condition 'a IS NULL' can be used. If all selected fields are replaced then we will also remove all involved tables and return the answer without any join. Thus, the following query will be replaced with a row of two constants: @verbatim SELECT MAX(b), MIN(d) FROM t1,t2 WHERE a=const AND bconst @endverbatim (assuming a index for column d of table t2 is defined) */ #include #include #include #include "ft_global.h" #include "my_base.h" #include "my_bitmap.h" #include "my_dbug.h" #include "my_inttypes.h" #include "my_sys.h" #include "my_table_map.h" #include "mysql_com.h" #include "sql/field.h" #include "sql/handler.h" #include "sql/item.h" #include "sql/item_cmpfunc.h" #include "sql/item_func.h" #include "sql/item_sum.h" // Item_sum #include "sql/key.h" // key_cmp_if_same #include "sql/range_optimizer/range_optimizer.h" #include "sql/sql_bitmap.h" #include "sql/sql_class.h" #include "sql/sql_const.h" #include "sql/sql_error.h" #include "sql/sql_lex.h" #include "sql/sql_list.h" #include "sql/sql_opt_exec_shared.h" #include "sql/sql_select.h" #include "sql/table.h" static bool find_key_for_maxmin(bool max_fl, Index_lookup *ref, Item_field *item_field, Item *cond, uint *range_fl, uint *key_prefix_length); static bool reckey_in_range(bool max_fl, Index_lookup *ref, Item_field *item_field, Item *cond, uint range_fl, uint prefix_len); static bool maxmin_in_range(bool max_fl, Item_field *item_field, Item *cond); /** Get exact count of rows in all tables. This is called, when at least one of the table handlers support HA_COUNT_ROWS_INSTANT, but not HA_STATS_RECORDS_IS_EXACT (NDB is one such storage engine). @param tables List of tables @retval Product of number of rows in all tables. ULLONG_MAX for error. */ static ulonglong get_exact_record_count(Table_ref *tables) { ulonglong count = 1; for (Table_ref *tl = tables; tl; tl = tl->next_leaf) { ha_rows tmp = 0; const int error = tl->table->file->ha_records(&tmp); if (error != 0) return ULLONG_MAX; count *= tmp; } return count; } /** Use index to read MIN(field) value. @param table Table object @param ref Reference to the structure where we store the key value @param item_field Field used in MIN() @param range_fl Whether range endpoint is strict less than @param prefix_len Length of common key part for the range @retval 0 No errors HA_ERR_... Otherwise */ static int get_index_min_value(TABLE *table, Index_lookup *ref, Item_field *item_field, uint range_fl, uint prefix_len) { int error; if (!ref->key_length) error = table->file->ha_index_first(table->record[0]); else { /* Use index to replace MIN/MAX functions with their values according to the following rules: 1) Insert the minimum non-null values where the WHERE clause still matches, or 2) a NULL value if there are only NULL values for key_part_k. 3) Fail, producing a row of nulls Implementation: Read the smallest value using the search key. If the interval is open, read the next value after the search key. If read fails, and we're looking for a MIN() value for a nullable column, test if there is an exact match for the key. */ if (!(range_fl & NEAR_MIN)) /* Closed interval: Either The MIN argument is non-nullable, or we have a >= predicate for the MIN argument. */ error = table->file->ha_index_read_map( table->record[0], ref->key_buff, make_prev_keypart_map(ref->key_parts), HA_READ_KEY_OR_NEXT); else { /* Open interval: There are two cases: 1) We have only MIN() and the argument column is nullable, or 2) there is a > predicate on it, nullability is irrelevant. We need to scan the next bigger record first. Open interval is not used if the search key involves the last keypart, and it would not work. */ assert(prefix_len < ref->key_length); error = table->file->ha_index_read_map( table->record[0], ref->key_buff, make_prev_keypart_map(ref->key_parts), HA_READ_AFTER_KEY); /* If the found record is outside the group formed by the search prefix, or there is no such record at all, check if all records in that group have NULL in the MIN argument column. If that is the case return that NULL. Check if case 1 from above holds. If it does, we should read the skipped tuple. */ if (item_field->field->is_nullable() && ref->key_buff[prefix_len] == 1 && /* Last keypart (i.e. the argument to MIN) is set to NULL by find_key_for_maxmin only if all other keyparts are bound to constants in a conjunction of equalities. Hence, we can detect this by checking only if the last keypart is NULL. */ (error == HA_ERR_KEY_NOT_FOUND || key_cmp_if_same(table, ref->key_buff, ref->key, prefix_len))) { assert(item_field->field->is_nullable()); error = table->file->ha_index_read_map( table->record[0], ref->key_buff, make_prev_keypart_map(ref->key_parts), HA_READ_KEY_EXACT); } } } return error; } /** Use index to read MAX(field) value. @param table Table object @param ref Reference to the structure where we store the key value @param range_fl Whether range endpoint is strict greater than @retval 0 No errors HA_ERR_... Otherwise */ static int get_index_max_value(TABLE *table, Index_lookup *ref, uint range_fl) { return (ref->key_length ? table->file->ha_index_read_map( table->record[0], ref->key_buff, make_prev_keypart_map(ref->key_parts), range_fl & NEAR_MAX ? HA_READ_BEFORE_KEY : HA_READ_PREFIX_LAST_OR_PREV) : table->file->ha_index_last(table->record[0])); } /** Substitute constants for some COUNT(), MIN() and MAX() functions in an aggregated (implicitly grouped) query @param[in] thd thread handler @param[in] select query block @param[in] fields All fields to be returned @param[in] conds WHERE clause @param[out] decision outcome for successful execution = AGGR_REGULAR regular execution required = AGGR_COMPLETE values available = AGGR_DELAYED execution with ha_records() required = AGGR_EMPTY source tables empty, aggregates are NULL or zero (for COUNT) @returns false if success, true if error This function is called for queries with aggregate functions and no GROUP BY, thus the result set will contain a single row only. First, the function will analyze the source tables and WHERE clause to see whether the query qualifies for optimization. If not, the decision AGGR_REGULAR is returned. Second, the function walks over all expressions in the SELECT list. If the expression can be optimized with a storage engine operation that is O(1) (MIN or MAX) or O(0) (instant COUNT), the value is looked up and inserted in the value buffer, and the corresponding Item is marked as being const. If the expression is a COUNT operation that can be evaluated efficiently by the storage manager (but still O(N)), indicated with HA_COUNT_ROWS_INSTANT, it will be marked as such. When all SELECT list expressions have been processed, there are four possible outcomes: - An empty result from the source tables is indicated, and the output state is AGGR_EMPTY. Notice that the result of aggregation is still one row, filled with zero for COUNT operations and NULL values for all other expressions. - All expressions have been given values, indicated with output state AGGR_COMPLETE. - All expressions have been given values, except for one or more COUNT operations that will be evaluated in execution. This is indicated with AGGR_DELAYED. - Some expressions must be evaluated as part of a regular execution, indicated with AGGR_REGULAR. Notice that some of the expressions may have been given values and are marked as const, but no expressions will be candidates for delayed execution. */ bool optimize_aggregated_query(THD *thd, Query_block *select, const mem_root_deque &fields, Item *conds, aggregate_evaluated *decision) { DBUG_TRACE; // True means at least one aggregate must be calculated by regular execution bool aggr_impossible = false; // True means COUNT expressions will be calculated by ha_records() // storage engine operations during execution phase bool aggr_delayed = false; bool recalc_const_item = false; // Calculated row count, valid only if have_exact_count is true. ulonglong row_count = 1; // True when all tables have an exact count bool have_exact_count = true; // True if all tables have contents and can be read bool tables_filled = true; // The set of tables optimized for MIN or MAX table_map removed_tables = 0; // The set of inner tables of outer join(s) table_map inner_tables = 0; // The set of tables in the join, excluding the inner tables of outer join table_map used_tables = 0; Table_ref *tables = select->leaf_tables; int error; *decision = AGGR_REGULAR; // Default return value // Local flag that indicates that ha_records() can be called in the // execution phase only. bool delay_ha_records_to_exec_phase = false; const table_map where_tables = conds ? conds->used_tables() : 0; /* A subquery is optimized once but executed possibly multiple times. If the value of the set function depends on the join's emptiness (like MIN() does), and the join's emptiness depends on the outer row or something nondeterministic, we cannot mark the set function as constant: */ if (where_tables & (OUTER_REF_TABLE_BIT | RAND_TABLE_BIT)) return false; if (!select->sj_nests.empty()) // Cannot optimize when there is a semijoin or antijoin return false; /* Analyze outer join dependencies, and, if possible, compute the number of returned rows. */ for (Table_ref *tl = tables; tl; tl = tl->next_leaf) { // Don't replace expression on a table that is part of an outer join if (tl->is_inner_table_of_outer_join()) { inner_tables |= tl->map(); /* We can't optimise LEFT JOIN in cases where the WHERE condition restricts the table that is used, like in: SELECT MAX(t1.a) FROM t1 LEFT JOIN t2 join-condition WHERE t2.field IS NULL; */ if (tl->map() & where_tables) return false; } else used_tables |= tl->map(); /* If the storage manager of 'tl' gives exact row count as part of statistics (cheap), compute the total number of rows. If there are no outer table dependencies, this count may be used as the real count. Schema tables are filled after this function is invoked, so we can't get row count. Derived tables aren't filled yet, their number of rows are estimates. FORCE INDEX implies that user wants a specific index to be used. So skip using stats.records. */ tables_filled &= !(tl->schema_table || tl->uses_materialization()); const ulonglong table_flags = tl->table->file->ha_table_flags(); if ((table_flags & HA_STATS_RECORDS_IS_EXACT) && tables_filled && !tl->table->force_index) { error = tl->fetch_number_of_rows(); if (error) { tl->table->file->print_error(error, MYF(ME_FATALERROR)); return true; } row_count *= tl->table->file->stats.records; } else { /* Note: If at least one of the tables can't be optimized, then all tables will be read in the execution phase (i.e. end_send_count). Example: SELECT COUNT(*) FROM t_myisam, t_innodb; */ delay_ha_records_to_exec_phase |= !(table_flags & HA_COUNT_ROWS_INSTANT) || tl->table->force_index; have_exact_count = false; } } /* Iterate through all items in the SELECT clause and replace COUNT(), MIN() and MAX() with constants (if possible). */ for (Item *item : fields) { if (item->type() == Item::SUM_FUNC_ITEM && !item->m_is_window_function) { if (item->is_outer_reference()) { aggr_impossible = true; continue; } Item_sum *item_sum = down_cast(item); enum Item_func::Functype func_type = conds != nullptr && conds->type() == Item::FUNC_ITEM ? down_cast(conds)->functype() : Item_func::UNKNOWN_FUNC; switch (item_sum->sum_func()) { case Item_sum::COUNT_FUNC: { Item_sum_count *item_count = down_cast(item_sum); /* If the expr in COUNT(expr) can never be null we can change this to the number of rows in the tables if this number is exact and there are no outer joins. */ if (conds == nullptr && !item_count->get_arg(0)->is_nullable() && !inner_tables && tables_filled) { if (delay_ha_records_to_exec_phase) { aggr_delayed = true; } else { if (!have_exact_count) { row_count = get_exact_record_count(tables); if (row_count == ULLONG_MAX) { /* Error from handler in counting rows. Don't optimize count() */ aggr_impossible = true; continue; } have_exact_count = true; // count is now exact } } } /* For result count of full-text search: If 1. it is a single table query, 2. the WHERE condition is a single MATCH expression, 3. the table engine can provide the row count from FTS result, and 4. the expr in COUNT(expr) can not be NULL, we do the full-text search now, and replace with the actual count. Note: Item_func_match::init_search() will be called again later in the optimization phase by init_fts_funcs(), but search will still only be done once. */ else if (tables->next_leaf == nullptr && // 1 (func_type == Item_func::FT_FUNC || func_type == Item_func::MATCH_FUNC) && // 2 (tables->table->file->ha_table_flags() & HA_CAN_FULLTEXT_EXT) && // 3 !item_count->get_arg(0)->is_nullable()) // 4 { Item_func_match *fts_item = func_type == Item_func::FT_FUNC ? down_cast(conds) : down_cast( down_cast(conds) ->arguments()[0]); fts_item->get_master()->set_hints(nullptr, FT_NO_RANKING, HA_POS_ERROR, false); if (fts_item->init_search(thd)) break; row_count = fts_item->get_count(); have_exact_count = true; } else aggr_impossible = true; // See comment above for get_exact_record_count() if (!thd->lex->is_explain() && !aggr_impossible && !aggr_delayed) { item_count->make_const((longlong)row_count); recalc_const_item = true; } break; } case Item_sum::MIN_FUNC: case Item_sum::MAX_FUNC: { int is_max = (item_sum->sum_func() == Item_sum::MAX_FUNC); /* If MIN/MAX(expr) is the first part of a key or if all previous parts of the key is found in the COND, then we can use indexes to find the key. */ Item *expr = item_sum->get_arg(0)->real_item(); if (expr->type() == Item::FIELD_ITEM) { uchar key_buff[MAX_KEY_LENGTH]; Index_lookup ref; uint range_fl, prefix_len; ref.key_buff = key_buff; Item_field *item_field = down_cast(expr); Table_ref *tr = item_field->table_ref; TABLE *table = tr->table; /* If table already has been read, we