1、 background

In the project , We often use update sentence , that update Statement will lock the records in the table ？ Here we use some simple cases to simulate ._{Here is my own understanding , If that place is misunderstood , Welcome to point out}

2、 Pre knowledge

2.1 Isolation level of database

mysql> show variables like 'transaction_isolation';
+-----------------------+-----------------+
| Variable_name         | Value           |
+-----------------------+-----------------+
| transaction_isolation | REPEATABLE-READ |
+-----------------------+-----------------+
1 row in set (0.00 sec)

2.2 Database version

mysql> select version();
+-----------+
| version() |
+-----------+
| 8.0.28    |
+-----------+
1 row in set (0.00 sec)

2.3 Database storage engine

mysql> show variables like '%storage_engine%';
+---------------------------------+-----------+
| Variable_name                   | Value     |
+---------------------------------+-----------+
| default_storage_engine          | InnoDB    |
| default_tmp_storage_engine      | InnoDB    |
| disabled_storage_engines        |           |
| internal_tmp_mem_storage_engine | TempTable |
+---------------------------------+-----------+
4 rows in set (0.01 sec)

2.4 Is the lock on the record or on the index

The lock is added to the index , If there is no index in the table , Is it added to the table ？ It's not , It is also added to the index , There will be a default .

Record locks always lock index records, even if a table is defined with no indexes. For such cases, InnoDB creates a hidden clustered index and uses this index for record locking

Reference link ： dev.mysql.com/doc/refman/…

2.5 update...where The basic unit of locking is

UPDATE ... WHERE ... sets an exclusive next-key lock on every record the search encounters Here we can understand that the unit of locking is : next-key lock

2.6 Row-level locks

2.6.1 Record Locks

Record locks , That is, only one record will be locked . In fact, it locks the index of this record . A record lock is a lock on an index record. For example, SELECT c1 FROM t WHERE c1 = 10 FOR UPDATE; prevents any other transaction from inserting, updating, or deleting rows where the value of t.c1 is 10.

2.6.2 Gap Locks

Clearance lock , A gap lock is a lock on the gap between index records , That is, lock an interval . Front and back sections , Not including the record itself .

Clearance lock If using Single column unique index value To update , Yes. degeneration become Record Lock.

The purpose of the clearance lock ：

Prevent new data from being inserted into the gap
Prevent existing data from being updated into the gap .

Gap locking is not needed for statements that lock rows using a unique index to search > for a unique row. (This does not include the case that the search condition includes only > some columns of a multiple-column unique index; in that case, gap locking does occur.)

2.6.3 Next-Key Locks

Next-Key Lock yes On the index record Of Record locks and Before indexing records Of Clearance lock on clearance The combination of . It also locks a section , Front open back close interval . Including the record itself .

If the index value includes 1,5,10,30, that next key The lock may cover the following sections

(negative infinity, 1]
(1, 115
(5, 10]
(10, 30]
(30, positive infinity)

negative infinity It means negative infinity .positive infinity It means positive infinity .

2.6.4 Test the table structure of the lock table

create table test_record_lock
(
    id   int         not null comment ' Primary key ',
    age  int         null comment ' Age , General index ',
    name varchar(10) null comment ' full name , No index ',
    constraint test_record_lock_pk
        primary key (id)
)
    comment ' Test record lock ';

create index test_record_lock_age_index
    on test_record_lock (age);

2.6.5 Test data in table

mysql> select * from test_record_lock;
+----+------+--------+
| id | age  | name   |
+----+------+--------+
|  1 |   10 |  Zhang San    |
|  5 |   20 |  Li Si    |
|  8 |   25 |  Wang Wu    |
+----+------+--------+
3 rows in set (0.00 sec)

2.7 View the current lock in the database

select * from performance_schema.data_locks;

Field explanation ：

Field	value	explain
`lock_type`	TABLE	The lock is on the watch
	RECORD	Lock the record
`lock_mode`	IX	Intention exclusive lock
	X perhaps S	next-key lock Lock the record itself and the gap before the record
	X,REC_NOT_GAP	Record Lock Lock only the record itself
	S,REC_NOT_GAP	Record Lock Lock only the record itself
	X,GAP	gap lock
	X,INSERT_INTENTION	Insert intention lock
`lock_data`	A specific number	Represents the value of the primary key
	value , value	The first value is ： The value of a normal index Second value ： Primary key value

_{doubt ：X,GAP Can it be understood as X The lock degenerated into GAP lock .}

3、 Test data is locked

3.1 Unique index test

This applies to the unique index of a single field , Not suitable for a unique index of multiple fields

3.1.1 Equivalent update - Records exist

explain ：

Add next-key lock, Then the locked record range is (1,5].
Because it's the only index , And the query value exists ,next-key lock Degenerate into record lock, That is to say, in the end, only id=5 This line of data . The rest of the data does not affect .

3.1.2 Equivalent query - Record does not exist -01

explain ：

Add next-key lock, Then the locked record range is (5,8].
Because it's the only index , And the query value does not exist ,next-key lock Degenerate into gap, That is, the final locked data range is (5,8). The rest of the data does not affect .

3.1.3 Equivalent update - Record does not exist -02

3.1.4 Scope update

1、 Less than or equal to the maximum critical value

At this point, you can find that all the scanned records in the table have been added next key lock( Lock on index )

2、 Greater than or equal to the minimum critical value

mysql> begin;
Query OK, 0 rows affected (0.01 sec)

mysql> update test_record_lock set name = 'aaa' where id >= 1;
Query OK, 3 rows affected (0.00 sec)
Rows matched: 3  Changed: 3  Warnings: 0

mysql> select LOCK_TYPE,INDEX_NAME,LOCK_MODE,LOCK_DATA from performance_schema.data_locks;
+-----------+------------+---------------+------------------------+
| LOCK_TYPE | INDEX_NAME | LOCK_MODE     | LOCK_DATA              |
+-----------+------------+---------------+------------------------+
| TABLE     | NULL       | IX            | NULL                   |
| RECORD    | PRIMARY    | X,REC_NOT_GAP | 1                      |
| RECORD    | PRIMARY    | X             | supremum pseudo-record |
| RECORD    | PRIMARY    | X             | 8                      |
| RECORD    | PRIMARY    | X             | 5                      |
+-----------+------------+---------------+------------------------+
5 rows in set (0.01 sec)

Only records smaller than the minimum threshold can be inserted into the table .

3、 normal range

3.2 General index test

3.2.1 Equivalent update - Records exist

explain ：

First, the general index age add next-key lock, The scope of locking is (10,20]
next-key lock And lock this record , So in id The value of the index is equal to 5 Added Record Lock
Because it is a normal index and the value still exists , Therefore, a clearance lock will be added to the next section of this record Gap Lock, The locked range is (20,25)

3.2.2 Equivalent update - Record does not exist

explain ：

obtain next-key lock The locked range is (10,20]
Because the record to be updated does not exist ,next-key lock Degenerate into gap lock, So the locked range is (10,20)
Because it is a normal index and the record does not exist , So there is no need to find the next interval again .

3.2.3 Scope update

explain ：

Range update of normal index ,next-key-lock It does not degenerate into gap lock.

3.3 No index update

It can be seen from the above figure , Updating data tables without indexes is dangerous , It needs to be handled with care . No index update , Causes a full table scan , This causes all the scanned records to be added with next-key lock.