Question

我有以下查询

SELECT MAKEDATE( LEFT(IF(i_id = 0, Job.FillDate, BillingDate), 4 ) , 
       RIGHT( IF(i_id = 0, Job.FillDate, BillingDate), 3 ) ) as day, 
       IFNULL(ROUND(SUM(GREATEST((IF(i_id = 0 AND ISNULL(BillAmt), (PaidByPlan1+PaidByPlan2+Cashpaid), IF(ISNULL(BillAmt),(PaidByPlan1+PaidByPlan2+Cashpaid), (BillAmt+Cashpaid)))),0.00)),2),0.00) as total_pay, 
       Count(*) as total_transactions 
FROM `job` AS `Job` 
LEFT JOIN `jobtrak` AS `JobTrack` ON (
  `Job`.`id` = `JobTrack`.`job_id` 
  AND  IF(i_id = 0, `Job`.`FillDate`, BillingDate) BETWEEN 2014183 AND 2014212 
  AND LENGTH( CONV( Jobstatus, 10, 2 ) ) NOT IN ( 5, 6, 8 ) 
  AND `Job`.`Active` = 0 ) 
GROUP BY IF(i_id = 0, `Job`.`FillDate`, BillingDate);

以下查询耗尽了我的mysql服务器并导致页面永远不会加载到我的应用程序上。所以，我查看了解释，它显示Job表正在检索所有493345条记录，JobTrack表从其表中获取所有286812条记录。为什么这样做？有什么建议??

Answer 1

`Job`.`id` = `Job`.`job_id`

您正在加入同一张表中的列

Answer 2

这是你的查询，清理了一下，所以我发现它更具可读性：

SELECT MAKEDATE( LEFT(IF(i_id = 0, j.FillDate, BillingDate), 4 ) , 
       RIGHT( IF(i_id = 0, j.FillDate, BillingDate), 3 ) ) as day, 
       IFNULL(ROUND(SUM(GREATEST((IF(i_id = 0 AND ISNULL(BillAmt), (PaidByPlan1+PaidByPlan2+Cashpaid), IF(ISNULL(BillAmt),(PaidByPlan1+PaidByPlan2+Cashpaid), (BillAmt+Cashpaid)))),0.00)),2),0.00) as total_pay, 
       Count(*) as total_transactions 
FROM job j LEFT JOIN
     jobtrak jt
     ON j.id = jt.job_id AND
        IF(i_id = 0, j.FillDate, BillingDate) BETWEEN 2014183 AND 2014212 AND
        LENGTH( CONV( Jobstatus, 10, 2 ) ) NOT IN ( 5, 6, 8 ) AND
        j.Active = 0 
GROUP BY IF(i_id = 0, j.FillDate, BillingDate);

j.Active =0子句中的表达式where正在执行 nothing 。因为您有left join，所以从job表返回所有行，即使谓词的计算结果为false也是如此。所以，我怀疑你的意思是：

SELECT MAKEDATE( LEFT(IF(i_id = 0, j.FillDate, BillingDate), 4 ) , 
       RIGHT( IF(i_id = 0, j.FillDate, BillingDate), 3 ) ) as day, 
       IFNULL(ROUND(SUM(GREATEST((IF(i_id = 0 AND ISNULL(BillAmt), (PaidByPlan1+PaidByPlan2+Cashpaid), IF(ISNULL(BillAmt),(PaidByPlan1+PaidByPlan2+Cashpaid), (BillAmt+Cashpaid)))),0.00)),2),0.00) as total_pay, 
       Count(*) as total_transactions 
FROM job j LEFT JOIN
     jobtrak jt
     ON j.id = jt.job_id AND
        IF(i_id = 0, j.FillDate, BillingDate) BETWEEN 2014183 AND 2014212 AND
        LENGTH( CONV( Jobstatus, 10, 2 ) ) NOT IN ( 5, 6, 8 )
WHERE j.Active = 0 
GROUP BY IF(i_id = 0, j.FillDate, BillingDate);

而JobStatus位于jobtrak表中。如果没有，它也应该移到where。

真正有用的唯一索引是jobtrack(job_id)和job(active)。表达式LENGTH( CONV( Jobstatus, 10, 2 ) )非常不寻常。在大多数业务应用程序中，通常不会比较十进制数的二进制表示的长度。

Answer 3

查询正在检索job表中的所有行，因为没有任何谓词限制从该表返回的行。（引用该表的谓词出现在ON操作的LEFT JOIN子句中，因此这些谓词仅影响从jobtrack表返回的行。）

就从jobtrak表中获取行而言，该表中的每一行可能“匹配”job表中的一行。如果job_trak.job_id是引用job.id的非NULL FOREIGN KEY，那将是真的。

如果要限制从job表返回的行，则需要在WHERE子句中包含谓词。

LEFT JOIN忽略ON子句并检查每一行。为什么？

3 个答案: