如果它具有索引,我如何优化MySQL查询,事件

Question

这是我的查询,需要17.9397 sec时间来得到回应:

SELECT allbar.iBarID AS iBarID,
       allbar.vName AS vName,
       allbar.tAddress AS tAddress,
       allbar.tDescription AS tDescription,

  (SELECT COUNT(*)
   FROM tbl_post p
   WHERE p.vBarIDs = allbar.iBarID) AS `total_post`,
       allbar.bar_usbg AS bar_usbg,
       allbar.bar_enhance AS bar_enhance,

  (SELECT count(*)
   FROM tbl_user
   WHERE FIND_IN_SET(allbar.iBarID,vBarIDs)
     AND (eType = 'Bartender'
          OR eType = 'Bar Manager'
          OR eType = 'Bar Owner')) AS countAss,
       allbar.eStatus AS eStatus
FROM
  (SELECT DISTINCT b.iBarID AS iBarID,
                   b.vName AS vName,
                   b.tAddress AS tAddress,
                   (CASE LENGTH(b.tDescription) WHEN 0 THEN '' WHEN LENGTH(b.tDescription) > 0
                    AND LENGTH(b.tDescription) < 50 THEN CONCAT(LEFT(b.tDescription, 50),'...') ELSE b.tDescription END) AS tDescription,
                   b.usbg AS bar_usbg,
                   b.enhance AS bar_enhance,
                   b.eStatus AS eStatus
   FROM tbl_bar b,
        tbl_user u
   WHERE b.iBarID <> '-10') AS allbar

我试过EXPLAIN,这是结果:

有人能解释一下这个EXPLAIN结果吗？

Answer 1

你应该完全重写那个查询,这完全是胡说八道.

在这一部分

  (SELECT DISTINCT b.<whatever>
   FROM tbl_bar b,
        tbl_user u
   WHERE b.iBarID <> '-10') AS allbar

你基本上做的是将表tbl_bar中的每一行与tbl_user中的每一行连接起来.然后过滤tbl_bar,当选择所有内容时(可能MySQL必须在执行此操作之前将所有内容写入临时表中)返回结果集而不重复.你永远不想这样做.特别是当你甚至没有从tbl_user中选择任何东西时.有连接时,请指定它.如果没有,请不要加入这些表或创建连接.我不知道你的表是否或如何连接,但它应该是这样的:

  (SELECT DISTINCT b.<whatever>
   FROM tbl_bar b
   JOIN tbl_user u ON b.user_id = u.id /*or whatever the connection is*/
   WHERE b.iBarID <> '-10') AS allbar

然后你有这个丑陋的子查询.

  (SELECT COUNT(*)
   FROM tbl_post p
   WHERE p.vBarIDs = allbar.iBarID) AS `total_post`,
       allbar.bar_usbg AS bar_usbg,
       allbar.bar_enhance AS bar_enhance,

这是由方式依赖(看你的解释输出).这意味着,对于外部查询的每一行执行此子查询(是的,如上所述,具有交叉连接的那一行).而不是这个子查询,加入外部查询中的表并使用GROUP BY.

到目前为止,查询应该如下所示:

SELECT 
b.iBarID AS iBarID,
b.vName AS vName,
b.tAddress AS tAddress,
b.tDescription AS tDescription,

COUNT(*) AS `total_post`,
allbar.bar_usbg AS bar_usbg,
allbar.bar_enhance AS bar_enhance

FROM
tbl_bar b
JOIN tbl_user u ON b.user_id = u.id
JOIN tbl_post p ON p.vBarIDs = b.iBarID
WHERE b.iBarID <> '-10'
GROUP BY b.iBarID 

(事实上​​,这不是正确的.规则是,SELECT子句中的每一列都应该在GROUP BY子句中或者具有聚合函数(类似count()或max()应用于它.否则显示每个组的随机行.但是这只是一个例子.你必须弄清楚细节.)

现在是最糟糕的部分.

  (SELECT count(*)
   FROM tbl_user
   WHERE FIND_IN_SET(allbar.iBarID,vBarIDs)
     AND (eType = 'Bartender'
          OR eType = 'Bar Manager'
          OR eType = 'Bar Owner')) AS countAss,
       allbar.eStatus AS eStatus

使用FIND_IN_SET()建议,您将多个值存储在一列中.再一次,你永远不想那样做.请阅读此答案Is storing a delimited list in a database column really that bad?,然后重新设计您的数据库.我不会帮你这个,因为这显然是一个单独的问题的东西.

所有这些并没有真正解释EXPLAIN结果.对于这个问题,我将不得不写一个完整的教程,我不会这样做,因为一切都在手册中,一如既往.