函数索引跟普通索引谁的效率高

环境:oracle 11g

在sqlplus中开启:

set linesize 500

set autotrace traceonly

set timing on

这里用Substr和 'xxx%'的查询效率做测试

drop table t purge;

create table t as

select * from dba_objects;

--实验1:返回多数据

--预计3313条记录

--加普通索引

create index t_object_type_idn on t(object_type);

select * from t where t.object_type like 'TAB%';

函数索引跟普通索引谁的效率高

--加函数索引

drop index t_object_type_idn;

create index t_object_type_idn_fun on t(substr(t.object_type,1,3));

select * from t where substr(t.object_type,1,3)='TAB';

发现:代价:函数索引小于普通索引

实验2:返回少数据100

drop table t purge;

create table t as

select * from dba_objects;

update t set t.object_type='XXXAAA' where rownum<=100;

--加普通索引

create index t_object_type_idn on t(object_type);

select * from t where t.object_type like 'XXX%';

--加函数索引

drop index t_object_type_idn;

create index t_object_type_idn_fun on t(substr(object_type,1,3));

select * from t where substr(t.object_type,1,3)='XXX';

发现:代价:函数索引大于普通索引

实验3:找到函数索引与普通索引性能差不多时的记录数;400

drop table t purge;

create table t as

select * from dba_objects;

update t set t.object_type='XXXAAA' where rownum<=400;

--加普通索引

create index t_object_type_idn on t(object_type);

select * from t where t.object_type like 'XXX%';

--加函数索引

drop index t_object_type_idn;

create index t_object_type_idn_fun on t(substr(object_type,1,3));

select * from t where substr(t.object_type,1,3)='XXX';

小结:发现平衡点在400左右;

当查询的结果数量小于400时,普通索引的性能高于函数索引;

当查询的结果数量大于400时,函数索引性能高于普通索引;

这里仅使用了dba_objectsz作为测试,实际的不同表临界值可能不一样;