MySQL back to table query optimization

1. What is return table query

Premise ： This test uses mysql 5.6 edition .

1. 1 mysql Storage engine for

mysql There are many categories of storage engines , The more common one is MyISAM and InnoDb Two kinds of , Specific storage The engine is unknown .

1. 2 InnoDb Storage engine

InnoDb The storage engine has two categories of indexes

Clustered index (clustered index)

General index (secondary index)

1. 3 The difference between a clustered index and a normal index

InnoDb The leaf node of the clustered index stores row records , therefore InnoDb There must be a clustered index and only one clustered index .

Clustered indexes are also conditional , It must be orderly , There is another point to be noted here , Order is not continuous , Such as ：1,3,5,7 Is ordered , But it's not continuous , Can still form a clustered index . If it's through uuid The generated primary key cannot form a clustered index .

InnoDb Leaf nodes store row records , Is a complete piece of data ,MyISAM Leaf nodes store record pointers .

InnoDB Clustered index The leaf node of the tree stores row records , therefore ,InnoDB There must be and only one clustered index

• A、 If you define PK, that PK Clustering index .
• B、 If there is no definition PK, that NOT NULL UNIQUE A column is a clustered index
• C、 otherwise InnoDb Will create a hidden row-id Columns as clustered indexes

Voice over ： therefore PK The inquiry is very fast , Locate row records directly .

InnoDB General index The leaf node stores the primary key value .

Voice over ： Be careful , Not store row record header pointer ,MyISAM The index leaf node stores the record pointer .

After a general understanding of the above concepts , So what is a watch ？ Suppose there is a table t(id PK, name KEY, sex, flag), Table has 4 Bar record . （id Is clustered index ,name It's a general index .）

CREATE TABLE `user` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`name` varchar(255) CHARACTER SET utf8 COLLATE utf8_general_ci NOT NULL,
`sex` varchar(5),
`flag` varchar(5),
PRIMARY KEY (`id`) USING BTREE)
ENGINE = InnoDB AUTO_INCREMENT = 1 CHARACTE
SET = utf8 COLLATE = utf8_general_ci ROW_FORMAT = Compact;

2. The difference between two different indexes storing the above data

Clustered index storage （ Store row records ） id by PK, Clustered index , Leaf nodes store row records

Nonclustered index （ Storage PK） name by KEY, General index , Leaf node storage PK value , namely id

Suppose we execute the following query statement ,mysql How is it implemented ？

select * from user where name='ls';

summary ： The pink path is the process of query execution , First through the general index name Navigate to the primary key value id=5, Then locate the row record through the cluster index . This is called back table query , First locate the primary key value , Repositioning row records , Its performance is lower than scanning the index tree .

You need to scan the index tree twice ：

• First locate the primary key value through the normal index id=5;
• To locate row records through a clustered index ;

3、 So we need to build a composite index to cover the queried fields ;

In this way, you can get... From an index tree SQL All required column data , There is no need to return the form , Faster .

No coverage index is used ( Note the single valued index )

create table user (
id int primary key,
name varchar(20),
sex varchar(5),
index(name)
)engine=innodb;

Query statement

explain select id, name, sex from `user` where name = "ls"

explain The results of the analysis

analysis :

keys The value of this column is name, Description hit index name, however sex The value of this column can only be queried by going back to the table

Use overlay index ( Note that the composite index )

create table user (
id int primary key,
name varchar(20),
sex varchar(5),
index(name, sex)
)engine=innodb;

Query statement

explain select id, name, sex from `user` where name = "ls"

explain The results of the analysis

analysis :

Can hit index coverage , There is no need to return the form .

4、 Which scenarios can be optimized with index coverage SQL？ How to avoid querying data in the back table ？

scene 1： Full table count Query optimization

The original table is ：

user(PK id, name, sex);

direct ：

select count(name) from user;

Cannot use index overlay .

Add index ：

alter table user add key(name);

We can use index coverage to improve efficiency .

scene 2： Column query return table optimization

select id,name,sex … where name=‘shenjian’;

This example will not be repeated , Index a single column (name) Upgrade to Federated index (name, sex), To avoid returning to the table .

scene 3： Paging query

select id,name,sex … order by name limit 500,100;

Index a single column (name) Upgrade to Federated index (name, sex), You can also avoid going back to the table .