Question

我有两个问题。

第一部分是关于我在下面做的速度。我正在获取数据集并获取每个DISTINCT，然后加入一个与标准相加/计数的子查询，然后重复这五次。在第一个LEFT JOIN之后我的运行时间为02秒，在第二个LEFT JOIN之后我的运行时间为04秒，但在第三个LEFT JOIN上运行时间为44秒。如果将该查询拉到一边并单独运行，则仅为01秒。完成查询的总运行时间为70秒。

第一部分问题：

在第三次加入时，是什么让这个查询显着变慢？

第二部分：我确信我以低效的方式完成了这项工作。我环顾了一会儿，无法找到一种方法来完成我正在做的事情，同时仍然能够为每个子查询设置不同的标准。

第二部分问题：

有没有更好的方法来完成此查询？

用户1413评论后

编辑。

开始时

SELECT DISTINCT在SQL Server执行计划视图中显示49％的成本。我更改了SELECT DISTINCT正在选择的表格，我的运行时间降到了01秒以下。

SELECT DISTINCT 
    od.location, currentTires, currentAlignments, currentLubes, 
    currentBatteries,currentSiping,currentCarcount
FROM 
    comm.dbo.ordetail as od
LEFT JOIN 
    (SELECT 
         od2.location, SUM(od2.qty_shipped) as currentTires
     FROM 
         comm.dbo.ordetail as od2
     JOIN 
         comm.dbo.invmas as invmas2 ON invmas2.item_num = od2.item#
     JOIN 
         comm.dbo.ordhrd as oh2 ON oh2.location = od2.location
                                AND oh2.inv_date = od2.inv_date
                                AND oh2.invoice# = od2.invoice#
     WHERE 
         (oh2.type_of_rec = '01' OR oh2.type_of_rec = '02')
         AND od2.inv_date >= '2017-02-01 00:00:00.000'
         AND od2.inv_date <= '2017-02-28 00:00:00.000'
         AND ( invmas2.category = 'uhp' OR invmas2.category = 'tour' OR invmas2.category = 'mass' OR invmas2.category = 'suv' OR invmas2.category = 'ltat' OR invmas2.category = 'ltmt' OR invmas2.category = 'lthwy' OR invmas2.category = 'snow' OR invmas2.category = 'stdls' )
    GROUP BY 
        od2.location) as currentTires ON od.location = currentTires.location
LEFT JOIN 
    (SELECT 
         od3.location, SUM(od3.qty_shipped) as currentAlignments
     FROM 
         comm.dbo.ordetail as od3
     JOIN 
         comm.dbo.invmas as invmas3 ON invmas3.item_num = od3.item#
     JOIN 
         comm.dbo.ordhrd as oh3 ON oh3.location = od3.location
                                AND oh3.inv_date = od3.inv_date
                                AND oh3.invoice# = od3.invoice#
     WHERE 
         (oh3.type_of_rec = '01' OR oh3.type_of_rec = '02')
         AND od3.inv_date >= '2017-02-01 00:00:00.000'
         AND od3.inv_date <= '2017-02-28 00:00:00.000'
         AND (od3.item# = '8501' OR od3.item# = '8502')
     GROUP BY 
         od3.location) as currentAlignments ON od.location = currentAlignments.location
LEFT JOIN 
    (SELECT 
         od4.location, SUM(od4.qty_shipped) as currentLubes
     FROM 
         comm.dbo.ordetail as od4
     JOIN 
         comm.dbo.invmas as invmas4 ON invmas4.item_num = od4.item#
     JOIN 
         comm.dbo.ordhrd as oh4 ON oh4.location = od4.location
                                AND oh4.inv_date = od4.inv_date
                                AND oh4.invoice# = od4.invoice#
     WHERE 
         (oh4.type_of_rec = '01' OR oh4.type_of_rec = '02')
         AND od4.inv_date >= '2017-02-01 00:00:00.000'
         AND od4.inv_date <= '2017-02-28 00:00:00.000'
         AND (od4.item# = '200fs' OR od4.item# = '200c' OR od4.item# = '200m' OR od4.item# = '200s')
     GROUP BY 
         od4.location) as currentLubes ON od.location = currentLubes.location
    LEFT JOIN ( SELECT od5.location,SUM(od5.qty_shipped) as currentBatteries
        FROM comm.dbo.ordetail as od5
        JOIN comm.dbo.invmas as invmas5
            ON invmas5.item_num = od5.item#
        JOIN comm.dbo.ordhrd as oh5
            ON oh5.location = od5.location
            AND oh5.inv_date = od5.inv_date
            AND oh5.invoice# = od5.invoice#
        WHERE ( oh5.type_of_rec = '01' OR oh5.type_of_rec = '02' )
            AND od5.inv_date >= '2017-02-01 00:00:00.000'
            AND od5.inv_date <= '2017-02-28 00:00:00.000'
            AND invmas5.manufact = 'inter'
        GROUP BY od5.location) as currentBatteries
        ON od.location = currentBatteries.location
    LEFT JOIN ( SELECT od6.location,SUM(od6.qty_shipped) as currentSiping
        FROM comm.dbo.ordetail as od6
        JOIN comm.dbo.invmas as invmas6
            ON invmas6.item_num = od6.item#
        JOIN comm.dbo.ordhrd as oh6
            ON oh6.location = od6.location
            AND oh6.inv_date = od6.inv_date
            AND oh6.invoice# = od6.invoice#
        WHERE ( oh6.type_of_rec = '01' OR oh6.type_of_rec = '02' )
            AND od6.inv_date >= '2017-02-01 00:00:00.000'
            AND od6.inv_date <= '2017-02-28 00:00:00.000'
            AND invmas6.manufact = 'inter'
        GROUP BY od6.location) as currentSiping
        ON od.location = currentSiping.location
    LEFT JOIN ( SELECT od7.location,COUNT(DISTINCT oh7.invoice#) as currentCarcount
        FROM comm.dbo.ordetail as od7
        JOIN comm.dbo.ordhrd as oh7
            ON oh7.location = od7.location
            AND oh7.inv_date = od7.inv_date
            AND oh7.invoice# = od7.invoice#
        WHERE ( oh7.type_of_rec = '01' OR oh7.type_of_rec = '02' )
            AND od7.inv_date >= '2017-02-01 00:00:00.000'
            AND od7.inv_date <= '2017-02-28 00:00:00.000'
            AND oh7.veh_make != ''
            AND od7.item# != ''
        GROUP BY od7.location) as currentCarcount
        ON od.location = currentCarcount.location
    ORDER BY od.location

