MySQL multi column index (composite index) features and usage scenarios

Create a table first , Have a surname ’first_name’、 name ’last_name’、 Father ID Etc ：

CREATE TABLE `users` (
  `id` int NOT NULL AUTO_INCREMENT COMMENT ' Primary key ID',
  `user_id` int NOT NULL COMMENT ' user ID',
  `first_name` varchar(32) DEFAULT NULL COMMENT ' surname ',
  `last_name` varchar(32) DEFAULT NULL COMMENT ' name ',
  `parent_id` int NOT NULL COMMENT 'parentID',
  `created_by` varchar(32) NOT NULL DEFAULT 'sys' COMMENT ' The creator ',
  `created_time` datetime NOT NULL DEFAULT CURRENT_TIMESTAMP COMMENT ' Creation time ',
  `updated_by` varchar(32) NOT NULL DEFAULT 'sys' COMMENT ' Reviser ',
  `updated_time` datetime NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT ' Modification time ',
  `is_deleted` varchar(1) NOT NULL DEFAULT '0' COMMENT ' Whether or not to delete ：0 Not delete 、1 deleted ',
  PRIMARY KEY (`id`),
  KEY `index_user_id` (`user_id`) USING BTREE,
  KEY `index_first_name` (`first_name`) USING BTREE,
  KEY `index_last_name` (`last_name`) USING BTREE
) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb3 COMMENT=' User table ';

Next we add some data , Put the code behind .

Single index

We know ,
where Only one index can be used ,mysql Will look for the index that it thinks is the best , The most restrictive index ,
But for multiple single column indexes ,mysql5.0 And above will be indexed and merged , And merge the results .

see sql Implementation plan of

EXPLAIN SELECT * FROM `users` WHERE first_name = 'Vm5' AND last_name = 'utD';

You can see type yes index_merge, This is index merging .

Index merging

Index merging can be divided into three types ：or union 、and The intersection 、and and or combination ,
Queries can use multiple single column indexes at the same time , And merge the results , It means extra consumption cpu、 Memory and other resources ,
Even so, it is cheaper than using only one index .

Index merging strategy is an optimization , If this happens, usually , It usually indicates that a multi column index containing related columns is needed .

Multi column index

Build an index on multiple fields according to the query requirements , To some extent, you can use the values of multiple columns to locate the specified row .

We added a that contains first_name、last_name Multi column index of index_first_last_name：

ALTER TABLE `users` ADD KEY `index_first_last_name` (`first_name`,`last_name`) USING BTREE;

Look again sql Implementation plan of

EXPLAIN SELECT * FROM `users` WHERE first_name = 'Vm5' AND last_name = 'utD';

Visible in WHERE first_name = ‘?’ AND last_name = '?' when ,
The multi column index limit is stronger than the previous single column index , Or more ,
mysql You can find a match immediately first_name then last_name.

But be careful ：

SELECT * FROM `users` WHERE first_name = 'Vm5' OR last_name = 'utD';

here ’or’ You can't use multiple column indexes .

Compared with multiple single column indexes , Will take up less disk space .

If you are interested, you can have a look at
adopt B+Tree Balanced multitree understanding InnoDB Clustered and nonclustered indexes of the engine

Order of multi column indexes

The order of multi column indexes is crucial ,
Experience tells us , Place the column with high selectivity on the left （ But not absolutely ）,
You need to consider the overall query requirements , Instead of a specific query .

image ：
KEY index_first_last_name (first_name,last_name)
Sometimes we just need to know the user's last name ,
Sometimes you may count the number of people with a certain surname , So here first_name To the left .

SELECT * FROM `users` WHERE last_name = 'utD';

Only last_name As a filter , There is also no need for multi column indexes .

Left most prefix

about ：
KEY a_b_c (a,b,c)
The leftmost column of a multi column index must appear in the filter criteria , And can't jump ,
Otherwise, the leftmost prefix will be invalid , Greatly reduce the query efficiency .

The scenarios where the leftmost prefix can be used are ：a、a&b、a&b&c,
a&c、b&c This kind is unusable .

Script for adding data

I'm through main Function to print sql, Then the output console To the file , Save as script .

	public static void main(String[] args) {
    
		System.out.println("INSERT INTO `users` (`user_id`, `first_name`, `last_name`, `parent_id`) VALUES");
		for (int i = 1; i < 130001; i++) {
    
			StringBuilder sqlSb = new StringBuilder();
			sqlSb.append("(");
			sqlSb.append(i + ", ");
			sqlSb.append("'" + RandomStringUtils.randomAlphanumeric(3) + "', ");
			sqlSb.append("'" + RandomStringUtils.randomAlphanumeric(3) + "', ");
			sqlSb.append(i/1000);
			sqlSb.append(")");
			if (i < 130000) {
    
				sqlSb.append(",");
			} else {
    
				sqlSb.append(";");
			}
			System.out.println(sqlSb.toString());
		}
	}

Right click class →run configurations→common

Reference article ：

Guo Er Ma ：MySQL The selection efficiency of single column index and combined index is the same as explain analysis
Yuwen 100： Correct understanding Mysql Column index and multi column index
duanx：mysql Index five ： Multi column index