Database - current read and snapshot read

The current reading ：insert update delete for update

Read the snapshot ： Non blocking read without lock ,select. The purpose is to improve concurrency

stay RC Under isolation level , The current read and snapshot read results are the same .

stay RR Under isolation level , The current read and snapshot read results may be different

for example

 Business 1  And business 2  Read a piece of data .
select * from user where id =1;
id     money
1       200

 Business 2  Modifying data 
update user set money = 1000 where id = 1;
commit;
 Business 1  read （ Read the snapshot ）：
select * from user where id  = 1;
id     money
1       200
 And in the execution （ The current reading ） 	select * from user where id  = 1 for update;
id     money
1       1000

RR Implementation principle of non blocking read at level ：

There are actually 3 The fields are ：
DB_TRX_ID : Transactions that modify this data ID
DB_ROLL_PRT: rollback pointer
DB_ROW_ID: Hidden lines ID

Business id=1, The data from 12 Change it to 32.

1. The bank adds an exclusive lock
2. Copy the data before modification , To undo log Inside
3. Modify the data of the current line , Fill in transactions id, rollback pointer
   
   Business id=2, Again on this trip Field3 Make changes .
4. Add an exclusive lock
5. Copy the data before modification , To undo
6. Modify the data of the current line , Fill in transactions id, rollback pointer

Every time you make a modification , Will generate one UNDO log data . When the transaction snapshot reads data, it will pass readView Algorithm , Use the current transaction id And log Of DB_TRX_ID Compare to determine which version of data is taken .