【MyBatis】【SQL】没有最快，只有更快，从一千万条记录中删除八百万条仅用1分9秒

这次直接使用delete from emp where cdate<'2018-02-02',看看究竟会发生什么。

Mapper里写好SQL：

<?xml version="1.0" encoding="UTF-8" ?>
<!DOCTYPE mapper PUBLIC "-//mybatis.org//DTD Mapper 3.0//EN" 
                    "http://mybatis.org/dtd/mybatis-3-mapper.dtd" >
<mapper namespace="com.hy.mapper.EmpMapper">
    <select id="selectById" resultType="com.hy.entity.Employee">
        select id,name,age,cdate as ctime  from emp where id=#{id}
    </select>
    
    <insert id="batchInsert">
        insert into emp(name,age,cdate)
        values
        <foreach collection="list" item="emp" separator=",">
            (#{emp.name},#{emp.age},#{emp.ctime,jdbcType=TIMESTAMP})
        </foreach>
    </insert>
    
    <insert id="singleInsert">
        insert into emp(name,age,cdate)
        values (#{name},#{age},#{ctime,jdbcType=TIMESTAMP})
    </insert>
    
    <select id="selectIdsByDate"  resultType="java.lang.Long">
        select id from emp where cdate&lt;#{date,jdbcType=TIMESTAMP} limit 10000
    </select>
    
    <delete id="deleteByIds">
        delete from emp where id in
        <foreach collection="list" open="(" close=")" separator="," item="id" index="i">
            #{id}
        </foreach>
    </delete>
    
    <delete id="deleteByDate">
        delete from emp where id in (select id from (select id from emp where cdate&lt;#{date,jdbcType=TIMESTAMP}) as tb)
    </delete>
    
    <delete id="deleteEmpByDate">
        delete from emp where cdate&lt;#{date,jdbcType=TIMESTAMP}
    </delete>
</mapper>

接口也做上：

package com.hy.mapper;

import java.util.List;

import org.apache.ibatis.annotations.Param;

import com.hy.entity.Employee;


public interface EmpMapper {
    Employee selectById(long id);
    int batchInsert(List<Employee> emps);
    // 用@Param标签指明和SQL的参数对应能避免出现org.apache.ibatis.binding.BindingException异常
    int singleInsert(@Param("name")String name,@Param("age")int age,@Param("ctime")String ctime);
    
    List<Long> selectIdsByDate(String date);
    
    int deleteByIds(List<Long> ids);
    
    int deleteByDate(String date);
    
    int deleteEmpByDate(String date);
}

代码写好：

package com.hy.action;

import java.io.Reader;
import java.util.ArrayList;
import java.util.List;

import org.apache.ibatis.io.Resources;
import org.apache.ibatis.session.SqlSession;
import org.apache.ibatis.session.SqlSessionFactory;
import org.apache.ibatis.session.SqlSessionFactoryBuilder;
import org.apache.log4j.Logger;

import com.hy.entity.Employee;
import com.hy.mapper.EmpMapper;

public class BatchDelete3 {
private static Logger logger = Logger.getLogger(SelectById.class);
    
    public static void main(String[] args) throws Exception{
        long startTime = System.currentTimeMillis();
        
        Reader reader=Resources.getResourceAsReader("mybatis-config.xml");
        
        SqlSessionFactory ssf=new SqlSessionFactoryBuilder().build(reader);
        reader.close();
        
        SqlSession session=ssf.openSession();
        
        try {
            EmpMapper mapper=session.getMapper(EmpMapper.class);
                
            int changed=mapper.deleteEmpByDate("2018-02-02");
            session.commit();
            
            System.out.println("All deleted="+changed);
        }catch(Exception ex) {
            logger.error(ex);
            session.rollback();
        }finally {
            session.close();
            
            long endTime = System.currentTimeMillis();
            logger.info("Time elapsed:" + toDhmsStyle((endTime - startTime)/1000) + ".");
        }
    }
    
    // format seconds to day hour minute seconds style
    // Example 5000s will be formatted to 1h23m20s
    private static String toDhmsStyle(long allSeconds) {
        String DateTimes = null;
        
        long days = allSeconds / (60 * 60 * 24);
        long hours = (allSeconds % (60 * 60 * 24)) / (60 * 60);
        long minutes = (allSeconds % (60 * 60)) / 60;
        long seconds = allSeconds % 60;
        
        if (days > 0) {
            DateTimes = days + "d" + hours + "h" + minutes + "m" + seconds + "s";
        } else if (hours > 0) {
            DateTimes = hours + "h" + minutes + "m" + seconds + "s";
        } else if (minutes > 0) {
            DateTimes = minutes + "m" + seconds + "s";
        } else {
            DateTimes = seconds + "s";
        }

        return DateTimes;
    }
}

然后塞了一千万条数据一执行，本以为会出现超时异常，回滚段异常，log区异常之类的，结果完全没有，反而还跑出了个最快结果：

All deleted=8035199
 INFO [main] - Time elapsed:1m9s.

数据库的情况也证实了删除操作的正确性：

看来MySql这边千万级数据要删除也就是直接进行的事情，不知道在拿另一环境中的的21张三四百万级的Oracle数据库实验又会是怎样的结果。

凭感觉，无论是插值还是删除，我虚拟机上的MySql（mysql Ver 14.14 Distrib 5.6.45, for Linux (x86_64) using EditLine wrapper）比单位实装的Oracle要迅速多了。

--END-- 2019年10月14日14:23:54