如何在 Oracle 中编写长 group by/case 子句？

Question

我经常有一些查询要按 case 子句进行分组，例如：

SELECT
    CASE WHEN FIELD IS NULL                     |
    AND (SOMETHING > 100 OR SOMETHINGELSE < 50) |
    AND THATDATE >= add_months(sysdate, -4)     | -> Case
        THEN 0 ELSE 1                           |                                                               
    END AS dim_1,                               |                                                               
    dim_2,
    sum(SOMEFIELD_TOTAL) as measure_1
FROM MY_TABLE
GROUP BY
    CASE WHEN FIELD IS NULL                     | 
    AND (SOMETHING > 100 OR SOMETHINGELSE < 50) | 
    AND THATDATE >= add_months(sysdate, -4)     | -> Same case
        THEN 0 ELSE 1                           |                                                                
    END,                                        |                                                                
    dim_2

有没有办法可以将我为整个案例定义的别名用于group by，例如group by dim_1？

找到this explanation，还是想知道有没有出路。

Answer 1

我认为您过于担心重复表达。 Oracle 和任何其他 DBMS 都会将该表达式识别为重复项并仅计算一次。如果您只是不想重复它（这只是 剪切和粘贴 琐事提醒您），您可以使用派生表或公用表表达式。

SELECT 
    dim_1, dim_2, sum(SOMEFIELD_TOTAL) as measure_1
FROM (
    SELECT
        CASE WHEN FIELD IS NULL                     |
        AND (SOMETHING > 100 OR SOMETHINGELSE < 50) |
        AND THATDATE >= add_months(sysdate, -4)     | -> Case
            THEN 0 ELSE 1                           |                                                               
        END AS dim_1,                               |                                                               
        dim_2,
        SOMEFIELD_TOTAL
    FROM MY_TABLE
) AS SQ
GROUP BY
    dim_1, dim_2

看起来并没有好多少，是吗？

Answer 2

您可以使用subquery factoring 也称为named subquery 和common table expression (CTE)（感谢@a_horse_with_no_name）

WITH t as 
(
SELECT
    CASE WHEN FIELD IS NULL                     
    AND (SOMETHING > 100 OR SOMETHINGELSE < 50) 
    AND THATDATE >= add_months(sysdate, -4)     
        THEN 0 ELSE 1 END AS dim_1,         
    dim_2,
    SOMEFIELD_TOTAL as measure_1
FROM MY_TABLE
)
SELECT dim_1, dim_2, sum(SOMEFIELD_TOTAL)
FROM t
GROUP BY dim_1, dim_2

这种方法的优点是你可以有很多子查询，而不必像在常规子查询中那样加入它们（内部选择）