[20160211]閉包傳遞3.txt

lfree發表於2016-02-11
[20160211]閉包傳遞3.txt

--所謂閉包傳遞是指sql語句的謂詞條件A=B and B=C 可以推出 A=C. oracle 的 最佳化器能夠利用這個特性最佳化sql語句。
--前一陣子看電子電子書<Apress.Expert.Oracle.SQL.Optimization.Deployment.and.Statistics.1430259779.pdf>測試連結:

http://blog.itpub.net/267265/viewspace-1981803/
http://blog.itpub.net/267265/viewspace-1987668/

--晚上以前學習oracle最佳化時,我練習過透過hint控制4個表的連線順序問題.自己重複測試看看,難道4個表就沒有問題嗎?

1.環境:
SCOTT@test01p> @ ver1

PORT_STRING          VERSION     BANNER                                                                       CON_ID
-------------------- ----------- ---------------------------------------------------------------------------- ------
IBMPC/WIN_NT64-9.1.0 12.1.0.1.0  Oracle Database 12c Enterprise Edition Release 12.1.0.1.0 - 64bit Production      0

create table t1 as select rownum   id,rownum||'t1' name from dual connect by level<=5;
create table t2 as select rownum+1 id,rownum||'t2' name from dual connect by level<=5;
create table t3 as select rownum+2 id,rownum||'t3' name from dual connect by level<=5;
create table t4 as select rownum+3 id,rownum||'t3' name from dual connect by level<=5;

-- 分析表.Method_Opt => 'FOR ALL COLUMNS SIZE 1 '

2.以前我記得我學習最佳化時,聽過一些最佳化的課程,裡面提到如何能很好的控制4個表的連線順序,能很好地解決一些問題.
  我當時覺得很簡單,實際上就是使用提示leading ,SWAP_JOIN_INPUTS 就很好的控制連線順序.舉一個例子:
  假設要使用t4->(t2,t3)->t1 這樣的連線順序,使用的提示是:
  leading (t2 t3 t4 t1) SWAP_JOIN_INPUTS(t4)

--測試看看:
select /*+ leading (t2 t3 t4 t1) SWAP_JOIN_INPUTS(t4)  */ * from t1,t2,t3,t4 where t1.id=t2.id  and t2.id=t3.id and t3.id=t4.id;

Plan hash value: 1146301945
--------------------------------------------------------------------------------------------------------------------------------------------------
| Id  | Operation            | Name | Starts | E-Rows |E-Bytes| Cost (%CPU)| E-Time   | A-Rows |   A-Time   | Buffers |  OMem |  1Mem | Used-Mem |
--------------------------------------------------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT     |      |      1 |        |       |    12 (100)|          |      2 |00:00:00.01 |      13 |       |       |          |
|*  1 |  HASH JOIN           |      |      1 |      5 |   140 |    12   (0)| 00:00:01 |      2 |00:00:00.01 |      13 |  1245K|  1245K|  557K (0)|
|*  2 |   HASH JOIN          |      |      1 |      5 |   105 |     9   (0)| 00:00:01 |      3 |00:00:00.01 |       9 |  1888K|  1888K| 1093K (0)|
|   3 |    TABLE ACCESS FULL | T4   |      1 |      5 |    35 |     3   (0)| 00:00:01 |      5 |00:00:00.01 |       3 |       |       |          |
|*  4 |    HASH JOIN         |      |      1 |      5 |    70 |     6   (0)| 00:00:01 |      4 |00:00:00.01 |       6 |  1888K|  1888K| 1064K (0)|
|   5 |     TABLE ACCESS FULL| T2   |      1 |      5 |    35 |     3   (0)| 00:00:01 |      5 |00:00:00.01 |       3 |       |       |          |
|   6 |     TABLE ACCESS FULL| T3   |      1 |      5 |    35 |     3   (0)| 00:00:01 |      5 |00:00:00.01 |       3 |       |       |          |
|   7 |   TABLE ACCESS FULL  | T1   |      1 |      5 |    35 |     3   (0)| 00:00:01 |      5 |00:00:00.01 |       4 |       |       |          |
--------------------------------------------------------------------------------------------------------------------------------------------------
Query Block Name / Object Alias (identified by operation id):
-------------------------------------------------------------
   1 - SEL$1
   3 - SEL$1 / T4@SEL$1
   5 - SEL$1 / T2@SEL$1
   6 - SEL$1 / T3@SEL$1
   7 - SEL$1 / T1@SEL$1
Predicate Information (identified by operation id):
---------------------------------------------------
   1 - access("T1"."ID"="T2"."ID")
   2 - access("T3"."ID"="T4"."ID")
   4 - access("T2"."ID"="T3"."ID")

3.當然我的問題要測試閉包問題,修改連線條件如下:
select /*+ leading (t2 t3 t4 t1) SWAP_JOIN_INPUTS(t4)  */ * from t1,t2,t3,t4 where t1.id=t2.id  and t2.id=t4.id and t3.id=t4.id;

--修改t2.id=t3.id為t2.id=t4.id,這樣實際上暗含有t2.id=t3.id條件的.看看執行計劃:

Plan hash value: 2268006538
----------------------------------------------------------------------------------------------------------------------------------------------------
| Id  | Operation              | Name | Starts | E-Rows |E-Bytes| Cost (%CPU)| E-Time   | A-Rows |   A-Time   | Buffers |  OMem |  1Mem | Used-Mem |
----------------------------------------------------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT       |      |      1 |        |       |    16 (100)|          |      2 |00:00:00.01 |      13 |       |       |          |
|*  1 |  HASH JOIN             |      |      1 |      5 |   140 |    16   (0)| 00:00:01 |      2 |00:00:00.01 |      13 |  1245K|  1245K|  525K (0)|
|*  2 |   HASH JOIN            |      |      1 |      5 |   105 |    13   (0)| 00:00:01 |      3 |00:00:00.01 |       9 |  1645K|  1645K| 1211K (0)|
|   3 |    TABLE ACCESS FULL   | T4   |      1 |      5 |    35 |     3   (0)| 00:00:01 |      5 |00:00:00.01 |       3 |       |       |          |
|   4 |    MERGE JOIN CARTESIAN|      |      1 |     25 |   350 |    10   (0)| 00:00:01 |     25 |00:00:00.01 |       6 |       |       |          |
|   5 |     TABLE ACCESS FULL  | T2   |      1 |      5 |    35 |     3   (0)| 00:00:01 |      5 |00:00:00.01 |       3 |       |       |          |
|   6 |     BUFFER SORT        |      |      5 |      5 |    35 |     7   (0)| 00:00:01 |     25 |00:00:00.01 |       3 |  2048 |  2048 | 2048  (0)|
|   7 |      TABLE ACCESS FULL | T3   |      1 |      5 |    35 |     1   (0)| 00:00:01 |      5 |00:00:00.01 |       3 |       |       |          |
|   8 |   TABLE ACCESS FULL    | T1   |      1 |      5 |    35 |     3   (0)| 00:00:01 |      5 |00:00:00.01 |       4 |       |       |          |
----------------------------------------------------------------------------------------------------------------------------------------------------
Query Block Name / Object Alias (identified by operation id):
-------------------------------------------------------------
   1 - SEL$1
   3 - SEL$1 / T4@SEL$1
   5 - SEL$1 / T2@SEL$1
   7 - SEL$1 / T3@SEL$1
   8 - SEL$1 / T1@SEL$1
Predicate Information (identified by operation id):
---------------------------------------------------
   1 - access("T1"."ID"="T2"."ID")
   2 - access("T2"."ID"="T4"."ID" AND "T3"."ID"="T4"."ID")

--注意看T2,T3的連線使用了CARTESIAN,看來是自己以前忽略了這個問題.

4.按照作者觀點,測試加入一個常量看看:

select /*+ leading (t2 t3 t4 t1) SWAP_JOIN_INPUTS(t4)  */ * from t1,t2,t3,t4 where t1.id=t2.id  and t2.id=t4.id and t3.id=t4.id and t1.id=4;

Plan hash value: 2268006538
----------------------------------------------------------------------------------------------------------------------------------------------------
| Id  | Operation              | Name | Starts | E-Rows |E-Bytes| Cost (%CPU)| E-Time   | A-Rows |   A-Time   | Buffers |  OMem |  1Mem | Used-Mem |
----------------------------------------------------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT       |      |      1 |        |       |    12 (100)|          |      1 |00:00:00.02 |      13 |       |       |          |
|*  1 |  HASH JOIN             |      |      1 |      1 |    28 |    12   (0)| 00:00:01 |      1 |00:00:00.02 |      13 |  1245K|  1245K|  697K (0)|
|*  2 |   HASH JOIN            |      |      1 |      1 |    21 |     9   (0)| 00:00:01 |      1 |00:00:00.01 |       9 |  1645K|  1645K|  731K (0)|
|*  3 |    TABLE ACCESS FULL   | T4   |      1 |      1 |     7 |     3   (0)| 00:00:01 |      1 |00:00:00.01 |       3 |       |       |          |
|   4 |    MERGE JOIN CARTESIAN|      |      1 |      1 |    14 |     6   (0)| 00:00:01 |      1 |00:00:00.01 |       6 |       |       |          |
|*  5 |     TABLE ACCESS FULL  | T2   |      1 |      1 |     7 |     3   (0)| 00:00:01 |      1 |00:00:00.01 |       3 |       |       |          |
|   6 |     BUFFER SORT        |      |      1 |      1 |     7 |     3   (0)| 00:00:01 |      1 |00:00:00.01 |       3 |  2048 |  2048 | 2048  (0)|
|*  7 |      TABLE ACCESS FULL | T3   |      1 |      1 |     7 |     3   (0)| 00:00:01 |      1 |00:00:00.01 |       3 |       |       |          |
|*  8 |   TABLE ACCESS FULL    | T1   |      1 |      1 |     7 |     3   (0)| 00:00:01 |      1 |00:00:00.01 |       4 |       |       |          |
----------------------------------------------------------------------------------------------------------------------------------------------------
Query Block Name / Object Alias (identified by operation id):
-------------------------------------------------------------
   1 - SEL$1
   3 - SEL$1 / T4@SEL$1
   5 - SEL$1 / T2@SEL$1
   7 - SEL$1 / T3@SEL$1
   8 - SEL$1 / T1@SEL$1
Predicate Information (identified by operation id):
---------------------------------------------------
   1 - access("T1"."ID"="T2"."ID")
   2 - access("T2"."ID"="T4"."ID" AND "T3"."ID"="T4"."ID")
   3 - filter("T4"."ID"=4)
   5 - filter("T2"."ID"=4)
   7 - filter("T3"."ID"=4)
   8 - filter("T1"."ID"=4)

-- 注意看裡面的filter條件.看來以後最佳化時要注意這個問題.

來自 “ ITPUB部落格 ” ,連結:http://blog.itpub.net/267265/viewspace-1987848/,如需轉載,請註明出處,否則將追究法律責任。

相關文章