may use the values already in record buffer. However, the logic around this is a bit unclear, thus we skip this possible optimization for now. */ if (table->const_table) { aggr_impossible = true; break; } /* We must not have accessed this table instance yet, because it must be private to this query block, as we already ensured that OUTER_REF_TABLE_BIT is not set. Or: if this field references an outer table in the form of an in-memory table with one row, it will have been read and closed again. In such a case, the OUTER_REF_TABLE_BIT is no longer set. */ assert(!table->file->inited); /* Look for a partial key that can be used for optimization. If we succeed, ref.key_length will contain the length of this key, while prefix_len will contain the length of the beginning of this key without field used in MIN/MAX(). Type of range for the key part for this field will be returned in range_fl. */ if ((inner_tables & tr->map()) || !find_key_for_maxmin(is_max, &ref, item_field, conds, &range_fl, &prefix_len)) { aggr_impossible = true; break; } if ((error = table->file->ha_index_init((uint)ref.key, true))) { table->file->print_error(error, MYF(0)); table->set_keyread(false); return true; } /* Necessary columns to read from the index have been determined by find_key_for_maxmin(); they are the columns involved in 'WHERE col=const' and the aggregated one. We may not need all columns of read_set, neither all columns of the index. */ assert(table->read_set == &table->def_read_set); assert(bitmap_is_clear_all(&table->tmp_set)); table->read_set = &table->tmp_set; table->mark_columns_used_by_index_no_reset(ref.key, table->read_set, ref.key_parts); // The aggregated column may or not be included in ref.key_parts. bitmap_set_bit(table->read_set, item_field->field->field_index()); error = is_max ? get_index_max_value(table, &ref, range_fl) : get_index_min_value(table, &ref, item_field, range_fl, prefix_len); /* Set table row status to "not started" unconditionally. This will prepare the table for regular access in the join execution machinery if this optimization is aborted and cannot be used. The row status does not affect column values read into record[0]. */ table->set_not_started(); table->read_set = &table->def_read_set; bitmap_clear_all(&table->tmp_set); /* Verify that the read tuple indeed matches the search key */ if (!error && reckey_in_range(is_max, &ref, item_field, conds, range_fl, prefix_len)) error = HA_ERR_KEY_NOT_FOUND; table->set_keyread(false); table->file->ha_index_end(); if (error) { if (error == HA_ERR_KEY_NOT_FOUND || error == HA_ERR_END_OF_FILE) { *decision = AGGR_EMPTY; return false; // No rows matching WHERE } /* HA_ERR_LOCK_DEADLOCK or some other error */ table->file->print_error(error, MYF(0)); return true; } removed_tables |= tr->map(); } else if (!expr->const_for_execution() || conds != nullptr || !have_exact_count) { /* We get here if the aggregate function is not based on a field. Example: "SELECT MAX(1) FROM table ..." This constant optimization is not applicable if 1. the expression is not constant, or 2. it is unknown if the query returns any rows. MIN/MAX must return NULL if the query doesn't return any rows. We can't determine this if: - the query has a condition, because, in contrast to the "MAX(field)" case above, the condition will not be evaluated against an index for this case, or - the storage engine does not provide exact count, which means that it doesn't know whether there are any rows. */ aggr_impossible = true; break; } item_sum->set_aggregator(item_sum->has_with_distinct() ? Aggregator::DISTINCT_AGGREGATOR : Aggregator::SIMPLE_AGGREGATOR); /* If row_count == 0 and there are no outer joins, set to NULL, otherwise set to the constant value. */ if (have_exact_count && row_count == 0 && !inner_tables) { item_sum->aggregator_clear(); // Mark the aggregated value as based on no rows item->no_rows_in_result(); } else item_sum->reset_and_add(); item_sum->make_const(); recalc_const_item = true; break; } default: aggr_impossible = true; break; } } else if (!aggr_impossible) { if (recalc_const_item) item->update_used_tables(); if (!item->const_for_execution()) aggr_impossible = true; } } if (thd->is_error()) return true; /* With a where clause, only ignore searching in the tables if MIN/MAX optimisation replaced all used tables. We do not use replaced values in case of: SELECT MIN(key) FROM table_1, empty_table removed_tables is != 0 if we have used MIN() or MAX(). */ if (removed_tables && used_tables != removed_tables) aggr_impossible = true; // We didn't remove all tables *decision = aggr_impossible ? AGGR_REGULAR : aggr_delayed ? AGGR_DELAYED : AGGR_COMPLETE; return false; } /** Test if the predicate compares a field with constants. @param func_item Predicate item @param[out] args Here we store the field followed by constants @param[out] inv_order Is set to true if the predicate is of the form 'const op field' @returns true if function is a simple predicate, false otherwise. */ bool is_simple_predicate(Item_func *func_item, Item **args, bool *inv_order) { Item *item; *inv_order = false; switch (func_item->argument_count()) { case 0: /* MULT_EQUAL_FUNC */ { Item_equal *item_equal = down_cast(func_item); args[0] = item_equal->get_first(); if (item_equal->members() > 1) return false; if (!