MySQL index transaction

MySQL Index transactions

1. Indexes

Index is a special kind of file , Contains a reference pointer to all records in the data table . You can index one or more columns in a table ,
And specify the type of index , Each index has its own data structure implementation .

1.1 effect

1. In the database surface , data , Indexes The relationship between , Similar to that on the bookshelf The book , The content of the book , Book catalogue The relationship between .
2. The function of index is similar to that of book catalogue , Can quickly locate , Retrieving data .

1.2 The underlying structure ( Basics )

The essence of indexing is to use some more complex data structures , Through this data structure, the records to be queried are organized , Thus, the speed of query can be accelerated .

In the data structure with high search efficiency, there are binary trees ( Red and black trees ,B+ Trees ), Hashtable . Which one is that ?

For the red and black trees , The basic structure is binary tree , If there are more elements , The height of the tree will increase , If the deeper the search data , This will lead to an increase in the number of comparisons in the query , Because the database needs to access the disk every time , But the reading time of the hard disk is much longer than the data processing time , The fewer times the database reads the hard disk, the better . More comparisons , Will slow down the speed .

For hash tables , His search time complexity 0(1), But we can only check the same situation , For range finding (< , > ) Or fuzzy search , There's nothing we can do .

The final answer is B+ Trees , Let's analyze B+ The structural characteristics and advantages of trees .

B+ The tree is made up of B The tree extends out , and B The tree is a N Fork tree , Its design idea is , Try to gather relevant data together , To read multiple data at once , Reduce the number of hard disk operations .

B Tree features :

1. A node can hold multiple values , For example, above , The most one node holds 4 It's worth
2. Unless the data is full , Otherwise, no new layer will be added . in other words ,B The tree pursues " layer " The less, the better. .
3. Values in child nodes , And the value in the parent node , There is a strict size correspondence . Generally speaking , If the parent node has a It's worth , Then there is a+1 Child node . Like in the picture above , The parent node has two values （7 and 16）, It corresponds to three child nodes , The first child node is less than 7 Value , The last child node is greater than 16 Value , The child node in the middle is 7 and 16 Between the value of the .

The time complexity of its search is log(N)N.

• B+ Tree structure

stay B Add several conditions on the basis of the tree

1. Several values are stored on each node , At most, there are several subtrees .
2. The elements of the parent node exist in the child node , Is the maximum of the child nodes ( minimum value ).
3. The leaf nodes at the bottom are connected by a linked list

advantage :

1. Fast query speed .
2. A single node stores more data , The height of the tree is lower , All comparisons will be greatly reduced .
3. All leaf nodes are connected by linked list , It is very convenient to find the range ( Because the leaf node contains the global data of the database ).
4. Every data row , Just save it on the leaf node , Non leaf nodes do not have to store real data rows , Just store the indexed id that will do , At this time, the leaf node takes up less space , It is possible to cache non leaf nodes in memory .

1.3 Use scenarios

Index usage scenarios :

1. Large amount of data , Often need to query , But it is often necessary to insert , Few modification operations .
2. Insensitive to space , More sensitive to time .

1.4 Use

Create a primary key constraint （PRIMARY KEY), Unique constraint （UNIQUE), Foreign key constraints (FOREIGN KEY) when , The index of the corresponding column will be created automatically

• Look at the index

show index from  Table name ;

• Create index

create index  Index name  on  Table name ( Name )

Case study ： Create a student table ,name Index of field

• Delete index

Case study ： Delete... From the student table name Index of field

drop index  Index name  on  Table name ;

2. Business

2.1 Transaction scenario

Case study :

--  The test table 
 create table accout(id int, name varchar, balance int);
--  insert data 
 insert into accout values(1, ' Zhang San ', 1000), (2, ' Li Si ', 2000);

Now Zhang San transfers money to Li Si 500 element

-- sql sentence 
update accout set balance = balance - 500 where name = ' Zhang San ';
update accout set balance = balance + 500 where name = ' Li Si ';

How to finish the first sentence sql after , There's something wrong with the database , Then Zhang San is less 500, But Li Si did not increase 500.

So how to deal with such emergencies ?

