子查询非嵌套(subquery unnesting) FILTER

SQL> set linesize 200
SQL> set pagesize 200
SQL> explain plan for select * from test1 where owner='SCOTT' or object_id in (select object_id from test2 where owner='SCOTT');

已解释。

SQL> select * from table(dbms_xplan.display());

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Plan hash value: 1896454807

----------------------------------------------------------------------------
| Id  | Operation	   | Name  | Rows  | Bytes | Cost (%CPU)| Time	   |
----------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |	   |  6171 |  1247K|   348   (1)| 00:00:05 |
|*  1 |  FILTER 	   |	   |	   |	   |		|	   |
|   2 |   TABLE ACCESS FULL| TEST1 |   103K|	20M|   348   (1)| 00:00:05 |
|*  3 |   TABLE ACCESS FULL| TEST2 |	 1 |	30 |   347   (1)| 00:00:05 |
----------------------------------------------------------------------------

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

   1 - filter("OWNER"='SCOTT' OR  EXISTS (SELECT 0 FROM "TEST2" "TEST2"
	      WHERE "OBJECT_ID"=:B1 AND "OWNER"='SCOTT'))
   3 - filter("OBJECT_ID"=:B1 AND "OWNER"='SCOTT')

Note
-----
   - dynamic sampling used for this statement (level=2)

已选择21行。


SQL> select count(*) from test1;

  COUNT(*)
----------
      1006

SQL> select count(*) from test2;

  COUNT(*)
----------
     87007


SQL> SELECT * FROM TABLE(DBMS_XPLAN.DISPLAY_CURSOR(NULL,NULL,'ALLSTATS LAST'));

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL_ID	a23r1uchdaafg, child number 0
-------------------------------------
select * from test1 where owner='SCOTT' or object_id in(select
object_id from test2 where owner='SCOTT')

Plan hash value: 1896454807

--------------------------------------------------------------------------------------
| Id  | Operation	   | Name  | Starts | E-Rows | A-Rows |   A-Time   | Buffers |
--------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |	   |	  1 |	     |	    6 |00:00:05.35 |	1247K|
|*  1 |  FILTER 	   |	   |	  1 |	     |	    6 |00:00:05.35 |	1247K|
|   2 |   TABLE ACCESS FULL| TEST1 |	  1 |	1000 |	 1006 |00:00:00.01 |	  24 |
|*  3 |   TABLE ACCESS FULL| TEST2 |   1000 |	   1 |	    0 |00:00:05.35 |	1247K|
--------------------------------------------------------------------------------------

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

   1 - filter(("OWNER"='SCOTT' OR  IS NOT NULL))
   3 - filter(("OBJECT_ID"=:B1 AND "OWNER"='SCOTT'))

Note
-----
   - dynamic sampling used for this statement (level=2)


已选择26行。


08-24
2013-08-24:12:04:21					  VALID 		N   N	N	     1


已选择6行。


注意看 A-Rows，Id=2 返回1006条记录 ，
然后 Id=3 执行了1000 次 最后返回7条数据1006-1000=6 
那么现在知道了 filter其实就相当于 NESTED LOOPS ，但是又不完全是 NESTED LOOPS
这里我们把离FILTER进的表也叫做驱动表，远的表叫被驱动表。
如果驱动表返回大量数据(比如100W)，最后FILTER过滤剩下10W条记录，那么被驱动表就会扫描100W-10W=90W次，这个时候不慢才怪了，逻辑读也就上去了。
大家在看执行计划的时候，看到FILTER就要引起注意了，绝大多数性能问题都是它引发的，所以我为啥说我培训是教你一眼看出性能问题。
注意：看到FILTER不一定它就慢，有时候还非得用FILTER不可，它跑得更快，后面有案例。