(args[1] = item_equal->const_arg())) return false; } break; case 1: /* field IS NULL */ item = func_item->arguments()[0]; if (item->type() != Item::FIELD_ITEM) return false; args[0] = item; break; case 2: /* 'field op const' or 'const op field' */ item = func_item->arguments()[0]; if (item->type() == Item::FIELD_ITEM) { args[0] = item; item = func_item->arguments()[1]; if (!item->const_for_execution()) return false; args[1] = item; } else if (item->const_for_execution()) { args[1] = item; item = func_item->arguments()[1]; if (item->type() != Item::FIELD_ITEM) return false; args[0] = item; *inv_order = true; } else return false; break; case 3: /* field BETWEEN const AND const */ item = func_item->arguments()[0]; if (item->type() != Item::FIELD_ITEM) return false; args[0] = item; for (int i = 1; i <= 2; i++) { item = func_item->arguments()[i]; if (!item->const_for_execution()) return false; args[i] = item; } } return true; } /** Check if a predicate should make the endpoint for the MIN/MAX optimization move towards -Inf for MAX or towards +Inf for MIN. It is assumed that the current endpoint is stored in the field accessed by "cond", and that "cond" represents a comparison of a field and constant values using the <, >, <=, >= or BETWEEN operator. */ static bool move_endpoint(bool is_max, Item *cond) { const Item_func::Functype type = down_cast(cond)->functype(); assert(is_function_of_type(cond, Item_func::LT_FUNC) || is_function_of_type(cond, Item_func::LE_FUNC) || is_function_of_type(cond, Item_func::GT_FUNC) || is_function_of_type(cond, Item_func::GE_FUNC) || is_function_of_type(cond, Item_func::BETWEEN)); if (type != Item_func::BETWEEN) { // For <, <=, >, >= operators, just check if "cond" evaluates to true on the // current endpoint. If it does not, it represents a stricter condition than // those seen before, so the endpoint should be moved. return cond->val_int() == 0; } // For BETWEEN, we check similarly, but only using the upper bound constant // for MAX and the lower bound constant for MIN. Item_func_between *const between = down_cast(cond); Item_func_inequality *compare_endpoints = nullptr; if (is_max) { compare_endpoints = new Item_func_le(between->get_arg(0), between->get_arg(2)); } else { compare_endpoints = new Item_func_ge(between->get_arg(0), between->get_arg(1)); } if (compare_endpoints == nullptr) return false; if (compare_endpoints->set_cmp_func()) return true; compare_endpoints->update_used_tables(); compare_endpoints->quick_fix_field(); return compare_endpoints->val_int() == 0; } /** Check whether a condition matches a key to get {MAX|MIN}(field):. For the index specified by the keyinfo parameter and an index that contains the field as its component (field_part), the function checks whether - the condition cond is a conjunction, - all of its conjuncts refer to columns of the same table, and - each conjunct is on one of the following forms: - f_i = const_i or const_i = f_i or f_i IS NULL, where f_i is part of the index - field {<|<=|>=|>|=} const - const {<|<=|>=|>|=} field - field BETWEEN const_1 AND const_2 As a side-effect, the key value to be used for looking up the MIN/MAX value is actually stored inside the Field object. An interesting feature is that the function will find the most restrictive endpoint by over-eager evaluation of the @c WHERE condition. It continually stores the current endpoint inside the Field object. For a query such as @code SELECT MIN(a) FROM t1 WHERE a > 3 AND a > 5; @endcode the algorithm will recurse over the conjunction, storing first a 3 in the field. In the next recursive invocation the expression a > 5 is evaluated as 3 > 5 (Due to the dual nature of Field objects as value carriers and field identifiers), which will obviously fail, leading to 5 being stored in the Field object. @param[in] max_fl Set to true if we are optimizing MAX(), false means we are optimizing %MIN() @param[in, out] ref Reference to the structure where the function stores the key value @param[in] keyinfo Reference to the key info @param[in] field_part Pointer to the key part for the field @param[in] cond WHERE condition @param[in] map Table map for the key @param[in,out] key_part_used Map of matchings parts. The function will output the set of key parts actually being matched in this set, yet it relies on the caller to initialize the value to zero. This is due to the fact that this value is passed recursively. @param[in,out] range_fl Says whether endpoints use strict greater/less than. @param[out] prefix_len Length of common key part for the range where MAX/MIN is searched for @retval false Index can't be used. @retval true We can use the index to get MIN/MAX value */ static bool matching_cond(bool max_fl, Index_lookup *ref, KEY *keyinfo, KEY_PART_INFO *field_part, Item *cond, table_map map, key_part_map *key_part_used, uint *range_fl, uint *prefix_len) { DBUG_TRACE; if (!cond) return true; if (!(cond->used_tables() & map)) { /* Condition doesn't restrict the used table */ return true; } if (cond->type() == Item::COND_ITEM) { if (((Item_cond *)cond)->functype() == Item_func::COND_OR_FUNC) return false; /* AND */ List_iterator_fast li(*((Item_cond *)cond)->argument_list()); Item *item; while ((item = li++)) { if (!matching_cond(max_fl, ref, keyinfo, field_part, item, map, key_part_used, range_fl, prefix_len)) return false; } return true; } if (cond->type() != Item::FUNC_ITEM) return false; // Not operator, can't optimize bool eq_type = false; // =, <=> or IS NULL bool is_null_safe_eq = false; // The operator is NULL safe, e.g. <=> bool noeq_type = false; // < or > bool less_fl = false; // < or <= bool is_null = false; // IS NULL bool between = false; // BETWEEN ... AND ... const Item_func::Functype functype = down_cast(cond)->functype(); switch (functype) { case Item_func::ISNULL_FUNC: is_null = true; [[fallthrough]]; case Item_func::EQ_FUNC: eq_type = true; break; case Item_func::EQUAL_FUNC: eq_type = is_null_safe_eq = true; break; case Item_func::LT_FUNC: noeq_type = true; [[fallthrough]]; case Item_func::LE_FUNC: less_fl = true; break; case Item_func::GT_FUNC: noeq_type = true; [[fallthrough]]; case Item_func::GE_FUNC: break; case Item_func::BETWEEN: between = true; // NOT BETWEEN is equivalent to OR and is therefore not a conjunction if (((Item_func_between *)cond)->negated) return false; break; case Item_func::MULT_EQUAL_FUNC: eq_type = true; break; default: return false; // Can't optimize function } Item *args[] = {nullptr, nullptr, nullptr}; bool inv; /* Test if this is a comparison of a field and constant */ if (!is_simple_predicate(down_cast(cond), args, &inv)) return false; // Test if the field and the constant values can be compared in an index. Field *const field = down_cast(args[0])->field; for (Item *value : {args[1], args[2]}) { if (value != nullptr && !comparable_in_index(cond, field, Field::itRAW, functype, value)) { return false; } } if (!is_null_safe_eq && !is_null && (args[1]->is_null() || (between && args[2]->is_null()))) return false; if (inv && !eq_type) less_fl = !less_fl; // Convert '<' -> '>' (etc) /* Check if field is part of the tested partial key */ uchar *key_ptr = ref->key_buff; KEY_PART_INFO *part; for (part = keyinfo->key_part;; key_ptr += part++->store_length) { if (part > field_part) return false; // Field is beyond the tested parts if (part->field->eq(field)) { break; // Found a part of the key for the field } } const bool is_field_part = part == field_part; if (!(is_field_part || eq_type)) return false; const key_part_map org_key_part_used = *key_part_used; if (eq_type || between || max_fl == less_fl) { const size_t length = (key_ptr - ref->key_buff) + part->store_length; if (ref->key_length < length) { /* Ultimately ref->key_length will contain the length of the search key */ ref->key_length = length; ref->key_parts = (part - keyinfo->key_part) + 1; } if (!