Question

我正在尝试使用下面的查询在一个包含现有表的表上运行更新。代码列记录需要更新，但只需更新其中一些。我的查询似乎工作，但任何应该留在代码列中的记录最终都被设置为NULL。如何修改此查询以保持这些记录不变？

查询

update t1 x
set x.code =
(select code from
(select distinct address, city, prov, aflag, rcode from t2) y
where x.address = y.address and x.city = y.city and x.state = y.state and x.flag = y.flag)
where x.aflag like '%b%';

表1：要更新的代码

   t1               
address city    state   flag    code
123      aaa      il     b       400
567      bbb      il     b       400
345      bbb      il     b      -500
789      ddd      il     b       600
546      ccc      il     b       700

表2：用于更新T1的代码列

t2              
address city    state   flag    code
   123   aaa      il      b     -555
   444   bbb      il      b     -555
   345   bbb      il      b     -555
   888   kkk      il      b     -555
   546   ccc      il      b     -555



What the query currently outputs

     current output             
    address city    state   flag    code
       123   aaa      il      b     400
       444   bbb      il      b     NULL
       345   bbb      il      b     -500
       888   kkk      il      b     NULL
       546   ccc      il      b     -700

我希望查询保持记录不受更新表的匹配，如下所示

  What I want               
address city    state   flag    code
  123    aaa      il      b     400
  444    bbb      il      b    -555
  345    bbb      il      b    -500
  888    kkk      il      b    -555
  546    ccc      il      b    -700

谢谢！

Answer 1

最好的方法是在exists子句中使用where：

update t1 x
    set x.code = (select code
                  from t2 y
                  where x.address = y.address and x.city = y.city and x.state = y.state and x.flag = y.flag and rownum = 1
                 )
    where x.aflag like '%b%' and
          exists (select code
                  from t2 y
                  where x.address = y.address and x.city = y.city and x.state = y.state and x.flag = y.flag and rownum = 1
                 );

Answer 2

t2都会告诉您要更新哪些行以及使用哪个值。为此，请编写一个可更新的查询：

update
(
  select t1.code, t2.code as new_code
  from t1
  join t2 on  t2.address = t1.address
          and t2.city    = t1.city city 
          and t2.state   = t1.state
          and t2.flag    = t1.flag
  where t1.flag like '%b%'
)
set code = new_code;

为了实现这一点，address, city, state, flag中的t2必须是唯一的，因此DBMS知道它正好使用更新数据获得一条记录。

对于您想要更新多个列的情况，可更新查询尤其有用。

Answer 3

有些人喜欢它，有些人不喜欢。它看起来与您原来的更新声明非常相似：

MERGE INTO t1 x
USING (SELECT DISTINCT address, city, state, flag, code from t2) y
   ON (x.address = y.address AND x.city = y.city AND x.state = y.state AND x.flag = y.flag)
 WHEN MATCHED THEN 
      UPDATE SET x.code = y.code
       WHERE x.flag LIKE '%b%';

如何防止update语句将某些记录设置为null？

3 个答案: