重庆思庄Oracle、Redhat认证学习论坛
标题: 如何看懂ORACLE的执行计划 [打印本页]
作者: mahan 时间: 2023-9-17 12:37
标题: 如何看懂ORACLE的执行计划
1、看懂Oracle执行计划
上面已经介绍了如何查看执行计划,现在简单介绍一下一些基本方法和相关理论知识
1.1 查看explain
找一条比较复杂的SQL:
SQL_ID 4qfq3t2ukm0y1, child number 0
-------------------------------------
SELECT /*+ monitor*/ A.USER_CODE, A.FULL_NAME, A.USER_PWD, C.UNIT_CODE,
C.UNIT_NAME FROM BASE_USER A LEFT JOIN (SELECT UR.USER_CODE,
UR.UNIT_CODE FROM APPR_USER_ROLE UR WHERE UR.USER_ROLE < 10) B ON
A.USER_CODE = B.USER_CODE LEFT JOIN LZCITY_APPROVE_UNIT_INFO C ON
B.UNIT_CODE = C.UNIT_CODE WHERE C.UNIT_CODE ='15803'
Plan hash value: 3288287052
------------------------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Starts | E-Rows | Cost (%CPU)| E-Time | A-Rows | A-Time | Buffers |
------------------------------------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | | 3 (100)| | 16 |00:00:00.01 | 38 |
| 1 | NESTED LOOPS | | 1 | 1 | 3 (0)| 00:00:01 | 16 |00:00:00.01 | 38 |
| 2 | NESTED LOOPS | | 1 | 1 | 3 (0)| 00:00:01 | 16 |00:00:00.01 | 22 |
| 3 | NESTED LOOPS | | 1 | 1 | 2 (0)| 00:00:01 | 16 |00:00:00.01 | 5 |
| 4 | TABLE ACCESS BY INDEX ROWID| LZCITY_APPROVE_UNIT_INFO | 1 | 1 | 1 (0)| 00:00:01 | 1 |00:00:00.01 | 3 |
|* 5 | INDEX UNIQUE SCAN | PK_LZCITY_APPROVE_UNIT_INFO | 1 | 1 | 0 (0)| | 1 |00:00:00.01 | 2 |
|* 6 | INDEX RANGE SCAN | PK_APPR_USER_ROLE | 1 | 1 | 1 (0)| 00:00:01 | 16 |00:00:00.01 | 2 |
|* 7 | INDEX UNIQUE SCAN | PK_BASE_USER | 16 | 1 | 0 (0)| | 16 |00:00:00.01 | 17 |
| 8 | TABLE ACCESS BY INDEX ROWID | BASE_USER | 16 | 1 | 1 (0)| 00:00:01 | 16 |00:00:00.01 | 16 |
------------------------------------------------------------------------------------------------------------------------------------------------
Query Block Name / Object Alias (identified by operation id):
-------------------------------------------------------------
1 - SEL$E3445A69
4 - SEL$E3445A69 / C@SEL$4
5 - SEL$E3445A69 / C@SEL$4
6 - SEL$E3445A69 / UR@SEL$2
7 - SEL$E3445A69 / A@SEL$3
8 - SEL$E3445A69 / A@SEL$3
Outline Data
-------------
/*+
BEGIN_OUTLINE_DATA
IGNORE_OPTIM_EMBEDDED_HINTS
OPTIMIZER_FEATURES_ENABLE('11.2.0.4')
DB_VERSION('11.2.0.4')
ALL_ROWS
OUTLINE_LEAF(@"SEL$E3445A69")
MERGE(@"SEL$2")
OUTLINE(@"SEL$A2E96217")
OUTER_JOIN_TO_INNER(@"SEL$E9F4A6F9" "B"@"SEL$1")
OUTER_JOIN_TO_INNER(@"SEL$E9F4A6F9" "C"@"SEL$4")
OUTLINE(@"SEL$2")
OUTLINE(@"SEL$E9F4A6F9")
MERGE(@"SEL$80808B20")
OUTLINE(@"SEL$6")
OUTLINE(@"SEL$80808B20")
MERGE(@"SEL$4")
MERGE(@"SEL$F1D6E378")
OUTLINE(@"SEL$5")
OUTLINE(@"SEL$4")
OUTLINE(@"SEL$F1D6E378")
MERGE(@"SEL$1")
OUTLINE(@"SEL$3")
OUTLINE(@"SEL$1")
INDEX_RS_ASC(@"SEL$E3445A69" "C"@"SEL$4" ("LZCITY_APPROVE_UNIT_INFO"."UNIT_CODE"))
INDEX(@"SEL$E3445A69" "UR"@"SEL$2" ("APPR_USER_ROLE"."UNIT_CODE" "APPR_USER_ROLE"."USER_CODE" "APPR_USER_ROLE"."AREA_SEQ"
"APPR_USER_ROLE"."USER_ROLE"))
INDEX(@"SEL$E3445A69" "A"@"SEL$3" ("BASE_USER"."USER_CODE"))
LEADING(@"SEL$E3445A69" "C"@"SEL$4" "UR"@"SEL$2" "A"@"SEL$3")
USE_NL(@"SEL$E3445A69" "UR"@"SEL$2")
USE_NL(@"SEL$E3445A69" "A"@"SEL$3")
NLJ_BATCHING(@"SEL$E3445A69" "A"@"SEL$3")
END_OUTLINE_DATA
*/
Predicate Information (identified by operation id):
---------------------------------------------------
5 - access("C"."UNIT_CODE"='15803')
6 - access("UR"."UNIT_CODE"='15803' AND "UR"."USER_ROLE"<10)
filter("UR"."USER_ROLE"<10)
7 - access("A"."USER_CODE"="UR"."USER_CODE")
1.2 explain执行顺序
所以不管是用F5方式还是set statistics_level=ALL方式,都有Operation参数,Operation表示sql执行过程,查看怎么执行的,有两个规则:
根据Operation缩进判断,缩进最多的最先执行;
Operation缩进相同时,最上面的是最先执行的;
如图执行计划,根据规则,可以得出执行顺序:INDEX UNIQUE SCAN->TABLE ACCESS BY INDEX ROWID->INDEX RANGE SCAN ->NESTED LOOPS ->INDEX UNIQUE SCAN->NESTED LOOPS ->TABLE ACCESS BY INDEX ROWID->NESTED LOOPS-> SELECT STATEMENT