*prefix_len && part + 1 == field_part) *prefix_len = length; if (is_field_part && eq_type) *prefix_len = ref->key_length; *key_part_used |= (key_part_map)1 << (part - keyinfo->key_part); } if (org_key_part_used == *key_part_used && /* The current search key is not being extended with a new key part. This means that the a condition is added a key part for which there was a previous condition. We can only overwrite such key parts in some special cases, e.g. a > 2 AND a > 1 (here range_fl must be set to something). In all other cases the WHERE condition is always false anyway. */ (eq_type || *range_fl == 0)) return false; if (org_key_part_used != *key_part_used || (is_field_part && (between || eq_type || max_fl == less_fl) && move_endpoint(max_fl, cond))) { /* It's the first predicate for this part or a predicate of the following form that moves upper/lower bounds for max/min values: - field BETWEEN const AND const - field = const - field {<|<=} const, when searching for MAX - field {>|>=} const, when searching for MIN */ if (is_null || (is_null_safe_eq && args[1]->is_null())) { /* If we have a non-nullable index, we cannot use it, since set_null will be ignored, and we will compare uninitialized data. */ if (!part->field->is_nullable()) return false; part->field->set_null(); *key_ptr = (uchar)1; } else { /* Update endpoints for MAX/MIN, see function comment. */ Item *value = args[between && max_fl ? 2 : 1]; /* A perfect save is necessary. Truncated / incorrect value can result in an incorrect index lookup. Truncation of trailing space is ignored for strings with a PAD SPACE collation. (When trailing space has been removed, TYPE_NOTE_TRUNCATED is returned. Not to be confused with TYPE_WARN_TRUNCATED, which is returned when non-space has been removed.) TODO(khatlen): It might be better if string literals that need truncation are handled during constant folding, so that we only need to check for TYPE_OK here. analyze_field_constant() does not yet handle string constants. */ const type_conversion_status retval = value->save_in_field_no_warnings(part->field, true); if (!(retval == TYPE_OK || (retval == TYPE_NOTE_TRUNCATED && part->field->is_text_key_type() && part->field->charset()->pad_attribute == PAD_SPACE))) { return false; } if (part->null_bit) *key_ptr++ = (uchar)(part->field->is_null()); part->field->get_key_image(key_ptr, part->length, Field::itRAW); } if (is_field_part) { if (between || eq_type) *range_fl &= ~(NO_MAX_RANGE | NO_MIN_RANGE | NEAR_MAX | NEAR_MIN); else { *range_fl &= ~(max_fl ? NO_MAX_RANGE : NO_MIN_RANGE); if (noeq_type) *range_fl |= (max_fl ? NEAR_MAX : NEAR_MIN); else *range_fl &= ~(max_fl ? NEAR_MAX : NEAR_MIN); } } } else if (eq_type) { if ((!is_null && !cond->val_int()) || (is_null && !part->field->is_null())) return false; // Impossible test } else if (is_field_part) *range_fl &= ~(max_fl ? NO_MIN_RANGE : NO_MAX_RANGE); return true; } /** Check whether we can get value for {max|min}(field) by using a key. If where-condition is not a conjunction of 0 or more conjuct the function returns false, otherwise it checks whether there is an index including field as its k-th component/part such that: -# for each previous component f_i there is one and only one conjunct of the form: f_i= const_i or const_i= f_i or f_i is null -# references to field occur only in conjucts of the form: field {<|<=|>=|>|=} const or const {<|<=|>=|>|=} field or field BETWEEN const1 AND const2 -# all references to the columns from the same table as column field occur only in conjucts mentioned above. -# each of k first components the index is not partial, i.e. is not defined on a fixed length proper prefix of the field. If such an index exists the function through the ref parameter returns the key value to find max/min for the field using the index, the length of first (k-1) components of the key and flags saying how to apply the key for the search max/min value. (if we have a condition field = const, prefix_len contains the length of the whole search key) @param[in] max_fl 0 for MIN(field) / 1 for MAX(field) @param[in,out] ref Reference to the structure we store the key value @param[in] item_field Field used inside MIN() / MAX() @param[in] cond WHERE condition @param[out] range_fl Bit flags for how to search if key is ok @param[out] prefix_len Length of prefix for the search range @note This function may set field->table->key_read to true, which must be reset after index is used! (This can only happen when function returns 1) @returns true if index can be used to optimize MIN()/MAX(), false otherwise. If true, ref, range_fl and prefix_len are updated */ static bool find_key_for_maxmin(bool max_fl, Index_lookup *ref, Item_field *item_field, Item *cond, uint *range_fl, uint *prefix_len) { Field *const field = item_field->field; if (!field->is_flag_set(PART_KEY_FLAG)) return false; // Not key field DBUG_TRACE; TABLE *const table = field->table; uint idx = 0; KEY *keyinfo, *keyinfo_end; for (keyinfo = table->key_info, keyinfo_end = keyinfo + table->s->keys; keyinfo != keyinfo_end; keyinfo++, idx++) { KEY_PART_INFO *part, *part_end; key_part_map key_part_to_use = 0; /* Perform a check if index is not disabled by ALTER TABLE or IGNORE INDEX. */ if (!table->keys_in_use_for_query.is_set(idx)) continue; uint jdx = 0; *prefix_len = 0; for (part = keyinfo->key_part, part_end = part + actual_key_parts(keyinfo); part != part_end; part++, jdx++, key_part_to_use = (key_part_to_use << 1) | 1) { if (!