表中将近有1100万数据，很多时候，我们要执行 字符串匹配，在SQL语句中，我们通常运用 like来达到我们搜索的目标。但经过实际测试发觉，like的效率与instr函数差别相当大。下面是一些测试结果：

　　SQL> set timing on

　　SQL> select count(*) from t where instr(title,’手册’)>0;

　　COUNT(*)

　　———-

　　65881

　　Elapsed: 00:00:11.04

　　SQL> select count(*) from t where title like ‘%手册%’;

　　COUNT(*)

　　———-

　　65881

　　Elapsed: 00:00:31.47

　　SQL> select count(*) from t where instr(title,’手册’)=0;

　　COUNT(*)

　　———-

　　11554580

　　Elapsed: 00:00:11.31

　　SQL> select count(*) from t where title not like ‘%手册%’;

　　COUNT(*)

　　———-

　　11554580

　　注:

　　instr(title,’手册’)>0 相当于like

　　instr(title,’手册’)=0 相当于not like

--------------------------------------------------------------------------------------------------------

SQL> create table test as select * from dba_objects;

表已创建。

SQL> select owner,object_name from test where owner like 'ROBINSON%';

OWNER                                                        OBJECT_NAME
------------------------------------------------------------ ------------------------------
ROBINSON                                                     TEST
ROBINSON                                                     T2
ROBINSON                                                     T1
ROBINSON                                                     T3
ROBINSON                                                     T4
ROBINSON                                                     HEX_TO_BIN

已选择6行。

SQL> create index i_owner on test(owner);

索引已创建。

SQL> update test set wner='SYSMAN' where owner!='ROBINSON';

已更新50353行。

SQL> update test set wner='SYSTEM' where object_id>1000 and object_id<1020;

已更新19行。

SQL> commit;

提交完成。

SQL> update test set wner='SCOTT' where object_id>2000 and object_id<2005;

已更新4行。

SQL> select distinct owner from test;

OWNER
------------------------------------------------------------
ROBINSON
SYSTEM
SCOTT
SYS

SQL> BEGIN
  2     DBMS_STATS.GATHER_TABLE_STATS(ownname => 'ROBINSON',
  3                                   tabname => 'TEST',
  4                                   estimate_percent => DBMS_STATS.AUTO_SAMPLE_SIZE,
  5                                   method_opt => 'for all columns size repeat',
  6                                   degree => DBMS_STATS.AUTO_DEGREE,
  7                                   cascade=>TRUE
  8                                   );
  9  END;
 10  /

PL/SQL 过程已成功完成。

SQL> set autot traceonly
SQL> select owner,object_name from test where owner not like 'SYS%';

已选择10行。

执行计划
----------------------------------------------------------
Plan hash value: 1357081020

--------------------------------------------------------------------------
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     |
--------------------------------------------------------------------------
|   0 | SELECT STATEMENT  |      | 25180 |   737K|   142   (3)| 00:00:02 |
|*  1 |  TABLE ACCESS FULL| TEST | 25180 |   737K|   142   (3)| 00:00:02 |
--------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

1 - filter("OWNER" NOT LIKE 'SYS%')

统计信息
----------------------------------------------------------
          0  recursive calls
          0  db block gets
          0  consistent gets
          0  physical reads
          0  redo size
          0  bytes sent via SQL*Net to client
          0  bytes received via SQL*Net from client
          0  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
         10  rows processed

从此处可以看到选择了10行，但是却走了全表扫描，现在对其加HINT提示

SQL> select /*+ INDEX (test I_OWNER) */ owner,object_name from test where owner not like 'SYS%';

已选择10行。

执行计划
----------------------------------------------------------
Plan hash value: 469605390

---------------------------------------------------------------------------------------
| Id  | Operation                   | Name    | Rows  | Bytes | Cost (%CPU)| Time     |
---------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT            |         | 25180 |   737K|   490   (1)| 00:00:06 |
|   1 |  TABLE ACCESS BY INDEX ROWID| TEST    | 25180 |   737K|   490   (1)| 00:00:06 |
|*  2 |   INDEX FULL SCAN           | I_OWNER | 25180 |       |   140   (3)| 00:00:02 |
---------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

2 - filter("OWNER" NOT LIKE 'SYS%')

统计信息
----------------------------------------------------------
          0  recursive calls
          0  db block gets
          0  consistent gets
          0  physical reads
          0  redo size
          0  bytes sent via SQL*Net to client
          0  bytes received via SQL*Net from client
          0  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
         10  rows processed

我们发现，强制加HINT提示所花的成本比全表扫描成本更高，这里要想优化这条SQL 必须改like

SQL> drop index i_owner;

索引已删除。

SQL>  create  index f_owner on test(instr(owner,'SYS',1,1)) ;

索引已创建。

SQL> select  owner,object_name from test where instr(owner,'SYS',1,1)=0;

已选择10行。

执行计划
----------------------------------------------------------
Plan hash value: 535182827

---------------------------------------------------------------------------------------
| Id  | Operation                   | Name    | Rows  | Bytes | Cost (%CPU)| Time     |
---------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT            |         |   504 | 15120 |    52   (0)| 00:00:01 |
|   1 |  TABLE ACCESS BY INDEX ROWID| TEST    |   504 | 15120 |    52   (0)| 00:00:01 |
|*  2 |   INDEX RANGE SCAN          | F_OWNER |   201 |       |    49   (0)| 00:00:01 |
---------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

2 - access(INSTR("OWNER",'SYS',1,1)=0)

统计信息
----------------------------------------------------------
          0  recursive calls
          0  db block gets
          0  consistent gets
          0  physical reads
          0  redo size
          0  bytes sent via SQL*Net to client
          0  bytes received via SQL*Net from client
          0  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
         10  rows processed

当使用函数索引后，COST由原来的142下降到现在的52

总结：对于LIKE语句，我们可以使用instr函数来进行SQL调优

本文来自CSDN博客，转载请标明出处：http://blog.csdn.net/robinson1988/archive/2010/03/23/5408253.aspx

来自 “ ITPUB博客 ” ，链接：http://blog.itpub.net/21916216/viewspace-630315/，如需转载，请注明出处，否则将追究法律责任。