oraclemerge误区

1、 如果在d1表中有一个字段有多个，并且在使用merge是使用该字段关联会产生ORA-30926的错误，如在上一节面我们使用相关联的字段是deptno，如果在d1中出现多行deptno的值一样就会报错。

--构造实验环境（延续上一节的d1、d2表）
SQL> update d1 set deptno = 10 where deptno = 30;
1 row updated.

SQL> select * from d1;
    DEPTNO DNAME          LOC
---------- -------------- -------------
        10 ACCOUNTING     NEW YORK
        20 RESEARCH       DALLAS
        10 SALES          CHICAGO
        40 OPERATIONS     BOSTON 

2、 执行merge操作，使用d1表的记录去更新d2表的记录，查看结果

MERGE INTO d2
USING d1
ON (d1.deptno = d2.deptno)
WHEN MATCHED THEN
        UPDATE SET d2.loc = d1.loc || '...';
		
USING d1
      *
ERROR at line 2:
ORA-30926: unable to get a stable set of rows in the source tables 

从上面得到的错误：没有稳定值。可以得出在 ON() 子句中的条件是d1.deptno = 10不具备有唯一性，所以建议在关联的列上创建主键或者创建unique index。还有一个解决办法就是将值相等的行合并成一行来处理（请注意：这样做可能会改变需求）。

3、 如果将d1和d2倒过来（使用d2的记录去更新d2的记录）就不会发生ORA-30926，原因是在d2中的deptno的值是唯一。

MERGE INTO d1
USING d2
ON (d1.deptno = d2.deptno)
WHEN MATCHED THEN
        UPDATE set d1.loc = d2.loc || '...'
		
SQL> select * from d1;
    DEPTNO DNAME          LOC
---------- -------------- -------------
        10 ACCOUNTING     FU JIAN...
        20 RESEARCH       DALLAS
        10 SALES          FU JIAN...
        40 OPERATIONS     BOSTON 

4、 delete子句的where必须在最后

--环境
SQL> select * from d1;

    DEPTNO DNAME          LOC
---------- -------------- -------------
        10 ACCOUNTING     NEW YORK
        20 RESEARCH       DALLAS
        30 SALES          CHICAGO
        40 OPERATIONS     BOSTON

SQL> select * from d2;

    DEPTNO DNAME          LOC
---------- -------------- -------------
        10 ACCOUNTING     FU JIAN
        30 SALES          CHICAGO
		
--错误示例
MERGE INTO d2
USING d1
ON (d1.deptno = d2.deptno)
WHEN MATCHED THEN
        UPDATE SET d2.loc = d1.loc || '...'
        DELETE WHERE (d2.deptno = 10)
        WHERE d1.deptno = 10;
		
        WHERE d1.deptno = 10
        *
ERROR at line 7:
ORA-00933: SQL command not properly ended

--正确示例
MERGE INTO d2
USING d1
ON (d1.deptno = d2.deptno)
WHEN MATCHED THEN
        UPDATE SET d2.loc = d1.loc || '...'
        WHERE d1.deptno = 10
        DELETE WHERE (d2.deptno = 10); 

5、在delete 子句只可以伤处目标表，而不能删除源表，以上实验中我们用的都是delete where (d1.deptno)。

SQL> select * from d1;

    DEPTNO DNAME          LOC
---------- -------------- -------------
        10 ACCOUNTING     NEW YORK
        20 RESEARCH       DALLAS
        30 SALES          CHICAGO
        40 OPERATIONS     BOSTON

SQL> select * from d2;

    DEPTNO DNAME          LOC
---------- -------------- -------------
        10 ACCOUNTING     FU JIAN
        30 SALES          CHICAGO

--DELETE WHERE (d1.deptno = 10)
MERGE INTO d2
USING d1
ON (d1.deptno = d2.deptno)
WHEN MATCHED THEN
        UPDATE SET d2.loc = d1.loc || '...'
        DELETE WHERE (d1.deptno = 10)		
		
SQL> select * from d2;

    DEPTNO DNAME          LOC
---------- -------------- -------------
        30 SALES          CHICAGO...		

SQL> rollback;
Rollback complete.		

--使用DELETE WHERE (d2.deptno = 10)
MERGE INTO d2
USING d1
ON (d1.deptno = d2.deptno)
WHEN MATCHED THEN
        UPDATE SET d2.loc = d1.loc || '...'
        DELETE WHERE (d2.deptno = 10)		
		
SQL> select * from d2;
    DEPTNO DNAME          LOC
---------- -------------- -------------
        30 SALES          CHICAGO... 

6、 merge当引用表中没数据的解决办法

SQL> select * from d1;
no rows selected

SQL> select * from d2;
    DEPTNO DNAME          LOC
---------- -------------- -------------
        10 ACCOUNTING     FU JIAN
        30 SALES          CHICAGO
		
MERGE INTO d2
USING d1
ON (d1.deptno = d2.deptno)
WHEN MATCHED THEN
        UPDATE SET d2.loc = 'AAAAA'
WHEN NOT MATCHED THEN
        INSERT VALUES(20, 'aaaaa', 'AAAAA')

0 rows merged.

SQL> select * from d2;
    DEPTNO DNAME          LOC
---------- -------------- -------------
        10 ACCOUNTING     FU JIAN
        30 SALES          CHICAGO 

按道理说：当d1中没有找到数据时应该会执行INSERT语句的。但是，从上可以看到并没有执行INSERT语句。我们可以这样取巧的使用让他实行INSERT语句：

MERGE INTO d2
USING (SELECT COUNT(*) CNT FROM d1) d
ON (d.cnt   0)
WHEN MATCHED THEN
        UPDATE SET d2.loc = 'AAAAA'
WHEN NOT MATCHED THEN
        INSERT VALUES(20, 'aaaaa', 'AAAAA')
		
SQL> select * from d2;
    DEPTNO DNAME          LOC
---------- -------------- -------------
        10 ACCOUNTING     FU JIAN
        30 SALES          CHICAGO
        20 aaaaa          AAAAA 

查看更多关于oraclemerge误区的详细内容...