以通配符（%）開始的like字符串，走索引

在對oracle的SQL優化過程中經常會遇到【like'%abc'】破壞索引的問題，但是如果真有此類需求，該如何在不破壞索引的基礎上進行查詢呢。

[sql] view plain copy
sys@mescp> select reverse('123') from dual;  
  
REVERSE('123')  
--------------------------------  
321  
1 row selected.  
  
sys@mescp> create table test_like as select object_id,object_name from dba_objects;  
  
Table created.  
  
sys@mescp> create index test_like__name on test_like(object_name);  
  
Index created.  
  
sys@mescp> create index test_like__name_reverse on test_like(reverse(object_name));  
  
Index created.  
  
sys@mescp> analyze table test_like compute statistics for table for all indexes;  
  
Table analyzed.  
  
sys@mescp> set autotrace trace exp  
  
-- '常量開頭的like , 會利用index ，沒問題…… '  
  
sys@mescp> select * from test_like where object_name like AS%';  
  
Execution Plan  
  
----------------------------------------------------------  
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=2 Card=655 Bytes=15720)  
1 0 TABLE ACCESS (BY INDEX ROWID) OF 'TEST_LIKE' (Cost=2 Card=655Bytes=15720)  
2 1 INDEX (RANGE SCAN) OF 'TEST_LIKE__NAME' (NON-UNIQUE) (Cost=2 Card=118)  
--'開頭和結尾都是 % ,對不起，很難優化'  
sys@mescp> select * from test_like where object_name like '%%';  
  
Execution Plan  
  
----------------------------------------------------------  
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=6 Card=655 Bytes=15720)  
1 0 TABLE ACCESS (FULL) OF 'TEST_LIKE' (Cost=6 Card=655 ytes=15720)  
  
-- '以常量結束，直接寫的時候是不能應用index的'  
sys@mescp> select * from test_like where object_name like '%S';  
Execution Plan  
----------------------------------------------------------  
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=6 Card=655 Bytes=15720)  
  
1 0 TABLE ACCESS (FULL) OF 'TEST_LIKE' (Cost=6 Card=655 Bytes=15720)  
--'以常量結束的，加個reverse 函數，又可以用上index了'  
  
sys@mescp> select * from test_like where reverse(object_name)like reverse('%AS');  
Execution Plan  
  
----------------------------------------------------------  
  
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=2 Card=655 Bytes=15720)  
1 0 TABLE ACCESS (BY INDEX ROWID) OF 'TEST_LIKE' (Cost=2 Card=655 Bytes=15720)  
  
2 1 INDEX (RANGE SCAN) OF 'TEST_LIKE__NAME_REVERSE' (NON-UNIQUE) (Cost=2 Card=118)  

了解SQL的同學，都知道，like關鍵字可以走索引，只要字符串不是以通配符（%）開始。

如果類似 like "%xxx" 的sql，如何走索引呢？基於REVERSE()函數來創建一個函數索引。

 

1、准備數據：

 

CREATE TABLE jka AS SELECT ROWNUM id, dbms_random.string('x',10) v FROM dual CONNECT BY LEVEL <= 10000;

2、創建原始索引：

 

 

CREATE INDEX jka_normal ON jka (v);

BEGIN
  dbms_stats.gather_table_stats
    (USER
    ,'JKA'
    ,estimate_percent=>100
    ,cascade=>TRUE);
END;

3、以下SQL基於jka_normal索引走范圍掃描：

 

 

SELECT * FROM jka WHERE v LIKE 'ABC%';

4、但是下面的SQL將走全表掃描（不會使用索引）：

 

 

SELECT * FROM jka WHERE v LIKE '%ABC';

5、現在，創建一個函數索引（不要與反向索引[REVERSE INDEX]混淆）：

 

 

CREATE INDEX jka_reverse ON jka(REVERSE(v));

BEGIN
  dbms_stats.gather_table_stats
    (USER
    ,'JKA'
    ,estimate_percent=>100
    ,cascade=>TRUE);
END;

6、下面的SQL將基於jka_reverse索引走范圍掃描：

 

 

SELECT * FROM jka WHERE REVERSE(v) LIKE REVERSE('%ABC');

問題解決。