I'm almost addicted to it. I can't sleep! Let a bug fuck me twice!

author ： Little brother Fu
Blog ：https://bugstack.cn

precipitation 、 Share 、 grow up , Let yourself and others have something to gain ！

One 、 Preface ： One Bug

I didn't expect one Bug, I was fucked twice ！

I'm probably addicted , I can't sleep , Sit up and open Mac And external display , this Bug There is no reason , Silently looking at the screen with abnormal printing , One is mine , The other one is also mine .

---

Recently, it may be the volume source code , I'm addicted to cigarettes . First, 《 Handwriting Spring》, Then the 《 Handwriting Mybatis》, But I didn't expect a small problem to make me 2 Time ！

Today, this question is mainly reflected in the common Mybatis, When inserting data , We can return the returned value of the library table index through the input parameter object . But through my own handwriting Mybatis, Every time I come back 0, Instead of finally inserting the index value of the library table . Because it's handwritten , Not directly Mybatis, So I will start from the file parsing 、 Object mapping 、SQL Query for 、 The packaging of the results has been checked , But the problem is not here ？！

• This is this. selectKey Configuration of , In the execution of the insert SQL after , Start to get the last index value .
• Usually as long as the configuration is OK , The return object also has a corresponding id Field , Then you can get the return value correctly .PS： Here's the problem , Written by little brother fu Mybatis It turns out that only one 0！

Two 、 analysis ： Abnormal diagnosis

Most of the R & D partners may not have read Mybatis Source code , So it's probably not clear what's going on here , Little brother Fu, draw a picture for you , Tell you what happened so that the result returned is 0 Of .

• Mybatis The process of processing can be divided into two parts , One part is parsing , The other part is to use . When parsing, put Mapper XML Medium insert Label statements are parsed out , At the same time, analyze selectKey label . After the final parsing , Use the parsed statement information MappedStatement The mapping statement class is stored . Convenient for follow-up in DefaultSqlSession When performing an operation , It can be downloaded from Configuration It is obtained from the configuration item for use .

• So here is a very important point , Is to perform insert When inserting , It also contains a query operation . That means , We'll be at once Insert in , Contains two execution statements . a key ：bug It happened here , Why? ？ Because when these two statements are executed at the beginning , When getting links , Each one is to get a new link , So in other words ,insert xxx、select LAST_INSERT_ID() In two connection During connection execution , It's not true , Unable to get the inserted index ID, Only under a link or a transaction ( once commit) Can have the characteristics of transaction , Get the auto increment after inserting data ID.

• And because this part was written at the beginning JdbcTransaction Realization Transaction When the interface gets the connection , Every time it's a new link , The code block is as follows ;
  

  • Here's a link to get , At the beginning, there was no if null The judgment of the , Every time I get the link directly , So this is not a link under the two SQL operation , So we will not get the right result , It is equivalent to just executing alone SELECT LAST_INSERT_ID() So the final query result is 0 When it's over ！ You can test copying this statement to SQL Query tool

3、 ... and 、 shock ： The same pit

But in fact, there is such a link problem , Write in little brother Fu's handwriting Spring Also encountered in .

stay Spring There is a part about transaction processing , In fact, the operation of these transactions is also correct JDBC Packaging operations for , Rely on the links obtained from the data source to manage transactions . And we usually use Spring It is also combined with Mybatis Configure the data source , Then operate multiple transactions under one transaction SQL At the time of statement , How to get the same link . Because from the above case , We know the characteristics of guaranteed transactions , Need to be under the same link , Even if there are multiple operations SQL

Due to the multiple SQL The operation of , It is equivalent to getting a new one every time Session There is a new link from the connection pool , But in order to achieve the characteristics of transactions , Therefore, multiple transactions are required SQL The transaction operation needs to be started before , Whether it's manual or annotation .

The transaction start action processing makes some restrictions on transaction propagation behavior and isolation level , In fact, the more important thing is to make multiple SQL The execution of get the link , The need is the same . So here we introduce ThreadLocal Based on the synchronization feature that it saves information under the same thread operation , In fact, the link obtained from the transaction here , In fact, it is saved to TransactionSynchronizationManager#resources Attribute .

Although it's just a small piece of content , But at the beginning of little brother Fu's handwriting Spring When , It was also left out . Until the time of the test , The link discovery transaction is always unsuccessful , At first, I thought that the whole facet logic was not cut in or my operation mode was wrong . Until the debugging code is checked step by step , It is found that there are many SQL The execution of is not the same link obtained , So there is no way to make the transaction effective .

Four 、 common ： Transaction failure

It may be such a small link problem , Sometimes it will cause a lot of exceptions , If we haven't learned the source code , Then you may not know how such a problem happened . Therefore, they often conduct in-depth research and exploration , So that you can explain a problem , It's simpler and more direct .

So you say , What are the reasons for transaction invalidation ？- Share some common , If you encounter other , You can send it to the comment area to have a look .

1. The database engine does not support transactions ： Here we use MySQL For example , Its MyISAM The engine does not support transaction operations ,InnoDB It's the engine that supports transactions , Generally, support transactions will use InnoDB.https://dev.mysql.com/doc/refman/8.0/en/storage-en… from MySQL 5.5.5 The default storage engine to start with is ：InnoDB, Before the default is ：MyISAM, So it's worth noting , The underlying engine doesn't support transactions. No matter what happens, it's just a waste of time .
2. The method is not public Of ： come from Spring Official documents 【When using proxies, you should apply the @Transactional annotation only to methods with public visibility. If you do annotate protected, private or package-visible methods with the @Transactional annotation, no error is raised, but the annotated method does not exhibit the configured transactional settings. Consider the use of AspectJ (see below) if you need to annotate non-public methods.】@Transactional It can only be used for public On the way , Otherwise, the transaction will not fail , If it is to be used in non public On the way , Can be opened AspectJ The proxy pattern .
3. Has not been Spring management ：// @Service - It is commented out here public class OrderServiceImpl implements OrderService { @Transactional public void placeOrder(Order order) { // ... } }
4. The data source is not configured with transaction manager ： Generally, it comes from the self-developed database routing component @Bean public PlatformTransactionManager transactionManager(DataSource dataSource) { return new DataSourceTransactionManager(dataSource); }
5. Abnormally swallowed .catch Then I ate it directly , Transaction exception cannot be rolled back . At the same time, configure the corresponding exception @Transactional(rollbackFor = Exception.class)

5、 ... and 、 summary ： Learning experience

Many technical problems like this , It all comes from little brother Fu's study of the source code , The first thing is to look at the source code when you encounter problems , Although it is often difficult to smooth out the whole logic , However, a little bit of accumulation will indeed enable R & D personnel to have a more solid understanding of Technology .

So the reason why I write now Spring、 Handwriting Mybatis, I also hope to sort out all such knowledge , Learn the design scheme of complex logic from it 、 Design principles and how to use design patterns to solve the problems of complex scenes .PS： Usually our business code complexity is difficult to reach this level , So I have seen ” God “ after , In the future, it will be easy to design the business undertaken .

The other is to control all kinds of technical details , And accumulated in this experience to apply relevant technical design to some similar SpringBoot Starter Development, etc , Only such breadth 、 Height 、 depth , Can we really improve our R & D ability .PS： It is also to go further on the road of technology , Whether it's advanced development 、 Architects 、CTO！