重写from语句中的mysql子查询

Question

这个查询很复杂，但所有单独的部分都可以工作。

查询从用户表中获取与传入的客户端id匹配的所有用户（有一个用户/客户端链接表）

具有挑战性的部分是获取用户已完成哪些 SCORM 课程的详细信息的中间部分（通过或完成的跟踪数据）。但是这个scorm跟踪表并没有scorm对象所在课程的数据，所以我必须计算出课程，课程中的scorm包，scorm包中的sco对象，才能成功确定是否用户已通过课程。

这曾经是在 PHP 中嵌套循环之后的循环内大约 15 个查询，然后在吐出页面/下载 csv 之前作为数组存储在 RAM 中。对于现在运行的用户数量来说，这真的很慢并且超时。

所以我必须找到更好的解决方案。

我已经完成了其他 4 个类似的查询页面——这是最后一个让我失望的页面，因为它是一个总结，即它计算他们完成了多少课程，然后将与他们已经分配的课程数量进行比较，并且给我们一个仪表板。

除了子查询中的 u.id 引用外，它正在工作。

SELECT DISTINCT u.id, u.firstname, u.lastname, u.username,

/* this part of the query gets the number of COMPLETED courses only   */
(SELECT COUNT(CASE WHEN refname=1  THEN 1 ELSE NULL END) AS complete  
FROM
    (SELECT CASE
    /* first check if the number of completions in the scorm tracking table is zero */
    WHEN COUNT(CASE WHEN t1.element = 'cmi.core.lesson_status' AND (t1.value = 'passed' OR t1.value = 'completed') THEN 1 ELSE NULL END) = 0 THEN 0 
    /* then do the original mod */
        WHEN (MOD(COUNT(CASE WHEN t1.element = 'cmi.core.lesson_status' AND (t1.value = 'passed' OR t1.value = 'completed') THEN 1 ELSE NULL END),
            (SELECT COUNT(sc9.id) AS scoes FROM mdl_scorm AS s9 LEFT JOIN mdl_scorm_scoes AS sc9 ON s9.id = sc9.scorm WHERE s9.course = c1.id AND sc9.launch = ''))) > 0 THEN 0
        ELSE 1
    END AS refname
    FROM mdl_user AS u1
    LEFT JOIN mdl_scorm_scoes_track AS t1 ON t1.userid = u1.id
    LEFT JOIN mdl_scorm AS s1  ON s1.id = t1.scormid
    LEFT JOIN mdl_course AS c1 ON c1.id = s1.course
    LEFT JOIN mdl_scorm_scoes AS sc1 ON s1.id = sc1.scorm
    WHERE u1.id = u.id 
    /* **this u.id is part of the OUTERMOST QUERY AND ISN'T SEEN BY THE INNER MOST NEST**    */
    AND sc1.launch = ''    
    GROUP BY c1.id, t1.attempt) AS tblname) 
AS complete,

/* this counts how many courses user has been allocated */
(SELECT COUNT(DISTINCT CONCAT(s2.course, t2.attempt)) AS allocated
FROM mdl_scorm AS s2
LEFT JOIN mdl_scorm_scoes_track AS t2 ON s2.id = t2.scormid
LEFT JOIN mdl_scorm_scoes AS sc2 ON sc2.id = t2.scoid
WHERE t2.userid = u.id),

sfhuc.clientid

FROM mdl_user AS u
LEFT JOIN mdl_sfh_user_client AS sfhuc ON sfhuc.userid = u.id
LEFT JOIN mdl_user_info_data ud_employeeid ON ud_employeeid.userid = u.id AND ud_employeeid.fieldid = 1 
LEFT JOIN mdl_user_info_data ud_jobdescription ON ud_jobdescription.userid = u.id AND ud_jobdescription.fieldid = 3
LEFT JOIN mdl_sfh_client AS sfhc ON sfhc.id = sfhuc.clientid
LEFT JOIN mdl_scorm_scoes_track AS t ON t.userid = u.id

WHERE 1=1
AND u.deleted = 0
AND u.timecreated > 0
AND sfhc.id = 2076441
GROUP BY u.id
ORDER BY u.lastname, u.firstname

Answer 1

• 我在FROM（又名“派生表”）中看不到任何子查询。

• 有些表只引用一列；看看是否可以删除表引用：

    sfhc:  (id)
    t:  (userid)
    sc2:  (id)
    c1:  (id)
    u1:  (id)
  

• 缩短代码有助于提高可读性：（注意布尔值被视为 1（真）或 0（假），使 SUM 工作。）

    COUNT(CASE WHEN t1.element = 'cmi.core.lesson_status'
        AND  (t1.value = 'passed'
          OR  t1.value = 'completed')
       THEN 1 ELSE NULL END
  
    -->
  
    SUM( t1.element = 'cmi.core.lesson_status' AND
         t1.value IN ('passed', 'completed') )
  

• CONCAT 是不必要的：

    COUNT(DISTINCT CONCAT(s2.course, t2.attempt))
  
    -->
  
    COUNT(DISTINCT s2.course, t2.attempt)
  

• 这些似乎完全没用：

    LEFT JOIN  mdl_user_info_data ud_employeeid
          ON  ud_employeeid.userid = u.id
         AND  ud_employeeid.fieldid = 1
  
    LEFT JOIN  mdl_user_info_data ud_jobdescription
          ON  ud_jobdescription.userid = u.id
         AND  ud_jobdescription.fieldid = 3
  

• 可能的索引：

    sfhuc:  (userid, clientid)
    t2:  (userid, scoid, scormid, attempt)
    t1:  (attempt, scormid, userid, value, element)
    mdl_scorm_scoes:  (launch, scorm)
  

这些都不一定能“解决”您的问题，但主要指向需要调查或重新思考的领域。