【MyBatis】【SQL】删除最快纪录诞生，从一千万条记录中删除八百万条仅用2分6秒

在 https://www.cnblogs.com/xiandedanteng/p/11669629.html 里我做个一个循环按时间查ID并删除之的程序，运行时间是4分7秒。

但是这个程序走了很多次没有必要的IO,也就是把不需要的指比如要删除的ID在Java所在的机器和DB服务器中传来传去，这个过程在没必要的耗时，不如把日期扔给DB告诉它删除之前的记录然后返回修改的记录数就行了，于是我书写了以下SQL语句：

<?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>
</mapper>

对应的接口是这样的：

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);
}

Java程序则简单多了：

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 BatchDelete2 {
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.deleteByDate("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;
    }
}

控制台输出也很有喜感：

All deleted=8035199
 INFO [main] - Time elapsed:2m6s.

数据库里也达到了想要的结果：

时间减半，相对于循环删除的方案https://www.cnblogs.com/xiandedanteng/p/11669629.html ，优势是显而易见的。

本作代码下载地址：https://files.cnblogs.com/files/xiandedanteng/InsertMillionComparison2019101014.rar

--END-- 2019年10月14日10:44:47