oracle中sql语句的优化

时间:2021-09-13 01:03:15

一、执行顺序及优化细则

1.表名顺序优化
(1) 基础表放下面,当两表进行关联时数据量少的表的表名放右边
表或视图:
Student_info (30000条数据)
Description_info (30条数据)
select *
from description_info di
,student_info si –学生信息表
where si.student_id = di.lookup_code(+)
and di.lookup_type(+) = ‘STUDENT_ID’

select *
from student_info si–学生信息表
,description_info di
where si.student_id = di.lookup_code(+)
and di.lookup_type(+) = ‘STUDENT_ID’
以student_info作为基础表,你会发现运行的速度会有很大的差距。

(2) 当出现多个表时,关联表被称之为交叉表,交叉表作为基础表
select *
from description_info di
,description_info di2
,student_info si –学生信息表
where si.student_id = di.lookup_code(+)
and di.lookup_type(+) = ‘STUDENT_ID’
and si.school_id = di.lookup_code(+)
and di.lookup_type(+) = ‘SCHOOL_ID’

select *
from student_info si–学生信息表
,description_info di
,description_info di2
where si.student_id = di.lookup_code(+)
and di.lookup_type(+) = ‘STUDENT_ID’
and si.school_id = di.lookup_code(+)
and di.lookup_type(+) = ‘SCHOOL_ID’
以student_info作为基础表,你会发现运行的速度会有很大的差距,
当基础表放在后面,这样的执行速度会明显快很多。

2.where执行顺序
where执行会从至下往上执行
select *
from student_info si –学生信息表
where si.school_id=10 –学院ID
and si.system_id=100–系ID
摆放where子句时,把能过滤大量数据的条件放在最下边
3. is null 和is not null
当要过滤列为空数据或不为空的数据时使用
select *
from student_info si –学生信息表
where si.school_id is null(当前列中的null为少数时用is not null,否则is null)

4.使用表别名
当查询时出现多个表时,查询时加上别名,
避免出现减少解析的时间字段歧义引起的语法错误。
5. where执行速度比having快
尽可能的使用where代替having
select from student_info si
group by si.student_id
having si.system_id!=100
and si.school_id!=10
(select from student_info si
wehre si.system_id!=100
and si.school_id!=10
group by si.student_id)
6. * 号引起的执行效率
尽量减少使用select * 来进行查询,当你查询使用*,
数据库会进行解析并将*转换为全部列。

二、替代优化

1、用>=替代>
select ui.user_name
from user_info ui–员工信息表
where ui.student_id>=10

select ui.user_name
from user_info ui–员工信息表
where ui.student_id>9
执行时>=会比>执行得要快

2、用UNION替换OR (适用于索引列)
select ui.user_name
from user_info ui–员工信息表
where ui.student_id=10
union
select ui.user_name
from user_info ui–员工信息表
where ui.student_id=2
上面语句可有效避免全表查询
select ui.user_name
from user_info ui–员工信息表
where ui.student_id=10
or ui.student_id=2
如果坚持要用OR, 可以把返回记录最少的索引列写在最前面

3、用in 代替or
select ui.user_name
from user_info ui–员工信息表
where ui.student_id=10
or ui.student_id=20
or ui.student_id=30
改成
select ui.user_name
from user_info ui–员工信息表
where ui.student_id in (10,20,30)
执行会更有效率

4、 Union All 与Union
Union All重复输出两个结果集合中相同记录
如果两个并集中数据都不一样.那么使用Union All 与Union是没有区别的,
select ui.user_name
from user_info ui–员工信息表
where ui.student_id=10
union All
select ui.user_name
from user_info ui–员工信息表
where ui.student_id=2

select ui.user_name
from user_info ui–员工信息表
where ui.student_id=10
union
select ui.user_name
from user_info ui–员工信息表
where ui.student_id=2
但Union All会比Union要执行得快

5、分离表和索引
总是将你的表和索引建立在另外的表空间内
决不要将这些对象存放到SYSTEM表空间里

三、一些优化技巧

1、计算表的记录数时

select count(si.student_id)
from Student_info si(student_id为索引)

select count(*) from Student_info si
执行时.上面的语句明显会比下面没有用索引统计的语句要快

2.使用函数提高SQL执行速度

当出现复杂的查询sql语名,可以考虑使用函数来提高速度
查询学生信息并查询学生(李明)个人信息与的数学成绩排名

select di.description student_name
,(select res.order_num–排名
from result res
where res.student_id = di.student_id
order by result_math) order_num
from description_info di
,student_info si –学生信息表
where si.student_id = di.lookup_code(+)
and di.lookup_type(+) = ‘STUDENT_ID’
and di.description = ‘李明’

