Range access on multiple columns using an index - mysql

Question

As per my understanding of the MySQL documentation which states:
The optimizer attempts to use additional key parts to determine the interval as long as the comparison operator is =, <=>, or IS NULL. If the operator is >, <, >=, <=, !=, <>, BETWEEN, or LIKE, the optimizer uses it but considers no more key parts.
https://dev.mysql.com/doc/refman/5.6/en/range-optimization.html#range-access-multi-part
I was trying to optimize the query the following query:
SELECT col3 FROM tbl2 WHERE col3 = 'fHI' AND col1 > 20 AND col4 > 0.5;
where tbl2 is defined as:
CREATE TABLE `tbl2` (
`col1` int NOT NULL,
`col2` varchar(512) NOT NULL DEFAULT '',
`col3` varchar(512) DEFAULT NULL,
`col4` double DEFAULT NULL,
PRIMARY KEY (`col2`),
KEY `cd1` (`col3`,`col1`,`col4`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1
I was under the impression that the columns used would be col3 and col1 since the predicate on col1 uses the > operator. However, the optimizer plan says that it uses all three index key columns as per the plan.
{
"query_block": {
"select_id": 1,
"cost_info": {
"query_cost": "92.10"
},
"table": {
"table_name": "tbl2",
"access_type": "index",
"possible_keys": [
"cd1"
],
"key": "cd1",
"used_key_parts": [
"col3",
"col1",
"col4"
],
"key_length": "528",
"rows_examined_per_scan": 901,
"rows_produced_per_join": 881,
"filtered": "97.78",
"using_index": true,
"cost_info": {
"read_cost": "4.00",
"eval_cost": "88.10",
"prefix_cost": "92.10",
"data_read_per_join": "901K"
},
"used_columns": [
"col1",
"col3",
"col4"
],
"attached_condition": "((`test`.`tbl2`.`col3` = 'fHI') and (`test`.`tbl2`.`col1` > 20) and (`test`.`tbl2`.`col4` > 0.5))"
}
}
}
Is the documentation wrong or is col4 being used in a different way?

Answer 1

mysql> explain SELECT col3 FROM tbl2 WHERE col3 = 'fHI' AND col1 > 20 AND col4 > 0.5;
+----+-------------+-------+------------+-------+---------------+------+---------+------+------+----------+--------------------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+-------+------------+-------+---------------+------+---------+------+------+----------+--------------------------+
| 1 | SIMPLE | tbl2 | NULL | index | cd1 | cd1 | 528 | NULL | 1 | 100.00 | Using where; Using index |
+----+-------------+-------+------------+-------+---------------+------+---------+------+------+----------+--------------------------+
I think the answer is that EXPLAIN is sloppy in multiple ways.
I cannot explain where "528" comes from.
Anyway, note that the Optimizer Trace mentions only
"ranges": [
"fHI <= col3 <= fHI AND 20 < col1"
],
Optimizer Trace:
{
"steps": [
{
"join_preparation": {
"select#": 1,
"steps": [
{
"expanded_query": "/* select#1 */ select `tbl2`.`col3` AS `col3` from `tbl2` where ((`tbl2`.`col3` = 'fHI') and (`tbl2`.`col1` > 20) and (`tbl2`.`col4` > 0.5)) limit 33"
}
]
}
},
{
"join_optimization": {
"select#": 1,
"steps": [
{
"condition_processing": {
"condition": "WHERE",
"original_condition": "((`tbl2`.`col3` = 'fHI') and (`tbl2`.`col1` > 20) and (`tbl2`.`col4` > 0.5))",
"steps": [
{
"transformation": "equality_propagation",
"resulting_condition": "((`tbl2`.`col1` > 20) and (`tbl2`.`col4` > 0.5) and multiple equal('fHI', `tbl2`.`col3`))"
},
{
"transformation": "constant_propagation",
"resulting_condition": "((`tbl2`.`col1` > 20) and (`tbl2`.`col4` > 0.5) and multiple equal('fHI', `tbl2`.`col3`))"
},
{
"transformation": "trivial_condition_removal",
"resulting_condition": "((`tbl2`.`col1` > 20) and (`tbl2`.`col4` > 0.5) and multiple equal('fHI', `tbl2`.`col3`))"
}
]
}
},
{
"substitute_generated_columns": {
}
},
{
"table_dependencies": [
{
"table": "`tbl2`",
"row_may_be_null": false,
"map_bit": 0,
"depends_on_map_bits": [
]
}
]
},
{
"ref_optimizer_key_uses": [
{
"table": "`tbl2`",
"field": "col3",
"equals": "'fHI'",
"null_rejecting": true
}
]
},
{
"rows_estimation": [
{
"table": "`tbl2`",
"range_analysis": {
"table_scan": {
"rows": 1,
"cost": 2.45
},
"potential_range_indexes": [
{
"index": "PRIMARY",
"usable": false,
"cause": "not_applicable"
},
{
"index": "cd1",
"usable": true,
"key_parts": [
"col3",
"col1",
"col4",
"col2"
]
}
],
"best_covering_index_scan": {
"index": "cd1",
"cost": 0.35,
"chosen": true
},
"setup_range_conditions": [
],
"group_index_range": {
"chosen": false,
"cause": "not_group_by_or_distinct"
},
"skip_scan_range": {
"potential_skip_scan_indexes": [
{
"index": "cd1",
"usable": false,
"cause": "prefix_not_const_equality"
}
]
},
"analyzing_range_alternatives": {
"range_scan_alternatives": [
{
"index": "cd1",
"ranges": [
"fHI <= col3 <= fHI AND 20 < col1"
],
"index_dives_for_eq_ranges": true,
"rowid_ordered": false,
"using_mrr": false,
"index_only": true,
"rows": 1,
"cost": 0.36,
"chosen": false,
"cause": "cost"
}
],
"analyzing_roworder_intersect": {
"usable": false,
"cause": "too_few_roworder_scans"
}
}
}
}
]
},
{
"considered_execution_plans": [
{
"plan_prefix": [
],
"table": "`tbl2`",
"best_access_path": {
"considered_access_paths": [
{
"access_type": "ref",
"index": "cd1",
"chosen": false,
"cause": "range_uses_more_keyparts"
},
{
"rows_to_scan": 1,
"access_type": "scan",
"resulting_rows": 1,
"cost": 0.35,
"chosen": true
}
]
},
"condition_filtering_pct": 100,
"rows_for_plan": 1,
"cost_for_plan": 0.35,
"chosen": true
}
]
},
{
"attaching_conditions_to_tables": {
"original_condition": "((`tbl2`.`col3` = 'fHI') and (`tbl2`.`col1` > 20) and (`tbl2`.`col4` > 0.5))",
"attached_conditions_computation": [
],
"attached_conditions_summary": [
{
"table": "`tbl2`",
"attached": "((`tbl2`.`col3` = 'fHI') and (`tbl2`.`col1` > 20) and (`tbl2`.`col4` > 0.5))"
}
]
}
},
{
"finalizing_table_conditions": [
{
"table": "`tbl2`",
"original_table_condition": "((`tbl2`.`col3` = 'fHI') and (`tbl2`.`col1` > 20) and (`tbl2`.`col4` > 0.5))",
"final_table_condition ": "((`tbl2`.`col3` = 'fHI') and (`tbl2`.`col1` > 20) and (`tbl2`.`col4` > 0.5))"
}
]
},
{
"refine_plan": [
{
"table": "`tbl2`"
}
]
}
]
}
},
{
"join_execution": {
"select#": 1,
"steps": [
]
}
}
]
}