如何通过explain和dbms_xplan包分析执行计划

这篇文章将为大家详细讲解有关如何通过explain和dbms_xplan包分析执行计划，文章内容质量较高，因此小编分享给大家做个参考，希望大家阅读完这篇文章后对相关知识有一定的了解。

该工具需要访问一张特殊的表plan_table，该表用于存储执行计划，在Oracle 10g之前需要用脚本utlxplan.sql创建：

建表：

@?\rdbms\admin\utlxplan.sql

建同义词：

create public synonym plan_table for plan_table;

授权：

grant all on plan_table to public;

Oracle 10g之后不再需要创建表plan_table，而是增加了数据字典表plan_table$，然后基于plan_table$创建了公共同义词供用户使用。

explain基本语法：

explain plan [set statement_id = stmt_id] for sql_statement;

explain指令的执行结果存储于表plan_table中，有几种方法获取执行计划的详细信息：

1、直接查询plan_table表

解释计划

explain plan for select count(*) from scott.emp;

查询结果

col id for 999

col operation for a50

col options for a20

col object_name for a20

select id,

lpad( , 2 * depth) || operation || || options ||

decode(id, 0, substr(optimizer, 1, 6) || Cost = || to_char(cost)) operation,

options,

object_name,

position

from plan_table

where plan_id = (select max(plan_id) from plan_table)

order by id;

ID OPERATION OPTIONS OBJECT_NAME POSITION

---- -------------------------------------------------- -------------------- -------------------- ----------

0 SELECT STATEMENT ALL_RO Cost = 1 1

1 SORT AGGREGATE AGGREGATE 1

2 INDEX FULL SCAN FULL SCAN PK_EMP 1

或者用以下查询，包含了执行计划树的level层次关系

col "Execute Plan" for a100

select id || || parent_id || || lpad( , 2 * level - 1) ||

operation || || options || || object_name || (Cost= || cost || ) as "Execute Plan"

from plan_table

start with id = 0

connect by prior id = parent_id;

Execute Plan

----------------------------------------------------------------------------------------------------

0 SELECT STATEMENT (Cost=1)

1 0 SORT AGGREGATE (Cost=)

2 1 INDEX FULL SCAN PK_EMP (Cost=1)

2、通过程序包dbms_xplan获得执行计划

1)获得最近一次explain的执行计划

col "PLAN_TABLE_OUTPUT" for a100

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

PLAN_TABLE_OUTPUT

----------------------------------------------------------------------------------------------------

Plan hash value: 2937609675

-------------------------------------------------------------------

| Id | Operation | Name | Rows | Cost (%CPU)| Time |

-------------------------------------------------------------------

| 0 | SELECT STATEMENT | | 1 | 1 (0)| 00:00:01 |

| 1 | SORT AGGREGATE | | 1 | | |

| 2 | INDEX FULL SCAN| PK_EMP | 14 | 1 (0)| 00:00:01 |

-------------------------------------------------------------------

2)通过指定的语句ID来查询

select * from table(dbms_xplan.display(plan_table, stmt_id));

3)通过SQL_ID和子游标来查询，该函数并不要求先做explain，显示的信息也较详细，另外还有一个format参数可以做更详细的定制。

select * from table(dbms_xplan.display_cursor(sql_id, child_number, format));

sql_id为null时显示最近一次执行的SQL的执行计划，但注意要保持set serveroutput off，否则最后一句SQL将不是你运行的SQL，child number为null，则返回所有子游标的执行计划。

可以通过在SQL语句中加入注释，方便的获取SQL_ID和CHILD_NUMBER信息，如以下SQL，先执行一次

select /*12345*/ m.material_id, m.material_no, m.part_no, m.material_name, m.validaterule, m.validaterule2

from cmes.c_material_t m

where (regexp_like(641234HNGA080001A, m.validaterule, c) or

regexp_like(641234HNGA080001A, m.validaterule2, c))

and m.deleted_flag = 0

and rownum = 1;

查出它的ID：

select sql_id, child_number from v$sql where sql_text like %12345%;

SQL_ID CHILD_NUMBER

------------- ------------

9jk2r7a64s470 0

cc274s1r7ab6w 0

因为以上包含"12345"注释的语句被执行了两条，所以因取先执行的第一条为实际的ID。

查看执行计划：

col PLAN_TABLE_OUTPUT for a100;

select * from table(dbms_xplan.display_cursor(9jk2r7a64s470, 0));

PLAN_TABLE_OUTPUT

----------------------------------------------------------------------------------------------------

SQL_ID 9jk2r7a64s470, child number 0

-------------------------------------

select /*12345*/ m.material_id, m.material_no, m.part_no, m.material_name, m.validaterule, m.validaterule2

from cmes.c_material_t m

where (regexp_like(641234HNGA080001A, m.validaterule, c) or

regexp_like(641234HNGA080001A, m.validaterule2, c))

and m.deleted_flag = 0

and rownum = 1;