而且我们将上面order_num排名写成一个fuction时
create or replace package body order_num_pkg is
function order_num(p_student_id number) return_number is
v_return_number number;
begin
select res.order_num –排名
into v_return_number
from result res
where res.student_id = di.student_id
order by result_math;
return v_return_number;
exception
when others then
null;
return null;
end;
end order_num_pkg;
执行
select di.description student_name
,order_num_pkg.order_num(di.student_id) order_num
from description_info di
,student_info si –学生信息表
where si.student_id = di.lookup_code(+)
and di.lookup_type(+) = ‘STUDENT_ID’
and di.description = ‘李明’
执行查询时的速度也会有所提高

3.减少访问数据库的次数

执行次数的减少(当要查询出student_id=100的学生和student_id=20的学生信息时)
select address_id
from student_info si –学生信息表
where si.student_id=100

select address_id
from student_info si –学生信息表
where si.student_id=20
都进行查询.这样的效率是很低的
而进行
(
select si.address_id,si2.address_id
from student_info si –学生信息表
,student_info si2
where si.student_id=100
and si2.student_id=20

select decode(si.student_id,100,address_id)
,decode(si.student_id,20,address_id)
from student_info si
)
执行速度是提高了,但可读性反而差了..
所以这种写法个人并不太推荐

4、用Exists(Not Exists)代替In(Not In)

在执行当中使用Exists或者Not Exists可以高效的进行查询

5、Exists取代Distinct取唯一值的

取出关联表部门对员工时,这时取出员工部门时,出现多条..
select distinct di.dept_name
from departments_info di –部门表
,user_info ui –员工信息表
where ui.dept_no = di.dept_no
可以修改成
select di.dept_name
from departments_info di –部门表
where exists (select ‘X’
from user_info ui –员工信息表
where di.dept_no = ui.dept_no)
6、用表连接代替Exists
通过表的关联来代替exists会使执行更有效率
select ui.user_name
from user_info ui–员工信息表
where exists (select ‘x ’
from departments_info di–部门表
where di.dept_no = ui.dept_no
and ui.dept_cat = ‘IT’);
执行是比较快,但还可以使用表的连接取得更快的查询效率
select ui.user_name
from departments_info di
,user_info ui –员工信息表
where ui.dept_no = di.dept_no
and ui.department_type_code = ‘IT’

代码是经测试并进行优化所写,
以上只例子,具体使用还是要针对各个不同的具体的业务使用用Exists(Not Exists)代替In(Not In)

四、索引篇

1、运算导致的索引失效

select di.description student_name
,(select res.order_num–排名
from result res
where res.student_id = di.student_id
order by result_math) order_num
from description_info di
,student_info si –学生信息表
where si.student_id = di.lookup_code(+)
and di.lookup_type(+) = ‘STUDENT_ID’
and si.student_id+0=100/student_id索引将失效/

2、类型转换导致的索引失效

select di.description student_name
,(select res.order_num–排名
from result res
where res.student_id = di.student_id
order by result_math) order_num
from description_info di
,student_info si –学生信息表
where si.student_id = di.lookup_code(+)
and di.lookup_type(+) = ‘STUDENT_ID’
and di.student_id=’100’

student_id为number类型的索引,当执行下列语句,
oracle会自动转换成
select di.description student_name
,(select res.order_num–排名
from result res
where res.student_id = di.student_id
order by result_math) order_num
from description_info di
,student_info si –学生信息表
where si.student_id = di.lookup_code(+)
and di.lookup_type(+) = ‘STUDENT_ID’
and di.student_id=to_number(‘100’)
所幸,只是解析并转换类型,并没有导到失效,
但要是写成下面,将会使用其失效
select di.description student_name
,(select res.order_num–排名
from result res
where res.student_id = di.student_id
order by result_math) order_num
from description_info di
,student_info si –学生信息表
where si.student_id = di.lookup_code(+)
and di.lookup_type(+) = ‘STUDENT_ID’
and to_char(di.student_id)=’100’

