• Mysql 中Left/Right join on后面and和where条件查询的差异-Mysql SQL运算符是有优先级


    一、Mysql中Left/Right join on后面and和where条件查询的差异

    1、建两张测试表,一张商户定义表、一张商户操作状态明细表

    1)商户定义表

    CREATE TABLE hope.merchant (
      MERCHANT_id varchar(100) COLLATE utf8mb4_unicode_ci DEFAULT NULL
    ) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;

    2)商户操作明细表

    CREATE TABLE trade_info (
      MERCHANT_NO varchar(100) COLLATE utf8mb4_unicode_ci DEFAULT NULL,
      TRADE_STATUS varchar(30) COLLATE utf8mb4_unicode_ci DEFAULT NULL,
      CREATE_TIME datetime DEFAULT current_timestamp()
    ) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;

    2.测试Left/Right join on后面and和where条件查询的差异

    普通left join演示

    select a.MERCHANT_id,b.MERCHANT_NO,b.TRADE_STATUS from hope.merchant a left join hope.trade_info b
      on a.MERCHANT_id=b.MERCHANT_NO;

    select count(1) from hope.merchant a left join hope.trade_info b
      on a.MERCHANT_id=b.MERCHANT_NO;--》82行数据

    结果:可以看到,部分商户在某段时间内事没有任何操作的,所以,左表hope.merchant的商户,没有对应的操作,会用Null补足。

    1)Left/Right join on后面带and

     select a.MERCHANT_id,b.MERCHANT_NO,b.TRADE_STATUS,b.CREATE_TIME from hope.merchant a left join hope.trade_info b
      on a.MERCHANT_id=b.MERCHANT_NO and b.CREATE_TIME>'2018-12-19 19:00:00';

    select count(1) from hope.merchant a left join hope.trade_info b
      on a.MERCHANT_id=b.MERCHANT_NO and b.CREATE_TIME>'2018-12-19 19:00:00';  ----》69行数据

    2)Left/Right join on后面带where

    select a.MERCHANT_id,b.MERCHANT_NO,b.TRADE_STATUS,b.CREATE_TIME from hope.merchant a left join hope.trade_info b
      on a.MERCHANT_id=b.MERCHANT_NO where b.CREATE_TIME>'2018-12-19 19:00:00';

    select count(1) from hope.merchant a left join hope.trade_info b
      on a.MERCHANT_id=b.MERCHANT_NO where b.CREATE_TIME>'2018-12-19 19:00:00'; --》43行数据

    3)对比结果

    语句一是on and ,一句是 on where,下的条件类似,但是结果却是完全不同。带where的过滤掉了左连接中左表没有任何操作记录的商户号,

    但是使用and的却没有,并得到了我想要的结果。Why???

    二、Mysql 运算符是有优先级

    1.总结SQL优先级

    where是对已经搜索出的结果的过滤条件

    on是两者join形成新表时用到的匹配条件

    大家可能看出一些规律,在on的情况下and(或者or)与on同时对前面的集合起作用,而存在where的时候,先对集合进行on条件的抽取,在根据where的条件进行抽取.这是个值得注意的地方.
    说下原因:
    (1)ON后面的筛选条件主要是针对的是关联表【而对于主表筛选条件不适用】;
    (2)对于主表的筛选条件应放在where后面,不应该放在ON后面;
    (3)对于关联表我们要区分对待。如果是要条件查询后才连接应该把查询件放置于ON后;如果是想再连接完毕后才筛选就应把条件放置于where后面;
    (4)对于关联表我们其实可以先做子查询再做join。

    2.总结sql类型,尽量不写子查询

    1)对于关联表我们其实可以先做子查询再做join,此SQL运行3.5秒

    select MERCHANT_ID,case when rait is null then 0 else  rait end
    from (
    select MERCHANT_ID,
    TRUNCATE(count(case when c.trade_status ='SUCCESS' then 1 else null end)*100/count(case when c.trade_status !='FAILED' then 1 else null end),2) rait 
    from ifpay_ccpay.merchant a left join (
    select * from ifpay_ccpay.trade_info b where b.CREATE_TIME > date_add(now(), interval - 30 minute)
    and b.method='H5') c on a.MERCHANT_ID=c.MERCHANT_NO group by a.MERCHANT_ID) d;

    2)不做子查询运行14ms

    select MERCHANT_ID,case when rait is null then 0 else  rait end
    from (
    select MERCHANT_ID,
    TRUNCATE(count(case when b.trade_status ='SUCCESS' then 1 else null end)*100/count(case when b.trade_status !='FAILED' then 1 else null end),2) rait 
    from ifpay_ccpay.merchant a left join ifpay_ccpay.trade_info b on a.MERCHANT_ID=b.MERCHANT_NO
    and b.CREATE_TIME > date_add(now(), interval - 30 minute)
    and b.method='H5'  group by a.MERCHANT_ID) D;

  • 相关阅读:
    CARLA——The External Sensor Interface (ESI)
    Eventbased Vision meets Deep Learning on Steering Prediction for Selfdriving Cars
    DDD20 EndtoEnd Event Camera Driving Dataset: Fusing Frames and Events with Deep Learning for Improved Steering Prediction
    CARLA——Core implementations: synchrony, snapshots and landmarks
    Generative Adversarial Imitation Learning
    EventBased Vision Enhanced: A Joint Detection Framework in Autonomous Driving
    CARLA——Sensors in CARLA
    Imitation Learning via OffPolicy Distribution Matching
    CARLA入门
    颜色
  • 原文地址:https://www.cnblogs.com/xibuhaohao/p/10148248.html
Copyright © 2020-2023  润新知