具有多列的oracle汇总函数

Question

我有一个简单的查询：

WITH data(val1, val2, val3) AS
     ( SELECT 'a' ,'a-details' ,'1' FROM DUAL
     UNION ALL
     SELECT 'b' ,'b-details' ,'2' FROM DUAL
     UNION ALL
     SELECT 'c' ,'c-details' ,'3' FROM DUAL
     )
SELECT NVL(val1,'Total Result'),
     val2,
     SUM(val3) tot
from data
group by rollup(val1, val2);

我得到的输出如下：

VAL1                             VAL2                                    TOT
-------------------------------- -------------------------------- ----------
a                                a-details                                 1 
a                                                                          1 
b                                b-details                                 2 
b                                                                          2 
c                                c-details                                 3 
c                                                                          3 
Total Result                                                               6 

但我需要一个输出：

VAL1                             VAL2                                    TOT
-------------------------------- -------------------------------- ----------
a                                a-details                                 1 
b                                b-details                                 2 
c                                c-details                                 3 
Total Result                                                               6 

提前致谢。

Answer 1

GROUPING_ID表达式

您可以使用GROUPING_ID表达式来过滤所需的小计级别：

WITH data AS
     ( SELECT 'a' AS val1 ,'a-details' AS val2 , '1' AS val3 FROM DUAL
     UNION ALL
     SELECT 'b' ,'b-details' ,'2' FROM DUAL
     UNION ALL
     SELECT 'c' ,'c-details' ,'3' FROM DUAL
     )
SELECT NVL(val1,'Total Result'),
     val2,
     SUM(val3) tot
from data
group by ROLLUP(val1, val2)
HAVING GROUPING_ID(val1, val2) IN (0, 3);

输出：

NVL(VAL1,'TOTALRESULT') VAL2             TOT
----------------------- --------- ----------
a                       a-details          1 
b                       b-details          2 
c                       c-details          3 
Total Result                               6 

GROUPING_ID为没有小计的行返回0，为第一级返回1，依此类推，我们可以看一下它返回的值：

WITH data AS
     ( SELECT 'a' AS val1 ,'a-details' AS val2 , '1' AS val3 FROM DUAL
     UNION ALL
     SELECT 'b' ,'b-details' ,'2' FROM DUAL
     UNION ALL
     SELECT 'c' ,'c-details' ,'3' FROM DUAL
     )
SELECT NVL(val1,'Total Result'),
     val2,
     SUM(val3) tot,
     GROUPING_ID(val1, val2) AS grp_id
from data
group by ROLLUP(val1, val2);

NVL(VAL1,'TOTALRESULT') VAL2             TOT     GRP_ID
----------------------- --------- ---------- ----------
a                       a-details          1          0 
a                                          1          1 
b                       b-details          2          0 
b                                          2          1 
c                       c-details          3          0 
c                                          3          1 
Total Result                               6          3 

Check at SQLFiddle

有关汇总及相关主题的更多信息：Tim Hall about Rollup and Cube

（修改）

GROUPING函数

关于评论。您可以使用GROUPING功能：

  

GROUPING - 接受一个列作为参数，如果是，则返回“1”   该列包含由a作为小计的一部分生成的空值   ROLLUP或CUBE操作或“0”表示任何其他值，包括   存储空值。

返回值的示例：

WITH data AS
     ( SELECT 'a' AS val1 ,'a-details' AS val2 , '1' AS val3 FROM DUAL
     UNION ALL
     SELECT 'b' ,'b-details' ,'2' FROM DUAL
     UNION ALL
     SELECT 'c' ,'c-details' ,'3' FROM DUAL
     )
SELECT NVL(val1,'Total Result'),
     val2,
     SUM(val3) tot,
     grouping(val1),
     grouping(val2)
from data
group by ROLLUP(val1, val2);

输出：

NVL(VAL1,'TOTALRESULT') VAL2             TOT GROUPING(VAL1) GROUPING(VAL2)
----------------------- --------- ---------- -------------- --------------
a                       a-details          1              0              0 
a                                          1              0              1 
b                       b-details          2              0              0 
b                                          2              0              1 
c                       c-details          3              0              0 
c                                          3              0              1 
Total Result                               6              1              1

所以你的查询应该是这样的：

WITH data AS
     ( SELECT 'a' AS val1 ,'a-details' AS val2 , '1' AS val3 FROM DUAL
     UNION ALL
     SELECT 'b' ,'b-details' ,'2' FROM DUAL
     UNION ALL
     SELECT 'c' ,'c-details' ,'3' FROM DUAL
     )
SELECT NVL(val1,'Total Result'),
     val2,
     SUM(val3) tot
from data
group by ROLLUP(val1, val2)
HAVING GROUPING(val1) = 1
   OR (GROUPING(val1) + GROUPING(val2) = 0);

输出：

NVL(VAL1,'TOTALRESULT') VAL2             TOT
----------------------- --------- ----------
a                       a-details          1 
b                       b-details          2 
c                       c-details          3 
Total Result                               6

使用AskTom的GROUPING函数的想法，here。

Answer 2

我发现使用GROUPING SET子句指定所需的确切集合相当容易：

WITH data(val1, val2, val3) AS
     ( SELECT 'a' ,'a-details' ,'1' FROM DUAL
     UNION ALL
     SELECT 'b' ,'b-details' ,'2' FROM DUAL
     UNION ALL
     SELECT 'c' ,'c-details' ,'3' FROM DUAL
     )
SELECT NVL(val1,'Total Result'),
     val2,
     SUM(val3) tot
from data
group by grouping sets ((val1, val2),());

我怀疑它更有效率，因为它直接指定要计算的水平。

http://sqlfiddle.com/#!4/8301d/3

CUBE和ROLLUP可以方便地自动生成大量聚合级别（例如，维度层次结构中的每个级别），如果您想从大型级别中消除一小部分级别，则可能存在使用GROUPING ID的情况CUBE生成的集合，但GROUPING SET精确设计用于指定特定的聚合级别。