How to execute an SQL statement in MySQL

1.mysql Basic framework ：

• The connector ： Identity authentication and authority related ( Sign in MySQL When ).

• The query cache : When executing a query statement , The cache will be queried first （MySQL 8.0 Remove after version , Because this function is not very practical ）.

• analyzer : If you don't hit the cache ,SQL The statement goes through the parser , The analyzer says that we should first look at your SQL What is the statement , Check your SQL Is the sentence syntax correct .

• Optimizer ： according to MySQL Think the best plan to carry out .

• actuator : Execute statement , Then return the data from the storage engine .

Simply speaking MySQL It is mainly divided into Server Layer and storage engine layer ：

• Server layer ： Mainly including connectors 、 The query cache 、 analyzer 、 Optimizer 、 Actuators etc. , All cross-storage engine functionality is implemented in this layer , Like stored procedures 、 trigger 、 View , Functions, etc , There is also a general log module binglog Log module .

• Storage engine ： Mainly responsible for data storage and reading , Using a plug-in architecture that can be replaced , Support InnoDB、MyISAM、Memory Wait for multiple storage engines , among InnoDB The engine has its own logging module redolog modular . Now the most commonly used storage engine is InnoDB, It is from MySQL 5.5.5 The version has been used as the default storage engine from the beginning .

2. analysis of sentences ：

2.1 Query statement

sql It can be divided into two kinds , One is a query , One is updating （ increase , to update , Delete ）. Let's analyze the query statement first , The statement is as follows ：

select * from tb_student  A where A.age='18' and A.name='  Zhang San  ';

Combined with the above instructions , Let's analyze the execution flow of this statement ：

• First check whether the statement has permission , If you don't have permission , Return error message directly , If you have authority , stay MySQL8.0 Version before , The cache will be queried first , With this sql Statement for key Query in memory for results , If there is a direct cache , without , Perform the next step .

• Lexical analysis by Analyzer , extract sql Key elements of the statement , For example, extracting the above statement is a query select, Extract the table name to be queried as tb_student, Need to query all columns , The query criteria are id='1'. And then judge this sql Is there any grammatical error in the statement , For example, whether the keywords are correct, etc , If the check is OK, go to the next step .

• Next, the optimizer determines the execution scheme , above sql sentence , There can be two implementation options ：

  a. First look up the names in the student list “ Zhang San ” Of the students , Then determine if the age is  18.
  b. First find out the age of the students  18  Year old student , Then check the name as “ Zhang San ” Of the students .

  Then the optimizer selects the best execution efficiency scheme according to its own optimization algorithm （ Optimizer thinks , Sometimes it's not the best ）. After confirming the implementation plan, we are ready to start the implementation .

• Verify authority , If you don't have permission, an error message will be returned , If you have permission, you will call the database engine interface , Returns the execution result of the engine .

2.2 UPDATE statement

The above is a query sql The implementation process of , So let's see how an update statement is executed ？sql The statement is as follows ：

update tb_student A set A.age='19' where A.name='  Zhang San  ';

This statement will basically follow the flow of the previous query , Just keep a log when updating , This will introduce the log module ,MySQL Built in log module binlog（ Archive log ） , All storage engines can use , That we use a lot InnoDB The engine also comes with a log module redo log（ Redo log ）, We'll take InnoDB Mode to explore the execution process of this statement . The process is as follows ：

• First, query the data of Zhang San , If you have a cache , Cache is also used .

• Then get the query statement , hold age Change it to 19, Then invoke the engine. API Interface , Write this line of data ,InnoDB The engine keeps the data in memory , Simultaneous recording redo log, here redo log Get into prepare state , Then tell the actuator , Execution completed , Ready to submit .

• After the actuator receives the notice, it records binlog, Then call the engine interface , Submit redo log Is submitted status .

• Update complete .

Why use two log modules ：

This is because in the beginning MySQL Not with InnoDB engine ( InnoDB The engine is inserted by other companies in the form of plug-ins MySQL Of ) ,MySQL The engine is MyISAM, But we know that redo log yes InnoDB Engine specific , No other storage engine , This leads to the absence of crash-safe The ability of (crash-safe Even if the database is restarted abnormally , No records submitted before will be lost ),binlog Logs can only be used to archive .

This is not to say that only one log module is not allowed , It's just InnoDB The engine just goes through redo log To support affairs . that , Another student will ask , I use two log modules , But don't be so complicated , Why? redo log To introduce prepare Pre submission status ？ Here we use the counter evidence method to explain why we should do this ？

• First write redo log Submit directly , Then write binlog, Hypothesis complete redo log after , The machine hangs up ,binlog The log was not written , So after the machine restarts , This machine will go through redo log Restore data , But at this point bingog This data is not recorded , When the machine is backed up later , You lose this piece of data , At the same time, master-slave synchronization will also lose this piece of data .

• First write binlog, Then write redo log, Let's say it's done binlog, The machine rebooted unexpectedly , Because there is no redo log, This record cannot be recovered by this computer , however binlog There's another record , So the same thing as above , There will be data inconsistencies .

If the redo log The two-stage submission is not the same , finish writing sth. binglog after , Then submit redo log Will prevent the above problems , So as to ensure the consistency of data . So here comes the question , Is there an extreme situation ？ hypothesis redo log In pre commit state ,binglog It's finished , What happens when an abnormal restart occurs ？ This depends on MySQL The processing mechanism of ,MySQL The process is as follows ：

• Judge redo log Is it complete , If the judgment is complete , Submit immediately .

• If redo log It's just withholding but not commit state , This is the time to judge binlog Is it complete , If it's complete, submit redo log, Rollback transaction if incomplete .

This solves the problem of data consistency .