Plan hash value: 1524529232

-----------------------------------------------------------------------------------

| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |

-----------------------------------------------------------------------------------

| 0 | SELECT STATEMENT | | | | 2 (100)| |

|* 1 | COUNT STOPKEY | | | | | |

|* 2 | TABLE ACCESS FULL| C_MATERIAL_T | 2 | 116 | 2 (0)| 00:00:01 |

-----------------------------------------------------------------------------------

Predicate Information (identified by operation id):

---------------------------------------------------

1 - filter(ROWNUM=1)

2 - filter(("M"."DELETED_FLAG"=0 AND ( REGEXP_LIKE

(641234HNGA080001A,"M"."VALIDATERULE",c,<not feasible>)

查看更详细的执行计划信息，需要先设置统计级别为ALL，否则没有A-ROWS等信息

alter session set statistics_level = all;

或者在语句级别使用HINT

select /*+gather_plan_statistics*/

select m.material_id, m.material_no, m.part_no, m.material_name, m.validaterule, m.validaterule2

from cmes.c_material_t m

where (regexp_like(641234HNGA080001A, m.validaterule, c) or

regexp_like(641234HNGA080001A, m.validaterule2, c))

and m.deleted_flag = 0

and rownum = 1;

select * from table(dbms_xplan.display_cursor(null, null, ALLSTATS));

PLAN_TABLE_OUTPUT

---------------------------------------------------------------------------------------------------

SQL_ID fmu73t3umxm1r, child number 0

-------------------------------------

select m.material_id, m.material_no, m.part_no, m.material_name, m.validaterule,

m.validaterule2 from cmes.c_material_t m where (regexp_like(641234HNGA080001A,

m.validaterule, c) or regexp_like(641234HNGA080001A, m.validaterule2,

c)) and m.deleted_flag = 0 and rownum = 1

Plan hash value: 1524529232

---------------------------------------------------------------------------------------------

| Id | Operation | Name | Starts | E-Rows | A-Rows | A-Time | Buffers |

---------------------------------------------------------------------------------------------

|* 1 | COUNT STOPKEY | | 1 | | 0 |00:00:00.01 | 16 |

|* 2 | TABLE ACCESS FULL| C_MATERIAL_T | 1 | 2 | 0 |00:00:00.01 | 16 |

---------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):

---------------------------------------------------

1 - filter(ROWNUM=1)

2 - filter((( REGEXP_LIKE (641234HNGA080001A,"M"."VALIDATERULE",c,-)

这里E-Rows是预估的返回行数，A-Rows是实际的返回行数。

format的高级应用可以显示更多的信息，特别是可以显示绑定变量的具体值，这个非常有用。以下是推荐的使用格式

select * from table(dbms_xplan.display_cursor(null,null,ADVANCED ALLSTATS LAST PEEKED_BINDS));

allstats：iostats + memstats，iostats显示该游标累计执行的io统计信息(buffers, reads)，memstats显示累计执行的pga使用信息(omem 1mem used-mem)

last：仅显示最后一次执行的统计信息

advanced：显示outline、query block name、column projection等信息

peeked_binds：打印解析时使用的绑定变量

如以下操作可以显示非常详细的执行计划和绑定变量等信息

alter session set statistics_level = all;

var a varchar2(20);

exec :a := EMP;

select object_id from dba_objects where object_name = :a;

select * from table(dbms_xplan.display_cursor(null,null,ADVANCED ALLSTATS LAST PEEKED_BINDS));

dbms_xplan程序包还有一个函数display_awr可以获取AWR报告中指定SQL_ID的执行计划

col PLAN_TABLE_OUTPUT for a100;

select * from table(dbms_xplan.display_awr(9jk2r7a64s470));

该函数获取的执行计划来自dba_hist_sql_plan视图，通过历史数据记录，甚至一些被老化的SQL执行计划仍然可以被查询到。

3、通过脚本utlxpls.sql或utlxplp.sql获得执行计划

@?\rdbms\admin\utlxpls.sql

或

@?\rdbms\admin\utlxplp.sql

PLAN_TABLE_OUTPUT

----------------------------------------------------------------------------------------------------

Plan hash value: 2937609675

-------------------------------------------------------------------

| Id | Operation | Name | Rows | Cost (%CPU)| Time |

-------------------------------------------------------------------

| 0 | SELECT STATEMENT | | 1 | 1 (0)| 00:00:01 |

| 1 | SORT AGGREGATE | | 1 | | |

| 2 | INDEX FULL SCAN| PK_EMP | 14 | 1 (0)| 00:00:01 |

-------------------------------------------------------------------

看一下这两个脚本的内容，该方法实际是调用了方法2中的程序包

get ?\rdbms\admin\utlxpls.sql

......

41* select plan_table_output from table(dbms_xplan.display(plan_table,null,serial));

42

get ?\rdbms\admin\utlxplp.sql

......

40* select * from table(dbms_xplan.display());

关于如何通过explain和dbms_xplan包分析执行计划就分享到这里了，希望以上内容可以对大家有一定的帮助，可以学到更多知识。如果觉得文章不错，可以把它分享出去让更多的人看到。