数据库调优：如何优化in操作

今天开发人员让我优化一个sql：

Select Item.*
From Item Item
Where Item.Item_Id in (15056, 14697);

　　这个item表很大，sql执行起来很慢。

Execution Plan
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=12795 Card=2 Bytes=1
316)

1 0 TABLE ACCESS (FULL) OF 'ITEM' (Cost=12795 Card=2 Bytes=131
6)



Statistics
----------------------------------------------------------
0 recursive calls
0 db block gets
134234 consistent gets
26568 physical reads
2184 redo size
3336 bytes sent via SQL*Net to client
503 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
2 rows processed

　　在这个例子中，item_id是item表的主键。这里慢的原因是in操作不能用到索引，做全表扫描，但为什么不能用索引呢？

　　实际上，oracle在处理in的时候是转换成or进行处理的，如：

　　a in (1,2)相当于 a=1 or a=2

　　而对于b*tree索引来说，对于or操作是不会用到索引的。知道原因，下面就要找出解决方法了。

　　方法一、从b*tree索引着手

　　这种方法就是想办法让查询可以走b*tree索引。要走btree索引，就要去掉or操作，所以把查询改成一个等价的sql：

SQL> Select Item.* From Item Item Where Item.Item_Id in (14697)
2 union all
3 Select Item.* From Item Item Where Item.Item_Id in (15056);


Execution Plan
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=6 Card=2 Bytes=1316)
1 0 UNION-ALL
2 1 TABLE ACCESS (BY INDEX ROWID) OF 'ITEM' (Cost=3 Card=1 B
ytes=658)

3 2 INDEX (UNIQUE SCAN) OF 'PK_ITEM' (UNIQUE) (Cost=2 Card
=1)

4 1 TABLE ACCESS (BY INDEX ROWID) OF 'ITEM' (Cost=3 Card=1 B
ytes=658)

5 4 INDEX (UNIQUE SCAN) OF 'PK_ITEM' (UNIQUE) (Cost=2 Card
=1)


Statistics
----------------------------------------------------------
0 recursive calls
0 db block gets
8 consistent gets
0 physical reads
0 redo size
3334 bytes sent via SQL*Net to client
503 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
2 rows processed

　　从输出结果可以看出，sql已经可以有索引了，效率也得到大幅提高。但这种办法有缺陷：
如果in list的个数不确定，且如果in list较多的话，拼sql都是一个问题。我这个应用就是这种情况。

　　方法二、从位图方式着手

　　我们知道，位图索引是可以被or操作使用的，我们可以根据这一特点先用btree索引查询数据，然后进行位图转换、联合。
oracle给我们提供了一个hint来解决这个问题：use_concat。原sql修改为：

SQL> Select /**//*+ use_concat */ 
Item.* From Item Item Where Item.Item_Id in (15056,14697);


Execution Plan
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=6 Card=2 Bytes=1316)
1 0 TABLE ACCESS (BY INDEX ROWID) OF 'ITEM' (Cost=6 Card=2 Byt
es=1316)

2 1 BITMAP CONVERSION (TO ROWIDS)
3 2 BITMAP OR
4 3 BITMAP CONVERSION (FROM ROWIDS)
5 4 INDEX (RANGE SCAN) OF 'PK_ITEM' (UNIQUE) (Cost=2)
6 3 BITMAP CONVERSION (FROM ROWIDS)
7 6 INDEX (RANGE SCAN) OF 'PK_ITEM' (UNIQUE) (Cost=2)

Statistics
----------------------------------------------------------
0 recursive calls
0 db block gets
8 consistent gets
0 physical reads
0 redo size
3336 bytes sent via SQL*Net to client
503 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
2 rows processed

　　从结果可以看到，oracle先通过主键索引分别对两个id进行查询，然后转换成位图进行合并，这样效率就得到提供，对原sql的改动也很小，基本满足了要求。

　　本文所有结果都是基于 as 3+ oracle9204平台。

　　这个转换受_b_tree_bitmap_plans决定。如果该参数为true，则可进行转换，否则不进行转换。而8i里，该参数缺省为false，所以只有存在bitmap索引时才会考虑将B树索引转换为bitmap; 而到了9i里，该参数缺省为true了，则对任何索引都有可能进行bitmap转换。

　　顺便提一下，今天在itpub有帖子讲到in能否用到btree索引问题，我这里总结一下：
　　1、如果in list是具体的term，如id in (1,1,2)之类的话，是不会用到b*tree的
　　2、如果in list是有一个子查询得到的，如id in (select id from t2)，那么这个查询实际两表关联。那么能否用到id列的索引要受到很多因素的制约了。

【责编:Peng】

--------------------next---------------------

Execution Plan
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=12795 Card=2 Bytes=1
316)

1 0 TABLE ACCESS (FULL) OF 'ITEM' (Cost=12795 Card=2 Bytes=131
6)



Statistics
----------------------------------------------------------
0 recursive calls
0 db block gets
134234 consistent gets
26568 physical reads
2184 redo size
3336 bytes sent via SQL*Net to client
503 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
2 rows processed

--------------------next---------------------

SQL> Select Item.* From Item Item Where Item.Item_Id in (14697)
2 union all
3 Select Item.* From Item Item Where Item.Item_Id in (15056);


Execution Plan
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=6 Card=2 Bytes=1316)
1 0 UNION-ALL
2 1 TABLE ACCESS (BY INDEX ROWID) OF 'ITEM' (Cost=3 Card=1 B
ytes=658)

3 2 INDEX (UNIQUE SCAN) OF 'PK_ITEM' (UNIQUE) (Cost=2 Card
=1)

4 1 TABLE ACCESS (BY INDEX ROWID) OF 'ITEM' (Cost=3 Card=1 B
ytes=658)

5 4 INDEX (UNIQUE SCAN) OF 'PK_ITEM' (UNIQUE) (Cost=2 Card
=1)


Statistics
----------------------------------------------------------
0 recursive calls
0 db block gets
8 consistent gets
0 physical reads
0 redo size
3334 bytes sent via SQL*Net to client
503 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
2 rows processed

--------------------next---------------------

SQL> Select /**//*+ use_concat */ 
Item.* From Item Item Where Item.Item_Id in (15056,14697);


Execution Plan
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=6 Card=2 Bytes=1316)
1 0 TABLE ACCESS (BY INDEX ROWID) OF 'ITEM' (Cost=6 Card=2 Byt
es=1316)

2 1 BITMAP CONVERSION (TO ROWIDS)
3 2 BITMAP OR
4 3 BITMAP CONVERSION (FROM ROWIDS)
5 4 INDEX (RANGE SCAN) OF 'PK_ITEM' (UNIQUE) (Cost=2)
6 3 BITMAP CONVERSION (FROM ROWIDS)
7 6 INDEX (RANGE SCAN) OF 'PK_ITEM' (UNIQUE) (Cost=2)

Statistics
----------------------------------------------------------
0 recursive calls
0 db block gets
8 consistent gets
0 physical reads
0 redo size
3336 bytes sent via SQL*Net to client
503 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
2 rows processed

--------------------next---------------------