Solution : Either both of them are finished , Or none of them

notes : It's not true that none of them are implemented , But through recovery , Restore the influence of the previous operation , And this process is called ’ Roll back ’.

So how does the transaction rollback of the database do ?

Every operation in the database , There are records inside , If there is a problem in the middle of the transaction , According to the previous records , To deal with .

2.2 Use of transactions

1)  Open transaction 
start transaction;
2)  perform sql sentence 
3)  Roll back or commit 
rollback/commit;
 notes : rollback It's all failure ,commit It's all success 

So the above transfer sql You can write like this

start transaction;
update accout set balance = balance - 500 where name = ' Zhang San ';
update accout set balance = balance + 500 where name = ' Li Si ';
commit;

2.3 The basic characteristics of affairs

1. Atomic Atomicity ： Atomicity requires that all statements either succeed , Or not all of it , The unsuccessful part
2. Consistency Uniformity ： It stipulates before and after the submission of things , There can only be the state of things before submission and the state of things after submission , From one consistent state to another consistent state , There can be no intermediate process state .
3. Durability persistence ： When a thing is submitted , The database state has changed forever . The data has been updated to the hard disk .
4. Isolation Isolation, ： The isolation of things , Based on atomicity and consistency , Because things are atomized , Quantized , therefore , Things can be executed concurrently in the form of multiple atomic packages , however , Everything doesn't interfere with each other .

To solve the problems caused by concurrent transaction execution , MySQL When the database is introduced " Isolation level ", Users can choose a level suitable for their current business scenario , So what are the problems caused by concurrent transaction execution ?

1. Dirty reading problem .( Read dirty data )

A business A, In the process of execution , A series of modifications have been made to the data , Before submitting to the database ( Before completing the transaction ), Another business B, Read the corresponding data , At this point B The data read are all temporary results , The follow-up may be immediately A It's changed . here B The reading behavior of is " Dirty reading ".

for instance :

I was in the exam , Answering questions , The students behind me are next to me , Secretly observe , And then put " answer " Take it away , Later, I found that the answer to this question is incorrect , I revised it again , At this time, this student saw " answer " Namely ’ Dirty data ’, The process of observation is ’ Dirty reading ’. How to solve it , I made an appointment with my classmates at a certain time before the exam , I checked , You can ’ Observe ’.

The solution is to lock the read operation , It is equivalent to reducing the degree of concurrency , Reduced efficiency , Improved isolation

2. It can't be read repeatedly

Business A After submitting , Business B Just started reading ( Reading is locked ) And then in B During the execution of ,A Once again , Modified the data . here B In execution , The results of two read operations may be inconsistent .

for instance :

Modify the above example , After the appointed time, the students still observed , I revised the answer again , The final answer is inconsistent . How to solve : Don't read before the appointment ( Lock added before ), When you observe , I won't change the answer .

So the isolation is improved , Reduced concurrency , More accurate data , Reduced efficiency .

3. Fantasy reading

Business B During reading , Business A It's been modified , No direct modification B Read data , But it affects B Read result set transactions B The result sets read twice are different .
This is “ Fantasy reading ” Unreal reading is equivalent to a special situation that cannot be repeated .

for instance :

It is agreed that I will not change when my classmates observe , So I went to change the answer of the multiple-choice question , Lead to different answers , How to solve : Be strict with your classmates ’ Observe ’ I can't do how to write , You can look at the questions in the back , So as to ensure that the read and write operations are strictly executed in serial ( After one execution , To execute another )

At this time, the isolation is the highest , Minimum concurrency , The data accuracy is the best , But the slowest .

MySQL Four gears are provided for us in , We are free to choose

Restriction level	effect
read uncommitted	Concurrency is the strongest , Isolation is the weakest .
read committed	Only data after submission can be read , Solved the problem of dirty reading . Concurrency decreases , Isolation has increased a little .
repeatable read	There are restrictions on reading and writing , Solved the unrepeatable read problem . The concurrency ability is further reduced , Isolation is further increased .
serializable	Strict serial execution , The phantom reading problem was solved . Minimum concurrency , Highest isolation , But it's slow