如何使SQL性能得到改进？

Question

Oracle模式：

tableA 
AA varchar2 PK  
BB number  PK          
CC varchar2 PK  
DD number  PK  
EE number 

SQL查询：

SELECT 
    X1.AA, 
    X1.CC, 
    X1.DD, 
    tableA.EE 
FROM 
    tableA, 
    (SELECT 
     AA, 
     CC, 
     MAX(tableA.DD) DD 
    FROM 
     tableA 
    WHERE 
     (tableA.BB = 7 OR tableA.BB = 1) AND 
     tableA.EE <> 1 
    GROUP BY 
     AA, 
     CC 
    ORDER BY 
     AA ASC, 
     CC DESC 
    ) X1, 
    (SELECT 
     AA, 
     MAX(tableA.CC) CC 
    FROM 
     tableA 
    WHERE 
     (tableA.BB = 7 OR tableA.BB = 1) AND 
     tableA.EE <> 1 
    GROUP BY 
     AA 
    ORDER BY 
     AA ASC 
    ) X2 
WHERE 
    X2.AA = X1.AA AND 
    X2.CC = X1.CC AND 
    X2.AA = tableA.AA AND 
    X2.CC = tableA.CC AND 
    X1.DD = tableA.DD 
ORDER BY 
    AA ASC 

Answer 1

我明白X1和X2表用于仅MAX值滤波。在这种情况下，查询可以使用窗口功能被改写：

SELECT AA, CC, DD, EE 
FROM (
    SELECT * 
     , MAX(CC) OVER (PARTITION BY AA ) As maxCC 
     , MAX(DD) OVER (PARTITION BY AA, CC) As maxDD 
    FROM tableA 
    WHERE BB In (1, 7) AND EE <> 1 
) A 
WHERE CC = maxCC AND DD = maxDD; 

这将节省一对夫妇的子查询，每个自己的过滤加入。

更新：考虑到Gordon Linoff的评论这样可以更准确地被改写为：

SELECT AA, CC, DD, EE 
FROM tableA 
WHERE (AA, CC, DD) IN (
    SELECT DISTINCT AA 
     , MAX(CC) OVER (PARTITION BY AA ) As CC 
     , MAX(DD) OVER (PARTITION BY AA, CC) As DD 
    FROM tableA 
    WHERE BB In (1, 7) AND EE <> 1 
); 

Answer 2

这是基于查询的YB的解释，这是对赛道非常大概什么OP真的很想要。但是，它会过滤正在输出的行。相反，过滤器只需要移动以计算最大值：

SELECT AA, CC, DD, EE 
FROM (SELECT a.*, 
      MAX(CASE WHEN BB In (1, 7) AND EE <> 1 THEN CC END) OVER 
       (PARTITION BY AA) As maxCC, 
      MAX(CASE WHEN BB In (1, 7) AND EE <> 1 THEN DD END) OVER 
       (PARTITION BY AA, CC) As maxDD 
     FROM tableA a 
    ) a 
WHERE CC = maxCC AND DD = maxDD;