我的子查询减慢了这个过程。有什么优化方法吗？答案

【问题标题】：My subquery is slowing down the process. Any optimization method?我的子查询减慢了这个过程。有什么优化方法吗？
【发布时间】：2017-06-30 16:12:11
【问题描述】：

我用子查询写过代码，但是太慢了。我正在寻找可以优化此查询过程的解决方案。它旨在返回 6 列 - PO 编号、PO 行编号、每个 PO 行的金额、每个 PO 编号的 PO 行计数、每个 PO 编号的行金额总和、按行金额对每个 PO 编号的分类。下面是我写的示例表和实际查询。非常感谢您的帮助！

|PO NUMBER |PO LINE|LINE AMOUNT|TOTAL PO LINES|TOTAL PO AMOUNT|TOTAL PO GROUP|
|1721574   |   1   |   10.76   |      1       |     10.76     |   $0-100     |
|1722154   |   1   |   30.00   |      1       |     30.00     |   $0-100     |
|1723538   |   1   |   15.97   |      4       |     65.63     |   $0-100     |
|1723538   |   2   |   11.23   |      4       |     65.63     |   $0-100     |
|1723538   |   3   |   15.97   |      4       |     65.63     |   $0-100     |
|1723538   |   4   |   22.46   |      4       |     65.63     |   $0-100     |
|1723877   |   1   |   15.70   |      1       |     15.70     |   $0-100     |

查询

select ph.ponumber, 
       pl.poline, 
       pl.polinebasemerchamount, 
       (select count(pl2.poline) 
               from dbo.polineflat as pl2 
               inner join dbo.poheader as ph2 
               on ph2.pokey = pl2.pokey 
               where ph2.pokey = ph.pokey 
               group by ph2.ponumber), 
       (select sum(pl2.polinebasemerchamount) 
               from dbo.polineflat as pl2 
               inner join dbo.poheader as ph2 
               on ph2.pokey = pl2.pokey 
               where ph2.pokey = ph.pokey 
               group by ph2.ponumber), 
       (select case 
                   when sum(pl2.PoLineBaseMerchAmount) between 0 and 100 then '$0-100'
                   when sum(pl2.polinebasemerchamount) between 101 and 500 then '$101-500'
                   when sum(pl2.polinebasemerchamount) between 501 and 1000 then '$501-1000'
                   else '1000+' end
               from dbo.polineflat as pl2 
               inner join dbo.poheader as ph2 
               on ph2.pokey = pl2.pokey 
               where ph2.pokey = ph.pokey 
               group by ph2.ponumber) 
from dbo.poheader as ph
inner join dbo.polineflat as pl on ph.pokey = pl.pokey

【问题讨论】：

看起来你可能正在计算三个不同的总和三到四次..
查询显示不完整
是的，计算四次总和会要了你的命。我的解决方案计算一次，然后有一个执行逻辑的父查询。
您使用的是什么版本的 SQL Server？
大家好，我使用的是最新的 Sql server 2016 Ver。 13.

标签： sql sql-server performance subquery

【解决方案1】：

我会计算一次总和，然后在父查询中执行您的案例：

SELECT ponumber, poline, polinebasemerchamount, polineCount, polineFlatSum,
    CASE 
        WHEN polineFlatSum between 0 and 100 then '$0-100'
        WHEN polineFlatSum between 101 and 500 then '$101-500'
        WHEN polineFlatSum  between 501 and 1000 then '$501-1000'
        ELSE '1000+' 
    END AS polineFlatSumString
from
(
    select ph.ponumber, 
           pl.poline, 
           pl.polinebasemerchamount, 
           (select count(pl2.poline) 
                   from dbo.polineflat as pl2 
                   inner join dbo.poheader as ph2 
                   on ph2.pokey = pl2.pokey 
                   where ph2.pokey = ph.pokey 
                   group by ph2.ponumber) AS polineCount, 
           (select sum(pl2.PoLineBaseMerchAmount) 
                   from dbo.polineflat as pl2 
                   inner join dbo.poheader as ph2 
                   on ph2.pokey = pl2.pokey 
                   where ph2.pokey = ph.pokey 
                   group by ph2.ponumber) AS polineFlatSum,        
    from dbo.poheader as ph
        inner join dbo.polineflat as pl on ph.pokey = pl.pokey
) T

【讨论】：

【解决方案2】：

三个sub-queries 可以使用Cross Apply 组合成一个查询。使用Cross Apply 而非相关子查询的这一优势可以在select 列表中返回多于一列

SELECT ph.ponumber, 
       pl.poline, 
       pl.polinebasemerchamount, 
       oa.poline_count, 
       oa.polinebasemerchamount_sum, 
       CASE 
         WHEN oa.polinebasemerchamount_sum BETWEEN 0 AND 100 THEN '$0-100' 
         WHEN oa.polinebasemerchamount_sum BETWEEN 101 AND 500 THEN '$101-500' 
         WHEN oa.polinebasemerchamount_sum BETWEEN 501 AND 1000 THEN '$501-1000' 
         ELSE '1000+' 
       END AS Range
FROM   dbo.poheader AS ph 
       CROSS Apply (SELECT Count(pl2.poline)              AS poline_count, 
                           Sum(pl2.polinebasemerchamount) AS polinebasemerchamount_sum                               
                    FROM   dbo.polineflat AS pl2 
                    WHERE  pl2.pokey = ph.pokey) oa

注意：我已经删除了子查询中的Group By，因为它应该没用，否则它会在原始查询中引发错误

更新：要进一步改进查询，请创建以下Non clustered indexes

--polineflat table
CREATE NONCLUSTERED INDEX IX_polineflat_pokey
    ON dbo.polineflat (pokey) Include (poline,polinebasemerchamount);  

--poheader table
CREATE NONCLUSTERED INDEX IX_poheader_pokey
    ON dbo.poheader (pokey) Include (ponumber,poline,polinebasemerchamount);

【讨论】：

非常感谢 Prdp。你的解决方案确实有效，我花了大约 4 分 40 秒才得到结果，比我原来的要快得多......仍然有点好奇是否有更快的方法来做到这一点......
@SanK - 我对代码做了一些小改动并添加了 Index.在您的数据库中运行此查询并在此处发布执行计划 brentozar.com/pastetheplan 并分享它。还要提到两个表的索引详细信息
哇，与您以前的解决方案相比，即使没有索引，这也快得惊人。非常感谢 Prdp！
@SanK Gald 它帮助了你。如果你对这个答案感到满意，你可以将此标记为答案

【解决方案3】：

试试这个...

select ph2.ponumber, 
       pl2.poline, 
       pl2.polinebasemerchamount, 
       count(pl2.poline) ,
       sum(pl2.polinebasemerchamount) ,
case  when sum(pl2.PoLineBaseMerchAmount) between 0 and 100 then '$0-100'
      when sum(pl2.polinebasemerchamount) between 101 and 500 then '$101-500'
      when sum(pl2.polinebasemerchamount) between 501 and 1000 then '$501-1000'
           else '1000+' end

from dbo.poheader as ph2
inner join dbo.polineflat as pl2
on ph.pokey = pl.pokey
inner join dbo.polineflat as pl2 
on ph.pokey = pl2.pokey
group by  ph2.ponumber, 
       pl2.poline, 
       pl2.polinebasemerchamount

【讨论】：

谢谢，我刚试过这个。显然，这个查询返回值非常快，但不幸的是，它以与我尝试做的不同的方式生成结果..

【解决方案4】：

看起来你正在使用选择进行分组

select ph.ponumber, 
   pl.poline, 
   pl.polinebasemerchamount, 
   count(pl.poline) as total_count,
   sum(pl.polinebasemerchamount) as total_sum,
case  when sum(pl.PoLineBaseMerchAmount) between 0 and 100 then '$0-100'
  when sum(pl.polinebasemerchamount) between 101 and 500 then '$101-500'
  when sum(pl.polinebasemerchamount) between 501 and 1000 then '$501-1000'
       else '1000+' end as total_group

from dbo.poheader as ph
inner join dbo.polineflat as pl
on ph.pokey = pl.pokey
group by     ph.ponumber, pl.poline, pl.polinebasemerchamount

【讨论】：

感谢 Mehrad，我能够执行您的查询，但我没有得到我想要的结果..尤其是在 Total Sum 和 Total Group 列中..

【解决方案5】：

在这些情况下，我实际上使用#temp 表。我发现它们的性能要好得多，而且我可以更好地控制我想要的结果。它也更容易理解（至少对我和其他阅读它们的人来说）。我回答这篇文章的目的是帮助您利用这种类型的编码。您可以复制粘贴并按 F5，而不是直接给出答案。

注意：请随意命名对您有意义的临时表。您还可以使用 LEFT JOIN/RIGHT JOIN 从 dbo.poheader 获取所有值，并查看这些值是否与您的 #temp 表匹配。您可能需要切换 ID 的 /temp 表以适应您需要的结果。我希望这可以帮助您了解这个想法。

 --Insert Subquery 1 into a #temp1

 select count(pl2.poline) as Count1, ph2.ponumber 
 Into #temp1
 from dbo.polineflat as pl2 
           inner join dbo.poheader as ph2 
           on ph2.pokey = pl2.pokey 
           where ph2.pokey = ph.pokey 
           group by ph2.ponumber)

  --Insert Subquery 2 into #temp2

  select sum(pl2.polinebasemerchamount) as Sum1, ph2.ponumber
  into #temp2
           from dbo.polineflat as pl2 
           inner join dbo.poheader as ph2 
           on ph2.pokey = pl2.pokey 
           where ph2.pokey = ph.pokey 
           group by ph2.ponumber

  --Insert Subquery 3 into #temp3. Pull pl2.polinebasemerchant from #temp2

   select case 
    when sum(#temp2.PoLineBaseMerchAmount) between 0 and 100 then '$0-100'
    when sum(#temp2.polinebasemerchamount) between 101 and 500 then'$101-500'
    when sum(#temp2..polinebasemerchamount) between 501 and 1000 then '$501-1000'
    else '1000+' end 'DollarAmounts',ph2.ponumber
    into #temp3
    from dbo.polineflat as pl2 
    inner join dbo.poheader as ph2 
     on ph2.pokey = pl2.pokey 
     where ph2.pokey = ph.pokey 
     group by ph2.ponumber


 -- Select here with the temp table groups
  select ph.ponumber, 
         pl.poline, 
         pl.polinebasemerchamount,
         #temp1.Count1,
         #temp2.Sum1,
         #temp3.DollarAmounts

   from dbo.poheader as ph
   join dbo.polineflat as pl on ph.pokey = pl.pokey

   --join your temp tables here
   Join #temp1 on #temp1.ponumber = ph.ponumber
   join #temp2 on #temp2.ponumber = ph.ponumber
   join #temp3 on #temp3.ponumber = ph.ponumber


   TRUNCATE TABLE #temp1
   TRUNCATE TABLE #temp2
   TRUNCATE TABLE #temp3

【讨论】：