示例数据输出：

Answer 1

这里可以做很多事情。对于初学者，下面的以下连接之一可以被删除，因为od5和od6都返回相同的项目。

LEFT JOIN ( SELECT od5.location,SUM(od5.qty_shipped) as currentBatteries
    FROM comm.dbo.ordetail as od5
    JOIN comm.dbo.invmas as invmas5
        ON invmas5.item_num = od5.item#
    JOIN comm.dbo.ordhrd as oh5
        ON oh5.location = od5.location
        AND oh5.inv_date = od5.inv_date
        AND oh5.invoice# = od5.invoice#
    WHERE ( oh5.type_of_rec = '01' OR oh5.type_of_rec = '02' )
        AND od5.inv_date >= '2017-02-01 00:00:00.000'
        AND od5.inv_date <= '2017-02-28 00:00:00.000'
        AND invmas5.manufact = 'inter'
    GROUP BY od5.location) as currentBatteries
    ON od.location = currentBatteries.location
LEFT JOIN ( SELECT od6.location,SUM(od6.qty_shipped) as currentSiping
    FROM comm.dbo.ordetail as od6
    JOIN comm.dbo.invmas as invmas6
        ON invmas6.item_num = od6.item#
    JOIN comm.dbo.ordhrd as oh6
        ON oh6.location = od6.location
        AND oh6.inv_date = od6.inv_date
        AND oh6.invoice# = od6.invoice#
    WHERE ( oh6.type_of_rec = '01' OR oh6.type_of_rec = '02' )
        AND od6.inv_date >= '2017-02-01 00:00:00.000'
        AND od6.inv_date <= '2017-02-28 00:00:00.000'
        AND invmas6.manufact = 'inter'
    GROUP BY od6.location) as currentSiping
    ON od.location = currentSiping.location

至于性能和可读性，请尝试以下内容。注意：我无法对此进行测试，但是，它可能会让有更多时间来调整查询性能的人深入了解。

DECLARE @LowDate DATETIME = '2017-02-01 00:00:00.000'
DECLARE @HighDate DATETIME = '2017-02-28 00:00:00.000'


SELECT
    DetailList.location,
    currentTires = SUM(currentTires), 
    currentAlignments = SUM(currentAlignments),
    currentLubes = SUM(currentLubes),
    currentBatteries = SUM(currentBatteries),
    currentSiping = SUM(currentSiping),         
    carCount = SUM(hasCar)
FROM
(
    SELECT 
        od.location,
        currentLubes=CASE WHEN  ( od.item# = '200fs' OR od.item# = '200c' OR od.item# = '200m' OR od.item# = '200s' ) THEN od.qty_shipped ELSE NULL END,
        currentAlignments=CASE WHEN  ( od.item# = '8501' OR od.item# = '8502' ) THEN od.qty_shipped ELSE NULL END,
        currentSiping = CASE WHEN ( invmas.item_num = 'p15' OR invmas.item_num = 'u15' ) THEN od.qty_shipped ELSE NULL END,
        currentBatteries = CASE WHEN invmas.manufact = 'inter' THEN od.qty_shipped ELSE NULL END,
        currentTires=  CASE WHEN ( invmas.category = 'uhp' OR invmas.category = 'tour' OR invmas.category = 'mass' OR invmas.category = 'suv' OR invmas.category = 'ltat' OR 
                    invmas.category = 'ltmt' OR invmas.category = 'lthwy' OR invmas.category = 'snow' OR invmas.category = 'stdls' ) THEN od.qty_shipped ELSE NULL END,
    hasCar= CASE WHEN (oh.veh_make != '' AND od.item# !='') THEN 1 ELSE 0 END
    FROM
        comm.dbo.ordetail as od
        LEFT JOIN comm.dbo.invmas as invmas ON invmas.item_num = od.item# --FOR currentCarcount
        INNER JOIN comm.dbo.ordhrd as oh ON oh.location = od.location AND oh.inv_date = od.inv_date AND oh.invoice# = od.invoice#
    WHERE
        od.inv_date BETWEEN '2017-02-01 00:00:00.000' AND '2017-02-28 00:00:00.000'
        AND
        oh.type_of_rec IN('01','02')
)AS DetailList
GROUP BY
    DetailList.location

SQL Server 2014 - 带子查询的多个SUM / COUNT

1 个答案: