刚看了《oracle 高效设计》的讲到的,说到oracle中有标量子查询,性能要由于关联查询,但是我在SQL server中知道关联查询的性能要优于标量子查询。
我们来做个测试,看看性能:执行语句:
set
autotrace on
select a.username,count(
*
) from all_users a,all_objects b
where
a.username
=
b.owner(
+
)
group by a.username;
在SQL developer中执行,结果如下:
5356
recursive calls
0
db block gets
82152
consistent gets
再执下列语句:
set
autotrace
on
select
a.username,(
select
count
(
*
)
from
all_objects b
where
b.owner
=
a.username) cnt
from
all_users a;
执行结果,如下:
5371
recursive calls
0
db block gets
98645
consistent gets
这时发现,通过执行关联查询的性能要优于标量子查询。估计《Oracle 高效设计》是版本比较低,现在都是用的10gR2的版本有很大改变。
执行语句如下:
set
autotrace
on
select
a.username,
count
(
*
),
avg
(
object_id
)
from
all_users a,all_objects b
where
a.username
=
b.owner(
+
)
group
by
a.username;
执行结果:
5371
recursive calls
0
db block gets
82157
consistent gets
执行语句:
set
autotrace
on
select
username,to_number(substr(data,
1
,
10
)) cnt,to_number(substr(data,
11
))
avg
from
(
select
a.username,(
select
to_char(
count
(
*
),
'
fm0000000009
'
)
||
avg
(
object_id
)
from
all_objects b
where
b.owner
=
a.username) data
from
all_users a)
执行结果:
5356
recursive calls
0
db block gets
98556
consistent gets
执行语句:
set
autotrace
on
create
or
replace
type myType
as
object
(cnt
number
,
avg
number
);
select
username,a.data.cnt,a.data.
avg
from
(
select
username,(
select
myType(
count
(
*
),
avg
(
object_id
))
from
all_objects b
where
b.owner
=
a.username) data
from
all_users a) a;
执行结果:
5390
recursive calls
0
db block gets
98662
consistent gets
总结:
通过上面的测试发现,TOM的《Oracle高效设计》里谈到的标量子查询的性能要由于关联查询,在旧版本的中可能是可以的,在10gR2中是不成立的。