Conrad Fix 建议的子查询方法是常规方法。然而,这不太可能是最有效的方法,因为它需要 Oracle 访问表两次——一次计算平均年龄,一次撤回具有高于平均工资的行。如果您使用分析函数,您可以完成同样的事情,同时只需访问一次表并执行(大约)一半的逻辑 I/O 操作。
select *
from (select s.*, avg(age) over () avg_age
from student s)
where age > avg_age
传统方法需要 18 次一致的获取,并且必须对表进行两次完整扫描(请注意,我运行了几次测试以获得最低值以排除延迟块清除等事情)
SQL> ed
Wrote file afiedt.buf
1 select *
2 from hr.employees
3 where salary > (select avg(salary)
4* from hr.employees)
SQL> /
51 rows selected.
Execution Plan
----------------------------------------------------------
Plan hash value: 1945967906
---------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
---------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 5 | 345 | 6 (0)| 00:00:01 |
|* 1 | TABLE ACCESS FULL | EMPLOYEES | 5 | 345 | 3 (0)| 00:00:01 |
| 2 | SORT AGGREGATE | | 1 | 4 | | |
| 3 | TABLE ACCESS FULL| EMPLOYEES | 107 | 428 | 3 (0)| 00:00:01 |
---------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter("SALARY"> (SELECT AVG("SALARY") FROM "HR"."EMPLOYEES"
"EMPLOYEES"))
Statistics
----------------------------------------------------------
0 recursive calls
0 db block gets
18 consistent gets
0 physical reads
0 redo size
5532 bytes sent via SQL*Net to client
557 bytes received via SQL*Net from client
5 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
51 rows processed
然而,分析函数方法在一次表扫描中做同样的事情,只有 7 个一致的 get
SQL> select *
2 from (select e.*, avg(salary) over () avg_salary
3 from hr.employees e)
4 where salary > avg_salary
5 /
51 rows selected.
Execution Plan
----------------------------------------------------------
Plan hash value: 48081388
---------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
---------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 107 | 15622 | 3 (0)| 00:00:01 |
|* 1 | VIEW | | 107 | 15622 | 3 (0)| 00:00:01 |
| 2 | WINDOW BUFFER | | 107 | 7383 | 3 (0)| 00:00:01 |
| 3 | TABLE ACCESS FULL| EMPLOYEES | 107 | 7383 | 3 (0)| 00:00:01 |
---------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter("SALARY">"AVG_SALARY")
Statistics
----------------------------------------------------------
1 recursive calls
0 db block gets
7 consistent gets
0 physical reads
0 redo size
5220 bytes sent via SQL*Net to client
557 bytes received via SQL*Net from client
5 SQL*Net roundtrips to/from client
1 sorts (memory)
0 sorts (disk)
51 rows processed
不过,正如 Conrad 指出的那样,解析函数方法需要排序,因此它应该比传统方法使用更多的 PGA。您将以减少的 I/O 换取增加的 RAM。通常这是一个理想的权衡,但这是您应该注意的事情。