3、在索引列上进行计算引起的问题

select di.description student_name
,(select res.order_num–排名
from result res
where res.student_id = di.student_id
order by result_math) order_num
from description_info di
,student_info si –学生信息表
where si.student_id = di.lookup_code(+)
and di.lookup_type(+) = ‘STUDENT_ID’
and di.student_id-2=10
在索引列中进行运算,将会不使用索引而使用全表扫描
而将
select di.description student_name
,(select res.order_num–排名
from result res
where res.student_id = di.student_id
order by result_math) order_num
from description_info di
,student_info si –学生信息表
where si.student_id = di.lookup_code(+)
and di.lookup_type(+) = ‘STUDENT_ID’
and di.student_id=10+2
将会得到高效的运行速度

4、 Is not null引起的问题(student_id为索引)

不要把存在空值的列做为索引,否则无法使用索引
select ui.user_name
from user_info ui–员工信息表
where ui.student_id is not null–索引失效

select ui.user_name
from user_info ui–员工信息表
where ui.student_id>=-1–索引有效

5、Order by导致索引失效(student_id为索引)

select ui.user_name
from user_info ui–员工信息表
group by ui.student_id
而使用
select ui.user_name
from user_info ui–员工信息表
where ui.student_id>=-1
将使其有效,
在order by中只存在两种条件下可以使用索引
(ORDER BY中所有的列必须包含在相同的索引中并保持在索引中的排列顺序
ORDER BY中所有的列必须定义为非空. )

6、自动选择索引

如果表中有两个以上(包括两个)索引,其中有一个唯一性索引,而其他是非唯一性.
在这种情况下,ORACLE将使用唯一性索引而完全忽略非唯一性索引.

7、 !=导致索引失效

select ui.user_name
from user_info ui–员工信息表
where ui.student_id!=0
在Where中使用!=将会把索引失效

8、%导致的索引失效

select di.description student_name
,(select res.order_num–排名
from result res
where res.student_id = di.student_id
order by result_math) order_num
from description_info di
,student_info si –学生信息表
where si.student_id = di.lookup_code(+)
and di.lookup_type(+) = ‘STUDENT_ID’
and di.look_code Like ‘%12’/look_code为索引,索引将失效/

select di.description student_name
,(select res.order_num–排名
from result res
where res.student_id = di.student_id
order by result_math) order_num
from description_info di
,student_info si –学生信息表
where si.student_id = di.lookup_code(+)
and di.lookup_type(+) = ‘STUDENT_ID’
and di.look_code Like ‘12%’/索引有效/
以上只例子,具体还是要针对各个不同的具体的业务使用

五、oracle 中的not Exists与Not in的性能巨大差异
Not Exists与Not in的作用同样是排除数据,在oracle 中使用not in并不象mysql中的执行那么快,如(
select jt1.doc_num –单据号码
,oalc.description school_name –学校名称
,oalc2.description system_name –系名称
,oalc.description class_name –班级名称
from java_table1 jt1
,java_table_description oalc
,java_table_description oalc2
,java_table_description oalc3
where oalc.lookup_type(+) = ‘JAVA_SCHOOL_NAME’
and jt1.school_id = oalc.lookup_code(+)
and oalc2.lookup_type(+) = ‘JAVA_SYSTEM_NAME’
and jt1.system_id = oalc2.lookup_code(+)
and oalc3.lookup_type(+) = ‘JAVA_CLASS_NAME’
and jt1.class_id = oalc3.lookup_code(+)
and not exists
(select jt2.header_id
from java_table2 jt2 jt1.header_id = jt2.header_id))

select jt1.doc_num –单据号码
,oalc.description school_name –学校名称
,oalc2.description system_name –系名称
,oalc.description class_name –班级名称
from java_table1 jt1
,java_table_description oalc
,java_table_description oalc2
,java_table_description oalc3
where oalc.lookup_type(+) = ‘JAVA_SCHOOL_NAME’
and jt1.school_id = oalc.lookup_code(+)
and oalc2.lookup_type(+) = ‘JAVA_SYSTEM_NAME’
and jt1.system_id = oalc2.lookup_code(+)
and oalc3.lookup_type(+) = ‘JAVA_CLASS_NAME’
and jt1.class_id = oalc3.lookup_code(+)
and jt1.header_id not in (select jt2.header_id from java_table2 jt2)

当jt2表中的数据比较大时,就会出现巨大的差异,以上只能是我的个人理解与测试结果(java_table1 视图测试

数据量为36749,java_table2 为300条),如有其它可相互讨论

转自http://www.cnblogs.com/netjxz/archive/2009/09/21/1570991.html