Analysis of shardingsphere core source code

Download the source code （ Choose your favorite version ）

https://github.com/apache/incubator-shardingsphere/archive/4.0.0-RC2.zip

2、 Import project to IDEA

3、sharding-core-parse compile

Note that this project is compiled first because ： Some classes in the project need to rely on antlr4 SQL Parsing engine generation , Otherwise straight Connect and compile sharding The project reports the error that some classes cannot be found .

When compiling this project, you should pay attention to ：lifecycle Under the install 

4、 Compile the whole project

 

Pay attention to the use of maven-plugin compile , And remember to skip the test part when compiling

 

Translation time ：sharding-sql-test The project will report an error , Just ignore it .

Compile the complete ： function demo If it can be executed normally, the compilation is successful  

Apache Incubation version - Module partition

Kernel analysis  

Data fragmentation - Execution process

  Parsing engine

The parsing process is divided into lexical parsing and grammatical parsing . A lexical parser is used to SQL Disassembled into atomic symbols that cannot be further divided Number , be called Token. And according to the dictionaries provided by different database dialects , Categorize it as a keyword , expression , Literally Quantities and operators . Using the syntax parser again will SQL Convert to an abstract syntax tree .

SELECT id, name FROM t_user WHERE status = 'ACTIVE' AND age > 18

sql Analyze the overall structure ： 

Routing engine

The built-in fragmentation strategy can be roughly divided into mantissa modulus 、 Hash 、 Range 、 label 、 Time and so on . Configured by the user The fragmentation strategy is more flexible , You can customize the composite partition strategy according to the user's needs . 

SQL rewrite  

Is the truth obtained after routing

Execution engine  

The results merge  

Multiple data result sets to be obtained from each data node , Combine into a result set and return to the requesting customer correctly End , It's called result merging .

Distributed primary key  

Built in generator support ：UUID、SNOWFLAKE, And extract the distributed primary key generator The interface of , It is convenient for users to realize the user-defined self increasing primary key generator .

SNOWFLAKE Snowflake algorithm

It can ensure that the primary keys of different processes are not repeated , Ordering of primary keys of the same process . Binary form package contain 4 part , The sub table from high to low is ：1bit Sign bit 、41bit Time stamp bit 、10bit Working process bit and 12bit Serial number position .

• Sign bit (1bit)

Reserved symbol bits , Constant to zero .

• Time stamp bit (41bit)

41 The number of milliseconds a bit's timestamp can hold is 2 Of 41 The next power , The number of milliseconds used in a year is ：365 * 24 * 60 * 6 0 * 1000 Math.pow(2, 41) / (365 * 24 * 60 * 60 * 1000L) = 69.73 Years don't repeat ;

• Work progress bit (10bit)

The logo is on the Java It's the only thing in the process , If it is a distributed application deployment, it should ensure that each working process id It's different Of . This value defaults to 0, It can be set by properties .

• Serial number position (12bit)

This sequence is used to generate different... In the same millisecond ID. If the number generated in this millisecond exceeds 4096(2 Of 12 The next power ), Then the generator will wait until the next millisecond to continue to generate .

Mybatis-SQL Execution process  

The source code structure of sub database and sub table is as follows