Decode360's Blog

业精于勤而荒于嬉 QQ:150355677 MSN:decode360@hotmail.com

  BlogJava :: 首页 :: 新随笔 :: 联系 ::  :: 管理 ::
  302 随笔 :: 26 文章 :: 82 评论 :: 0 Trackbacks
    Oracle可以将Update和Insert操作合并为一个MERGE操作。第一可以更加方便操作、简化逻辑,第二在Oracle的内部进行了效率的优化。
 
    在下面的举例可以看出,如果使用PLSQL中的exception来判断是UPDATE还是INSERT,效率是及其低下的。而在SQL层级使用MERGE或者手动分两步进行INSERT|UPDATE,效率相差无几。使用MERGE的优势是逻辑简单,一般无特使要求均以使用MERGE函数为佳。而使用两步的优点是可以返回有多少记录是UPDATE多少记录是INSERT的,还可以精确反映在哪一步发生错误。所以后两种方法各有优势,第一种方法是需要极力避免的。
 
 
SQL> create table t1 as select object_id, object_name from all_objects;
 
Table created.
 
SQL> alter table t1 add constraint t1_pk primary key(object_id);
 
Table altered.
 
SQL> create table t2 as select * from t1;
 
Table created.
 
SQL> alter table t2 add constraint t2_pk primary key(object_id);
 
Table altered.
 
SQL>
SQL> analyze table t1 compute statistics
  2  for table
  3  for all indexes
  4  for all indexed columns;
 
Table analyzed.
 
SQL>
SQL> analyze table t2 compute statistics
  2  for table
  3  for all indexes
  4  for all indexed columns;
 
Table analyzed.
 
SQL>
SQL> set timing on
SQL> declare
  2          l_icnt number := 0;
  3          l_ucnt number := 0;
  4  begin
  5      for x in ( select * from t1 )
  6      loop
  7          begin
  8              insert into t2 ( object_id, object_name ) values ( x.object_id, x.object_name );
  9                          l_icnt := l_icnt+1;
 10          exception
 11              when dup_val_on_index then
 12                  update t2 set object_name = x.object_name where object_id = x.object_id;
 13                          l_ucnt := l_ucnt+1;
 14          end;
 15      end loop;
 16          dbms_output.put_line( 'Inserted ' || l_icnt );
 17          dbms_output.put_line( 'Updated  ' || l_ucnt );
 18  end;
 19  /
Inserted 0
Updated  29317
 
PL/SQL procedure successfully completed.
 
Elapsed: 00:01:04.07
SQL> rollback;
 
Rollback complete.
 
Elapsed: 00:00:00.86
SQL>
SQL> begin
  2      update ( select t1.object_name t1_oname, t2.object_name t2_oname
  3                 from t1, t2
  4                where t1.object_id = t2.object_id )
  5         set t1_oname = t2_oname;
  6
  7          dbms_output.put_line( 'Updated  ' || sql%rowcount );
  8
  9         insert into t1
 10      select * from t2 where t2.object_id not in ( select object_id from t1 );
 11
 12          dbms_output.put_line( 'Inserted ' || sql%rowcount );
 13  end;
 14  /
Updated  29317
Inserted 0
 
PL/SQL procedure successfully completed.
 
Elapsed: 00:00:02.44
SQL> rollback;
 
Rollback complete.
 
Elapsed: 00:00:01.07
SQL>
SQL> merge into t1
  2  using t2
  3  on ( t2.object_id = t1.object_id )
  4  when matched then
  5     update set t1.object_name = t2.object_name
  6  when not matched then
  7      insert (object_id, object_name) values( t2.object_id, t2.object_name);
 
29317 rows merged.
 
Elapsed: 00:00:02.87
SQL>
SQL> rollback;
 
Rollback complete.
 
Elapsed: 00:00:01.08
 
 
    另外,Tom还建议:除非是特殊情况,否则不要使用分段提交。因为这样会破坏事务的一致性,而且会影响整体的性能。
    如果一定要使用批量提交(例如undo表空间太小且不可增加时),加入以下代码即可(只能用cursor来update)
 
begin
    ...
for xc1 in c loop
    commit_point:=commit_point+1;
    if commit_point > 10000 then
    commit;
    commit_point:=0;
    end if;
    begin
        ...
    end;
end loop;
    ...
end;
 
 
 
 




-The End-

posted on 2009-04-01 21:23 decode360-3 阅读(256) 评论(0)  编辑  收藏 所属分类: SQL Dev

只有注册用户登录后才能发表评论。


网站导航: