zoukankan      html  css  js  c++  java
  • 转://WITH AS and materialize hints

    WITH AS: 就是将一个子查询部分独立出来,有时候是为了提高SQL语句的可读性,有时候是为了提高SQL语句性能。
    如果一个SQL语句中,某个表会被访问多次,而且每次访问的限制条件一样的话,就可以使用with as来提高性能。
    注意:如果 with as 短语没有被调用2次以上,CBO就不会讲这个短语获取的数据放入temp表,如果想要讲数据放入temp表需要使用materialize hint
    如果 with as 短语被调用了2次以上,CBO会自动将 with as 短语的数据放入一个临时表,这个时候不用写materialize hint

    举个例子(本例基于Scott用户)
    SQL> explain plan for
    with a as (select /*+ materialize */ ename,job,deptno from emp where sal>(select avg(sal) from emp))
    select * from a ;
    2 3
    Explained.

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

    PLAN_TABLE_OUTPUT
    -------------------------------------------------------------------------------------------------------
    Plan hash value: 2006423466

    -------------------------------------------------------------------------------------------------------
    | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
    -------------------------------------------------------------------------------------------------------
    | 0 | SELECT STATEMENT | | 1 | 26 | 8 (0)| 00:00:01 |
    | 1 | TEMP TABLE TRANSFORMATION | | | | | |
    | 2 | LOAD AS SELECT | SYS_TEMP_0FD9D6605_E16CE | | | | |
    |* 3 | TABLE ACCESS FULL | EMP | 1 | 21 | 3 (0)| 00:00:01 |
    | 4 | SORT AGGREGATE | | 1 | 4 | | |
    | 5 | TABLE ACCESS FULL | EMP | 14 | 56 | 3 (0)| 00:00:01 |
    | 6 | VIEW | | 1 | 26 | 2 (0)| 00:00:01 |
    | 7 | TABLE ACCESS FULL | SYS_TEMP_0FD9D6605_E16CE | 1 | 17 | 2 (0)| 00:00:01 |
    -------------------------------------------------------------------------------------------------------

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

    3 - filter("SAL"> (SELECT AVG("SAL") FROM "EMP" "EMP"))

    19 rows selected.

    去掉 /*+ materialize */ ,由于只访问了一次a,所以CBO不会将a的查询结果生成一个临时表

    SQL> explain plan for
    with a as (select ename,job,deptno from emp where sal>(select avg(sal) from emp))
    select * from a ; 2 3

    Explained.

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

    PLAN_TABLE_OUTPUT
    ----------------------------------------------------------------------------
    Plan hash value: 1876299339

    ----------------------------------------------------------------------------
    | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
    ----------------------------------------------------------------------------
    | 0 | SELECT STATEMENT | | 1 | 21 | 6 (0)| 00:00:01 |
    |* 1 | TABLE ACCESS FULL | EMP | 1 | 21 | 3 (0)| 00:00:01 |
    | 2 | SORT AGGREGATE | | 1 | 4 | | |
    | 3 | TABLE ACCESS FULL| EMP | 14 | 56 | 3 (0)| 00:00:01 |
    ----------------------------------------------------------------------------

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

    1 - filter("SAL"> (SELECT AVG("SAL") FROM "EMP" "EMP"))

    15 rows selected.

    WITH AS 语句调用一次 使用多次 需要写hints

    如果 表 只 扫描 1次,你些materialize hints 结果读了一次 还写入temp, 再从temp读出来
    临时表写入是1次,但是读要多次。

    继续测试:
    SQL> explain plan for
    with a as (select ename,job,deptno from emp where sal>(select avg(sal) from emp))
    select * from a union all select * from a; 2 3

    Explained.

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

    PLAN_TABLE_OUTPUT
    ------------------------------------------------------------------------------------------------------------------------------------------------------------

    --------------------------------------------
    Plan hash value: 2575088720

    --------------------------------------------------------------------------------------------------------
    | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
    --------------------------------------------------------------------------------------------------------
    | 0 | SELECT STATEMENT | | 2 | 52 | 4 (50)| 00:00:01 |
    | 1 | TEMP TABLE TRANSFORMATION | | | | | |
    | 2 | LOAD AS SELECT | SYS_TEMP_0FD9D6601_4DC46A | | | | |
    |* 3 | TABLE ACCESS FULL | EMP | 1 | 39 | 3 (0)| 00:00:01 |
    | 4 | SORT AGGREGATE | | 1 | 13 | | |
    | 5 | TABLE ACCESS FULL | EMP | 14 | 182 | 3 (0)| 00:00:01 |
    | 6 | UNION-ALL | | | | | |
    | 7 | VIEW | | 1 | 26 | 2 (0)| 00:00:01 |
    | 8 | TABLE ACCESS FULL | SYS_TEMP_0FD9D6601_4DC46A | 1 | 26 | 2 (0)| 00:00:01 |
    | 9 | VIEW | | 1 | 26 | 2 (0)| 00:00:01 |
    | 10 | TABLE ACCESS FULL | SYS_TEMP_0FD9D6601_4DC46A | 1 | 26 | 2 (0)| 00:00:01 |
    --------------------------------------------------------------------------------------------------------

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

    3 - filter("SAL"> (SELECT AVG("SAL") FROM "EMP" "EMP"))

    Note
    -----
    - dynamic sampling used for this statement (level=2)
    26 rows selected.

    充分证明 :
    1.当with as 语句没有被调用2次以上时,如果表需要访问多次,那么需要加hints /*+ materialize */
    2.如果with as 语句被调用2次以上时,自动会将 with as 短语的数据放入一个临时表,这个时候不用写materialize hint

  • 相关阅读:
    项目设计之---------- 模版模式利用
    项目设计之一------简单工厂模式利用
    项目设计之----命令模式的利用
    项目设计之一---------- 代码重构
    项目设计之---------- 类的设计原则
    项目设计之一-------------项目包的设计原则(原)
    模式经验/理解
    Java常量定义需要注意事项及static作用(复习)
    H2数据库使用 详解
    开源数据库 H2, HSQLDB, DERBY, PostgreSQL, MySQL区别/对比图表( 附加翻译) h2数据库
  • 原文地址:https://www.cnblogs.com/zfox2017/p/7878991.html
Copyright © 2011-2022 走看看