为什么这个查询花了这么长时间才能执行

Question

几个月前我写了这个查询.它运行良好.但是,这种查询日复一日地执行得越来越慢.

除了相同的表执行,此查询还检查多个表中的帐单历史记录.

这是查询 -

SELECT * FROM (
            SELECT *,                
            (SELECT username FROM users WHERE id = u_bills.UserId) AS username,
            (SELECT first_name FROM users WHERE id = u_bills.UserId) AS first_name,
            (SELECT last_name FROM users WHERE id = u_bills.UserId) AS last_name,
            (SELECT phone FROM users WHERE id = u_bills.UserId) AS phone,
            (SELECT email FROM users WHERE id = u_bills.UserId) AS email,
            (SELECT CPRate FROM cpt WHERE UserId = u_bills.UserId ORDER BY AddedDate DESC LIMIT 0,1) AS cprate,
            (SELECT (SELECT PopName FROM pops WHERE PopId = p.PopName) AS PopFullName FROM u_setupinfos AS p 
            WHERE UserId = u_bills.UserId) AS popname,
            (SELECT active FROM users WHERE id = u_bills.UserId) AS active,                
            (SELECT SUM(PaidAmount) AS PaidAmount FROM u_billhistory 
            WHERE UserId = u_bills.UserId AND MONTH(AddedDate) = MONTH(PaidDate)) AS PaidAmount,
            (SELECT PaidDate FROM u_billhistory 
             WHERE UserId = u_bills.UserId AND MONTH(AddedDate) = MONTH(PaidDate) 
             GROUP BY MONTH(PaidDate)) AS PaidDate,
            (SELECT PaymentMedia FROM u_billhistory 
            WHERE UserId = u_bills.UserId AND MONTH(AddedDate) = MONTH(PaidDate) 
            GROUP BY MONTH(PaidDate)) AS PaymentMedia,   
    (SELECT TransactionId FROM u_billhistory 
            WHERE UserId = u_bills.UserId AND MONTH(AddedDate) = MONTH(PaidDate) GROUP BY MONTH(PaidDate)) AS TransactionId,
            (SELECT GROUP_CONCAT(CONCAT(`BillHisId`,';',`PaidAmount`,';',`PaidDate`) separator '|') AS vals 
FROM u_billhistory 
WHERE UserId = u_bills.UserId 
AND YEAR(PaidDate) = YEAR(CURDATE())) AS TotalPaids
            FROM u_bills) AS m
        WHERE m.username = 'abc'

请查看EXPLAIN图像报告 -

我想组织这个查询.我需要建议使这个查询更快,如3到5秒,而不是1分半钟或更长.我在这个查询上犯了什么错误？

重要的提示-

我在查询中使用了几乎每个列的索引

Answer 1

它很慢,因为你让它为结果中每个记录的每个字段单独查找一个附加表...有时不止一个.您应该使用JOIN而不是嵌套的SELECT.对于某些分组结果,您可以加入一个SELECT执行自己的GROUP BY 的语句.

此外,单个列上的单独索引不会帮助!

想想像电话簿这样的索引.一个简单的电话簿使用姓氏和名字上的聚集索引.然后它还列出了每个项目的地址和电话号码.这就是聚集索引:只是数据的存储顺序.

其他索引就像补充一样."电话号码"上的非聚集索引就像本书末尾的补充,按顺序列出电话号码,然后显示姓氏和名字(但不是地址)."地址"上的索引也会这样做:一个单独的补充,按顺序列出地址,然后给你起名字.

使用这些索引补充,如果您有电话号码并想知道地址,您必须首先查找补充中的电话号码以查找名称,然后在主书中查找名称以获取地址.在这个例子中,手机补充确实有帮助,因为它比电话簿的完整扫描更好.

但是,这里要注意的重要一点是地址补充_ 对此搜索没有帮助,即使地址是您想要找到的确切字段.它可能对其他搜索很有用,但在这种情况下它没有帮助.此外,请记住,对您图书的每次更新都要求您不仅要更新图书,还要更新所有补充内容.它不仅仅是存储空间,还包括更新每个索引所花费的时间.你需要考虑它是否值得.如果您有一堆未使用的索引,那么没有它们会更好.

再回到您的特定问题,关于唯一指标,这将真正帮助你u_bills.username,user.UserId(希望集群为它的表)和两个单个索引u_billhistory.UserId和u_billhistory.PaidDate.

这是重构此查询的初步尝试:

SELECT ub.*, u.username, u.first_name, u.last_name, u.phone, u.email, u.active,
    m.PaidAmount, m.PaymentMedia, m.TransactionId, 
    y.TotalPaids, p.PopName As PopFullName, cpt2.CPRate       

FROM u_bills ub
INNER JOIN users u on u.id = ub.UserId
INNER JOIN u_setupinfos usi on usi.UserId = ub.UserID
INNER JOIN pops p ON p.PopId = usi.PopName
INNER JOIN (
    SELECT UserId, MIN(AddedDate) As cptDate
    FROM cpt
    GROUP BY UserID        
) cpt1 ON cp1.UserID = ub.UserID
INNER JOIN (
    SELECT UserID, AddedDate, MIN(CPRate) AS CPRate
    FROM cpt
    GROUP BY UserID, AddedDate
) cpt2 ON cpt2.UserID = ub.UserID AND cpt2.AddedDate = cpt1.ctpDate
INNER JOIN (
    SELECT UserID, MONTH(PaidDate) PaidMonth, SUM(PaidAmount) AS PaidAmount,
        PaymentMedia, TransactionId --these two fields need an aggregate function of some type!
    FROM u_billhistory
    GROUP BY UserId, MONTH(PaidDate)
) m ON m.UserID = ub.UserId AND m.PaidMonth = MONTH(ub.AddedDate)
INNER JOIN (
    SELECT UserID, YEAR(PaidDate) As PaidYear, 
        GROUP_CONCAT(CONCAT(`BillHisId`,';',`PaidAmount`,';',`PaidDate`) separator '|') AS TotalPaids
    FROM u_billhistory
    WHERE YEAR(PaidDate) = YEAR(CurDate())
    GROUP BY UserId, YEAR(PaidDate)
) y ON Y.UserId = ub.UserId 
WHERE username='abc'

请注意,它并不完整,因为我们没有关于您的架构和数据的所有信息,并且因为似乎有一些缺失GROUP BY和两个缺少的聚合函数.MySql这种方式很糟糕...一个更好的数据库不会允许你的查询运行,直到你修复这些东西.