将 MERGE 语句转换为 UPDATE 语句

Question

MERGE INTO TABLE1 t1
USING TABLE2 t2
ON (t1.ID = t2.ID) 
WHEN MATCHED THEN UPDATE 
  SET t1.PHONE_NUMBER = CASE 
                          WHEN t1.type in ('A','B') THEN  t2.phone_number 
                          ELSE NVL(t2.phone_number, t1.phone_number)
                        END

需要将上面的MERGE 转换成UPDATE 语句。

我尝试了以下方法：

UPDATE TABLE1 t1 
   SET t1.PHONE_NUMBER = (
                          SELECT t2.PHONE_NUMBER 
                            FROM TABLE2 t2, TABLE1 t3 
                           WHERE t3.type in ('A','B') 
                             AND t3.ID = t2.ID
                         )

如何在上述实现中加入CASE？

Answer 1

MERGE 语句将更有效率，您可能应该使用它；但是，您可以使用：

UPDATE table1 t1
SET PHONE_NUMBER = NVL(
                     ( SELECT t2.phone_number
                       FROM   table2 t2
                       WHERE  t1.id = t2.id ),
                     CASE
                     WHEN t1.type IN ( 'A', 'B' )
                     THEN NULL
                     ELSE t1.phone_number
                     END
                   )
WHERE EXISTS ( SELECT 1 FROM table2 t2 WHERE t1.id = t2.id );

Oracle 设置：

CREATE TABLE table1 ( id, phone_number, type ) AS
SELECT 1, 123456, 'A' FROM DUAL UNION ALL
SELECT 2, 123456, 'B' FROM DUAL UNION ALL
SELECT 3, 123456, 'C' FROM DUAL UNION ALL
SELECT 4, 123456, 'D' FROM DUAL UNION ALL
SELECT 5, 123456, 'E' FROM DUAL;

CREATE TABLE table2 ( id, phone_number ) AS
SELECT 1, 234567 FROM DUAL UNION ALL
SELECT 2, NULL   FROM DUAL UNION ALL
SELECT 3, 345678 FROM DUAL UNION ALL
SELECT 4, NULL   FROM DUAL;

输出：

运行更新后：

SELECT * FROM table1

输出与MERGE 语句相同：

身份证 |电话号码 |类型 -: | ------------: | :--- 1 | 234567 |一种 2 | 空 |乙 3 | 345678 | C 4 | 123456 | D 5 | 123456 |乙

db小提琴here

---

更新：

如果您查看EXPLAIN PLAN 中的MERGE 语句：

|计划表输出 | | :------------------------------------------------ ------------------------------ | |计划哈希值：3423411568 | | | | -------------------------------------------------- ------------------------------ | | |身份证 |操作 |姓名 |行 |字节 |成本 (%CPU)|时间 | | | -------------------------------------------------- ------------------------------ | | | 0 |合并声明 | | 4 | 168 | 7 (15)| 00:00:01 | | | | 1 |合并 |表 1 | | | | | | | | 2 |查看 | | | | | | | | |* 3 |哈希连接 | | 4 | 268 | 7 (15)| 00:00:01 | | | | 4 |表访问已满|表 2 | 4 | 104 | 3 (0)| 00:00:01 | | | | 5 |表访问已满|表 1 | 5 | 205 | 3 (0)| 00:00:01 | | | -------------------------------------------------- ----------------------------->

那么它只读取TABLE1 和TABLE2 各一次。

将其与EXPLAIN PLAN 的UPDATE 语句进行比较：

|计划表输出 | | :------------------------------------------------ ---------------------------------------- | |计划哈希值：735598124 | | | | -------------------------------------------------- ---------------------------------------- | | |身份证 |操作 |姓名 |行 |字节 |成本 (%CPU)|时间 | | | -------------------------------------------------- ---------------------------------------- | | | 0 |更新声明 | | 4 | 168 | 23 (22)| 00:00:01 | | | | 1 |更新 |表 1 | | | | | | | |* 2 |哈希连接半 | | 4 | 168 | 7 (15)| 00:00:01 | | | | 3 |表访问已满|表 1 | 5 | 145 | 3 (0)| 00:00:01 | | | | 4 |表访问已满|表 2 | 4 | 52 | 3 (0)| 00:00:01 | | | |* 5 |表访问完全 |表 2 | 1 | 26 | 3 (0)| 00:00:01 | | | -------------------------------------------------- ---------------------------------------- |

然后它将从TABLE1 读取一次，TABLE2 读取两次；所以MERGE 可能是一个性能更高的查询......但如果你愿意，你可以用UPDATE 来做到这一点。

db小提琴here

Answer 2

您可以使用带有RIGHT JOIN 的嵌套子查询：

update table1 t 
   set t.phone_number = (
                          select tt.phone_number
                            from( 
                                 select case 
                                        when t1.type in ('A','B') then  t2.phone_number 
                                        else nvl(t2.phone_number, t1.phone_number)
                                        end as phone_number, 
                                        nvl(t2.ID,t1.ID) as ID
                                   from table2 t2
                                  right join table1 t1 
                                     on t1.ID = nvl(t2.ID,t1.ID)
                                 ) tt 
                            where tt.ID = t.ID
                           );

Demo

谢谢你的小提琴@MTO