(table->file->index_flags(idx, jdx, false) & HA_READ_ORDER)) return false; // Due to lack of time, currently only ASC keyparts are supported. if (part->key_part_flag & HA_REVERSE_SORT) break; /* Check whether the index component is partial */ Field *part_field = table->field[part->fieldnr - 1]; if (part_field->is_flag_set(BLOB_FLAG) || part->length < part_field->key_length()) break; if (field->eq(part->field)) { ref->key = idx; ref->key_length = 0; ref->key_parts = 0; key_part_map key_part_used = 0; *range_fl = NO_MIN_RANGE | NO_MAX_RANGE; if (matching_cond(max_fl, ref, keyinfo, part, cond, item_field->table_ref->map(), &key_part_used, range_fl, prefix_len) && !(key_part_to_use & ~key_part_used)) { if (!max_fl && key_part_used == key_part_to_use && part->null_bit) { /* The query is on this form: SELECT MIN(key_part_k) FROM t1 WHERE key_part_1 = const and ... and key_part_k-1 = const If key_part_k is nullable, we want to find the first matching row where key_part_k is not null. The key buffer is now {const, ..., NULL}. This will be passed to the handler along with a flag indicating open interval. If a tuple is read that does not match these search criteria, an attempt will be made to read an exact match for the key buffer. */ /* Set the first byte of key_part_k to 1, that means NULL */ ref->key_buff[ref->key_length] = 1; ref->key_length += part->store_length; ref->key_parts++; assert(ref->key_parts == jdx + 1); *range_fl &= ~NO_MIN_RANGE; *range_fl |= NEAR_MIN; // Open interval } /* The following test is false when the key in the key tree is converted (for example to upper case) */ if (field->part_of_key.is_set(idx)) table->set_keyread(true); return true; } } } } return false; } /** Check whether found key is in range specified by conditions. @param[in] max_fl false for MIN(field) / true for MAX(field) @param[in] ref Reference to the key value and info @param[in] item_field Item representing field used in MIN/MAX expression @param[in] cond Condition to check nullptr means that all keys are within the range. @param[in] range_fl Says whether there is a condition to to be checked @param[in] prefix_len Length of the constant part of the key @returns true if the condition is not true for the found row, false otherwise. */ static bool reckey_in_range(bool max_fl, Index_lookup *ref, Item_field *item_field, Item *cond, uint range_fl, uint prefix_len) { if (key_cmp_if_same(item_field->field->table, ref->key_buff, ref->key, prefix_len)) return true; if (cond == nullptr) return false; // Constant conditions which were not evaluated earlier (in optimize_cond()) // should be evaluated now. if (cond->const_for_execution()) return (cond->val_int() == 0); if (range_fl & (max_fl ? NO_MIN_RANGE : NO_MAX_RANGE)) return false; return maxmin_in_range(max_fl, item_field, cond); } /** Check whether {MAX|MIN}(field) is in range specified by conditions. @param[in] max_fl false for MIN(field) / true for MAX(field) @param[in] item_field Item representing field used in MIN/MAX expression @param[in] cond Condition to check @returns true if condition is not true for the found row, otherwise false. */ static bool maxmin_in_range(bool max_fl, Item_field *item_field, Item *cond) { /* If AND/OR condition */ if (cond->type() == Item::COND_ITEM) { List_iterator_fast li(*down_cast(cond)->argument_list()); Item *item; while ((item = li++)) { if (maxmin_in_range(max_fl, item_field, item)) return true; } return false; } // Check that condition actually references the specific table if ((cond->used_tables() & item_field->table_ref->map()) == 0) return false; bool less_fl = false; switch (down_cast(cond)->functype()) { case Item_func::BETWEEN: return cond->val_int() == 0; case Item_func::LT_FUNC: case Item_func::LE_FUNC: less_fl = true; [[fallthrough]]; case Item_func::GT_FUNC: case Item_func::GE_FUNC: { Item *item = down_cast(cond)->arguments()[1]; /* In case of 'const op item' we have to swap the operator */ if (!item->const_for_execution()) less_fl = !less_fl; /* We only have to check the expression if we are using an expression like SELECT MAX(b) FROM t1 WHERE a=const AND b>const not for SELECT MAX(b) FROM t1 WHERE a=const AND bval_int() == 0; // Return true if condition is false return false; } case Item_func::EQ_FUNC: case Item_func::EQUAL_FUNC: case Item_func::MULT_EQUAL_FUNC: case Item_func::ISNULL_FUNC: break; default: // Keep compiler happy assert(false); // Impossible break; } return false; }