问题:请教HINT写法
我有一个SQL添加如下hint,目的是指定hash_join方式。
select /*+ordered use_hash(a,b,c,d) */ *
From a,b,c,d
Where ...
其中,
a只与b有关联关系,b只与c有关联关系,b只与c有关联关系,c只与d有关联关系,
数量级:a:1000条, b:100 万条, c:800万条 , d:100万条
执行计划为:
Hash Join
---Hash Join
----- Hash Join
------ a
----- b
----- c
---d
考虑到d表比较小,我能不能做到将d表作为驱动表、而a,b,c关联之后的结果作为prob外表呢,
通过Ordered好像是没有办法控制这样,加了就只能是a作为驱动表装载内存,b作为prob表与之关联,
之后的结果再作为驱动表,以此类推。
用Leading可以吗?请给出语法,谢谢。
解答:
oracle 10g中
hash_join可以通过no_swap_join_inputs/swap_join_inputs来强制控制build表,配合leading或者ordered可以控制多表之前的连接顺序。
比如t1,t2,t3,t4共4张表做hash_join
可以通过ordered+no_swap_join_inputs/swap_join_inputs来实现。
比如
如果想实现
( T3 hash-join (T1 hash-join T2)) hash-join T4
t1作为build表和T2做hash_join,然后t3作为build表和t1,t2的结果集作hash_join,在把t3,t1,t2的结果集作build表和t4做hash_join
通过sql可以写为
MYDB@MYDB10G >select
2 /*+
3 ordered
4 use_hash(t2)
5 use_hash(t3)
6 swap_join_inputs(t3)
7 use_hash(t4)
8 no_swap_join_inputs(t4)
9 */
10 * from t1,t2,t3,t4
11 where t1.object_id=t2.object_id
12 and t2.object_name=t3.object_name
13 and t3.owner=t4.owner
14 and t4.owner='MYDB'
15 /
已用时间: 00: 00: 00.07
执行计划
----------------------------------------------------------
Plan hash value: 3494725078
-------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes |TempSpc| Cost (%CPU)| Time |
-------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 2137 | 801K| | 182 (2)| 00:00:03 |
|* 1 | HASH JOIN | | 2137 | 801K| | 182 (2)| 00:00:03 |
|* 2 | HASH JOIN | | 52 | 14976 | | 167 (2)| 00:00:03 |
|* 3 | TABLE ACCESS FULL | T3 | 40 | 3840 | | 15 (0)| 00:00:01 |
|* 4 | HASH JOIN | | 11651 | 2184K| 1232K| 151 (1)| 00:00:02 |
| 5 | TABLE ACCESS FULL| T1 | 11651 | 1092K| | 15 (0)| 00:00:01 |
| 6 | TABLE ACCESS FULL| T2 | 11652 | 1092K| | 15 (0)| 00:00:01 |
|* 7 | TABLE ACCESS FULL | T4 | 41 | 3936 | | 15 (0)| 00:00:01 |
-------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - access("T3"."OWNER"="T4"."OWNER")
2 - access("T2"."OBJECT_NAME"="T3"."OBJECT_NAME")
3 - filter("T3"."OWNER"='MYDB')
4 - access("T1"."OBJECT_ID"="T2"."OBJECT_ID")
7 - filter("T4"."OWNER"='MYDB')
Note
-----
- dynamic sampling used for this statement
说明:
ordered表示依据from后面写的表的顺序来做联结~
写hints,分开些思路清晰~ ordered后 from t1 ,t2 ,t3 ,t4说明首先使用t1做驱动表来连接t2,如何连接呢?看后面的hint use_hash(t2)
代表连接t2的方式是hash_join;然后用use_hash(t3)表示连接t3的方式是hash-join,那么谁作build表呢?看后面的swap_join_inputs(t3)代表t3作build表和t1-t2的结果集作连接....依此类推~
标准的hint就应该这么写~ use_hash(x,y,z)这样的写法是不规范,这样只是说出了x,y,z地联接方式
ordered 是陈旧的hints,leading是用来代替ordered的~ leading不要求sql的写法(from后面的顺序不要求),直接可以在leading中定义连接顺序~
leading和ordered不能一起使用,也没必要一起使用~
至于"能不能举个Ordered不能实现而要用Leading的例子",上面已经说了,from后面的不能被修改的情况下可以使用leading.用法google一下,一大堆文档
