Question

我是编程新手，来自会计背景。我已经写了这个查询，但效果非常慢。我想知道是否有一种方法可以优化它的速度。

我要从中获取的表：

主表

enter image description here

事件表 enter image description here

结果表 enter image description here

认证表 enter image description here

以下是查询：

SELECT a.lgName AS lgname
  , a.wardName AS wardName
  , a.pUnitName AS pUnitName
  , SUM(a.pvc_collected) AS pvc
  , SUM(a.voter_reg_no) AS purvs
  , (   SELECT COUNT(pUnitName)
        FROM master
        WHERE wardName = a.wardName) AS No_Poll_Unitss
  , (   SELECT COUNT(reportedpu)
        FROM master
        WHERE wardName = a.wardName
        AND reportedpu = 1) AS reportedpu
  , (   SELECT COUNT(pUnitName)
        FROM master
        WHERE pUnitName = a.pUnitName) AS No_Poll_Unitss
  , (   SELECT COUNT(reportedpu)
        FROM master
        WHERE pUnitName = a.pUnitName
        AND reportedpu = 1) AS reportedpu
  , (   SELECT SUM(total)
        FROM accreditation
        WHERE pu_name IN (   SELECT pUnitName
                             FROM master
                             WHERE pUnitName = a.pUnitName)) AS acr
  , (   SELECT SUM(cno)
        FROM res
        WHERE pUnitName IN (   SELECT pUnitName
                               FROM master
                               WHERE pUnitName = a.pUnitName)
        AND category = 'tvc') AS tvc
  , (   SELECT SUM(cno)
        FROM res
        WHERE pUnitName IN (   SELECT pUnitName
                               FROM master
                               WHERE pUnitName = a.pUnitName)
        AND category = 'apc') AS apc
  , (   SELECT SUM(cno)
        FROM res
        WHERE pUnitName IN (   SELECT pUnitName
                               FROM master
                               WHERE pUnitName = a.pUnitName)
        AND category = 'ivc') AS ivc
  , (   SELECT SUM(cno)
        FROM res
        WHERE pUnitName IN (   SELECT pUnitName
                               FROM master
                               WHERE pUnitName = a.pUnitName)
        AND category = 'pdp') AS pdp
  , (   SELECT SUM(cno)
        FROM res
        WHERE pUnitName IN (   SELECT pUnitName
                               FROM master
                               WHERE pUnitName = a.pUnitName)
        AND category = 'lp') AS lp
  , (   SELECT SUM(cno)
        FROM res
        WHERE pUnitName IN (   SELECT pUnitName
                               FROM master
                               WHERE pUnitName = a.pUnitName)
        AND category = 'adc') AS adc
  , (   SELECT SUM(cno)
        FROM res
        WHERE pUnitName IN (   SELECT pUnitName
                               FROM master
                               WHERE pUnitName = a.pUnitName)
        AND category = 'sdp') AS sdp
  , (   SELECT SUM(cno)
        FROM res
        WHERE pUnitName IN (   SELECT pUnitName
                               FROM master
                               WHERE pUnitName = a.pUnitName)
        AND category = 'adp') AS adp
  , (   SELECT SUM(cno)
        FROM res
        WHERE pUnitName IN (   SELECT pUnitName
                               FROM master
                               WHERE pUnitName = a.pUnitName)
        AND category = 'other') AS oth
FROM master AS a
GROUP BY a.pUnitName, a.userCode
ORDER BY a.userCode ASC;

Answer 1

查询是非常棘手的，但是我想尝试的第一件事是以这种方式将所有这些子查询放入一个子查询中：

SELECT *
FROM master AS A
CROSS JOIN (   SELECT SUM(IF(category = 'adp', cno, 0)) AS adp
                 -- All other conditions
                 , SUM(IF(category = 'other', cno, 0)) AS other
               FROM res AS R
               WHERE pUnitName IN (   SELECT pUnitName
                                      FROM master
                                      WHERE pUnitName = A.pUnitName)
               AND category = 'other') AS oth;

这应该只运行一次从res表中进行选择，并为每个所需条件计算总和。另外，从子查询中删除该pUnitName条件当然应该是可能的，但是它需要更多有关数据集的知识。

Answer 2

在我们的案例中，最好的方法是在进行连接之前先汇总。您需要四种不同的聚合：

master由pUnitName
master由WardName
accreditation由pUnitName
res通过pUnitName`

然后将LEFT JOIN一起使用：

SELECT m.pUnitName, m.userCode,
       SUM(m.pvc_collected) as pvc,
       SUM(m.voter_reg_no) AS purvs,
       mw.Num_Poll_Units, mw.reportedpu,
       mu.Num_Poll_Units, mu.reportedpu,
       a.acr,
       r.tvc, r.apc, . . .
FROM master m LEFT JOIN
     (SELECT m2.wardName, COUNT(*) as Num_Poll_Units,
             SUM(m2.reportedpu = 1) as reportedpu
      FROM master m2
      GROUP BY m2.wardName
     ) mw
     USING (wardname) LEFT JOIN
     (SELECT m2.pUnitName, COUNT(*) as Num_Poll_Units,
             SUM(m2.reportedpu = 1) as reportedpu
      FROM master m2
      GROUP BY m2.pUnitName
     ) mu
     USING (pUnitName) LEFT JOIN
     (SELECT a.pu_name, SUM(a.total) as acr
      FROM accreditation a
      GROUP BY a.pu_name
     ) a
     ON a.pu_name = m.pUnitName LEFT JOIN
     (SELECT r.pUnitName,
             SUM(CASE WHEN category = 'tvc' THEN cno ELSE 0 END) as tvc,
             SUM(CASE WHEN category = 'apc' THEN cno ELSE 0 END) as apc,
             . . .
      FROM res r
      GROUP BY r.pUnitNmae
     ) r
     USING (pUnitName)
GROUP BY m.pUnitName, m.userCode,
         mw.Num_Poll_Units, mw.reportedpu,
         mu.Num_Poll_Units, mu.reportedpu,
         a.acr,
         r.tvc, r.apc, . . .
ORDER BY m.userCode ASC;

注意：最外面的查询未按WardName进行聚合。目前尚不清楚您对这些专栏的真正要求。您可能需要执行以下操作：

从mw.Num_Poll_Units中删除mw.reportedpu和GROUP BY。
将SELECT更改为SUM(mw.Num_Poll_Units)或AVG(mw.Num_Poll_Units)或使用任何适当的功能。

Answer 3

查询必须完全优化。我提到了一些使其优化的点：

使用交叉应用代替子查询
使用窗口功能
您还可以创建一些索引

您可以在Internet上搜索主题。或者你可以问你的问题。

如何优化SQL查询

3 个答案: