带有子查询优化的 SQL Server 存储过程

Question

我有四张桌子：

1. 销售人员 (personId int ,PersonName nvarchar(20)..)
2. personAndDistrict (autoId int,PersonId int,districtId int)：每个在特定地区工作的销售人员
3. PointOfSales (SalesId int,pointOfSalesId int,districtId int,pointOfSalesName nvarchar(20),...) 每个区都有很多 pointOfSales。
4. DailySales (SalesPersonId int,SoldAmount float,PointOfSalesId int, salesDate date,..)：每日销售数据。

为了保存每个销售人员的每日销售额，我首先需要显示他所有销售点的列表，其中包含每个销售点的总和（如果有）。

因此，我们需要显示特定销售人员在特定日期的所有销售点以及积分的总销售量。

我有存储过程：

ALTER PROCEDURE [dbo].[SP_PoinOfSales_GetDataForEntry]
      @SalesPersonId   int=null, @SalesDate  date=null
AS
   SELECT DISTINCT 
       par.SalesPersonId, 
       pos.pointOfsalesID, 
       pos.pointOfSalesName, 
       ISNULL((SELECT SUM(SoldAmount) AS Expr1 
               FROM   
                  dbo.DailySales AS ds 
               WHERE  
                  (SalesDate = CONVERT(VARCHAR, @salesDate, 102)) 
                  AND (SalesPersonId = par.SalesPersonId)  
                  AND (PointOfsalesID = ps.PointOfsalesID)), 0) AS SoldAmount, 
   FROM   
       dbo.PointOfSales AS pos 
   INNER JOIN 
       dbo.PersonAndDistrict AS par ON pos.districtId = par.districtId 
   WHERE  
       (@SalesPersonId IS NULL 
        OR par.SalesPersonId = @SalesPersonId ) 
   GROUP BY 
       pos.districtId, 
       pos.pointOfSalesName, 
       pos.PointOfsalesID, 
       par.SalesPersonId 
   ORDER BY 
       pos.PointOfsalesID, 
       pos.districtId 

仅返回 800 条记录（销售点）需要 6 秒以上！

那么，如何优化这个存储过程在最短的时间内执行？

Answer 1

总的来说，我发现 JOIN 中的内联视图比 SELECT 子句中的性能更好。所以我会这样重写。

SELECT DISTINCT par.salespersonid, 
            pos.pointofsalesid, 
            pos.pointofsalesname, 
            Isnull(t.expr1, 0), 
            as soldamount 
FROM   dbo.pointofsales AS pos 
       INNER JOIN dbo.personanddistrict AS par 
         ON pos.districtid = par.districtid 
       LEFT JOIN (SELECT SUM(soldamount) AS expr1, 
                         pointofsalesid, 
                         salespersonid 
                  FROM   dbo.dailysales AS ds 
                  WHERE  ( salesdate = CONVERT(VARCHAR, @salesDate, 102) ) 
                  GROUP  BY pointofsalesid, 
                            salespersonid) t 
         ON t.salespersonid = par.salespersonid 
            AND t.pointofsalesid = pos.pointofsalesid 
WHERE  ( @SalesPersonId IS NULL 
          OR par.salespersonid = @SalesPersonId ) 
GROUP  BY pos.districtid, 
          pos.pointofsalesname, 
          pos.pointofsalesid, 
          par.salespersonid 
ORDER  BY pos.pointofsalesid, 
          pos.districtid 

另一个选择是移到下面的 CTE，但这主要是一种样式，除非在 SQL 代码中重复 Inline 视图。

WITH t 
     AS (SELECT SUM(soldamount) AS expr1, 
                pointofsalesid, 
                salespersonid 
         FROM   dbo.dailysales AS ds 
         WHERE  ( salesdate = CONVERT(VARCHAR, @salesDate, 102) ) 
         GROUP  BY pointofsalesid, 
                   salespersonid) 
SELECT DISTINCT par.salespersonid, 
                pos.pointofsalesid, 
                pos.pointofsalesname, 
                Isnull(t.expr1, 0) AS soldamount, 
FROM   dbo.pointofsales AS pos 
       INNER JOIN dbo.personanddistrict AS par 
         ON pos.districtid = par.districtid 
       LEFT JOIN t 
         ON t.salespersonid = par.salespersonid 
            AND t.pointofsalesid = pos.pointofsalesid 
WHERE  ( @SalesPersonId IS NULL 
          OR par.salespersonid = @SalesPersonId ) 
GROUP  BY pos.districtid, 
          pos.pointofsalesname, 
          pos.pointofsalesid, 
          par.salespersonid 
ORDER  BY pos.pointofsalesid, 
          pos.districtid