针对leading使用:
10g中对leading做了加强~ 可以直接在后面写多表的连接顺序了,也就是说使用leading不需要from后面的固定顺序了
MYDB@MYDB10G >select
2 /*+
3 leading(t1 t2 t3 t4)
4 use_hash(t2)
5 use_hash(t3)
6 swap_join_inputs(t3)
7 use_hash(t4)
8 no_swap_join_inputs(t4)
9 */ * from t3,t4,t2,t1
10 where t1.object_id=t2.object_id
11 and t2.object_name=t3.object_name
12 and t3.owner=t4.owner
13 and t4.owner='MYDB'
14 /
已用时间: 00: 00: 00.01
执行计划
----------------------------------------------------------
Plan hash value: 3494725078
-----------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-----------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 2069 | 179K| 57 (4)| 00:00:01 |
|* 1 | HASH JOIN | | 2069 | 179K| 57 (4)| 00:00:01 |
|* 2 | HASH JOIN | | 50 | 3100 | 38 (3)| 00:00:01 |
|* 3 | TABLE ACCESS FULL | T3 | 40 | 1080 | 18 (0)| 00:00:01 |
|* 4 | HASH JOIN | | 82 | 2870 | 20 (5)| 00:00:01 |
| 5 | TABLE ACCESS FULL| T1 | 11584 | 248K| 16 (0)| 00:00:01 |
| 6 | TABLE ACCESS FULL| T2 | 82 | 1066 | 3 (0)| 00:00:01 |
|* 7 | TABLE ACCESS FULL | T4 | 41 | 1107 | 18 (0)| 00:00:01 |
-----------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - access("T3"."OWNER"="T4"."OWNER")
2 - access("T2"."OBJECT_NAME"="T3"."OBJECT_NAME")
3 - filter("T3"."OWNER"='MYDB')
4 - access("T1"."OBJECT_ID"="T2"."OBJECT_ID")
7 - filter("T4"."OWNER"='MYDB')
MYDB@MYDB10G >select
2 /*+
3 ordered
4 use_hash(t2)
5 use_hash(t3)
6 swap_join_inputs(t3)
7 use_hash(t4)
8 no_swap_join_inputs(t4)
9 */ * from t1,t2,t3,t4
10 where t1.object_id=t2.object_id
11 and t2.object_name=t3.object_name
12 and t3.owner=t4.owner
13 and t4.owner='MYDB'
14 /
已用时间: 00: 00: 00.00
执行计划
----------------------------------------------------------
Plan hash value: 3494725078
-----------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-----------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 2069 | 179K| 57 (4)| 00:00:01 |
|* 1 | HASH JOIN | | 2069 | 179K| 57 (4)| 00:00:01 |
|* 2 | HASH JOIN | | 50 | 3100 | 38 (3)| 00:00:01 |
|* 3 | TABLE ACCESS FULL | T3 | 40 | 1080 | 18 (0)| 00:00:01 |
|* 4 | HASH JOIN | | 82 | 2870 | 20 (5)| 00:00:01 |
| 5 | TABLE ACCESS FULL| T1 | 11584 | 248K| 16 (0)| 00:00:01 |
| 6 | TABLE ACCESS FULL| T2 | 82 | 1066 | 3 (0)| 00:00:01 |
|* 7 | TABLE ACCESS FULL | T4 | 41 | 1107 | 18 (0)| 00:00:01 |
-----------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - access("T3"."OWNER"="T4"."OWNER")
2 - access("T2"."OBJECT_NAME"="T3"."OBJECT_NAME")
3 - filter("T3"."OWNER"='MYDB')
4 - access("T1"."OBJECT_ID"="T2"."OBJECT_ID")
7 - filter("T4"."OWNER"='MYDB')