oracle标量子查询简介和表连接改写

之前小鱼就听过了标量子查询，不过对于其中的细节理解还是远远不够，借助一部分资料和自己测试对标量子查询做一点简单的分析和介绍。 Oracle允许在select子句中包含单行子查询，这个也就是oracle的标量子查询，标量子查询有点类似于外连接，当使用到外连接时

之前小鱼就听过了标量子查询，不过对于其中的细节理解还是远远不够，借助一部分资料和自己测试对标量子查询做一点简单的分析和介绍。

Oracle允许在select子句中包含单行子查询，这个也就是oracle的标量子查询，标量子查询有点类似于外连接，当使用到外连接时我们可以灵活的将其转化为标量子查询。

SQL> create table t1 as select * from all_users;

Table created.

SQL> create table t2 as select * from all_objects;

Table created.

SQL> select a.object_id,(select b.username from t1 b where a.owner=b.username) f
rom t2 a;

49812 rows selected.

Execution Plan
----------------------------------------------------------
Plan hash value: 1364172329

--------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 53276 | 1560K| 152 (1)| 00:00:02 |
|* 1 | TABLE ACCESS FULL| T1 | 1 | 17 | 2 (0)| 00:00:01 |
| 2 | TABLE ACCESS FULL| T2 | 53276 | 1560K| 152 (1)| 00:00:02 |
--------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

1 - filter("B"."USERNAME"=:B1)

Note
-----
- dynamic sampling used for this statement

Statistics
----------------------------------------------------------
13 recursive calls
0 db block gets
4282 consistent gets
0 physical reads
0 redo size
1176699 bytes sent via SQL*Net to client
37012 bytes received via SQL*Net from client
3322 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
49812 rows processed

标量子查询其实还是一个子查询，那么它究竟是如何查询的：首先走的是外部的查询，比如上一个sql语句执行计划，先全表扫描的T2 a，然后取T2 a表的每一行数据就去和T1 b去过滤，过滤条件是a.owner=b.username，如果符合则返回子查询的值，如果不符合则用null补充。当然这个时候还有个类似的filter去重的运算，对于t2 a中重复的数据行不用再去和t1 b去过滤。

而上面这个标量子查询的sql语句其实是等价于下面外连接sql语句的：
SQL> select a.object_id,b.username from t2 a,t1 b
2 where a.owner=b.username(+) ;

49812 rows selected.

Execution Plan
----------------------------------------------------------
Plan hash value: 535089106

------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 53276 | 2445K| 155 (2)| 00:00:02 |
|* 1 | HASH JOIN RIGHT OUTER| | 53276 | 2445K| 155 (2)| 00:00:02 |
| 2 | TABLE ACCESS FULL | T1 | 23 | 391 | 2 (0)| 00:00:01 |
| 3 | TABLE ACCESS FULL | T2 | 53276 | 1560K| 152 (1)| 00:00:02 |
------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

1 - access("A"."OWNER"="B"."USERNAME"(+))

Note
-----
- dynamic sampling used for this statement

Statistics
----------------------------------------------------------
9 recursive calls
0 db block gets
4043 consistent gets
0 physical reads
0 redo size
1176659 bytes sent via SQL*Net to client
37012 bytes received via SQL*Net from client
3322 SQL*Net roundtrips to/from client
2 sorts (memory)
0 sorts (disk)
49812 rows processed

而如果标量子查询中如果主查询的一行对应子查询返回有多个值，这个是不允许的
SQL> select a.username,b.object_id from t1 a,t2 b
2 where a.username=b.owner(+);

29742 rows selected.

Execution Plan
----------------------------------------------------------
Plan hash value: 1823443478

---------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
---------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 53276 | 2445K| 155 (2)| 00:00:02 |
|* 1 | HASH JOIN OUTER | | 53276 | 2445K| 155 (2)| 00:00:02 |
| 2 | TABLE ACCESS FULL| T1 | 23 | 391 | 2 (0)| 00:00:01 |
| 3 | TABLE ACCESS FULL| T2 | 53276 | 1560K| 152 (1)| 00:00:02 |
---------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

1 - access("A"."USERNAME"="B"."OWNER"(+))

Note
-----
- dynamic sampling used for this statement

Statistics
----------------------------------------------------------
7 recursive calls
0 db block gets
2726 consistent gets
0 physical reads
0 redo size
654542 bytes sent via SQL*Net to client
22294 bytes received via SQL*Net from client
1984 SQL*Net roundtrips to/from client
2 sorts (memory)
0 sorts (disk)
29742 rows processed

SQL> select a.username,(select b.object_id from t2 b where a.username=b.owner) f
rom t1 a;
select a.username,(select b.object_id from t2 b where a.username=b.owner) from t
1 a
*
ERROR at line 1:
ORA-01427: single-row subquery returns more than one row

这里由于a.username=b.owner，其中b.owner有多个相同的值，所以这里返回的b.object_id可能有多个值，这里就出现上述的ora-01427错误。

标量子查询中也可以有聚合函数的出现：
SQL> set autotrace traceonly;
SQL> select a.username,max(b.object_id) from t1 a,t2 b
2 where a.username=b.owner(+)
3 group by a.username;

23 rows selected.

Execution Plan
----------------------------------------------------------
Plan hash value: 577572187

----------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
----------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 53276 | 2445K| 159 (4)| 00:00:02 |
| 1 | HASH GROUP BY | | 53276 | 2445K| 159 (4)| 00:00:02 |
|* 2 | HASH JOIN OUTER | | 53276 | 2445K| 155 (2)| 00:00:02 |
| 3 | TABLE ACCESS FULL| T1 | 23 | 391 | 2 (0)| 00:00:01 |
| 4 | TABLE ACCESS FULL| T2 | 53276 | 1560K| 152 (1)| 00:00:02 |
----------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

2 - access("A"."USERNAME"="B"."OWNER"(+))

Note
-----
- dynamic sampling used for this statement

Statistics
----------------------------------------------------------
92 recursive calls
0 db block gets
777 consistent gets
685 physical reads
0 redo size
1169 bytes sent via SQL*Net to client
503 bytes received via SQL*Net from client
3 SQL*Net roundtrips to/from client
2 sorts (memory)
0 sorts (disk)
23 rows processed

SQL> select a.username,(select max(b.object_id) from t2 b where b.owner=a.userna
me) from t1 a;

23 rows selected.

Execution Plan
----------------------------------------------------------
Plan hash value: 367820

---------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
---------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 23 | 391 | 2 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | 30 | | |
|* 2 | TABLE ACCESS FULL| T2 | 533 | 15990 | 152 (1)| 00:00:02 |
| 3 | TABLE ACCESS FULL | T1 | 23 | 391 | 2 (0)| 00:00:01 |
---------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

2 - filter("B"."OWNER"=:B1)

Note
-----
- dynamic sampling used for this statement

Statistics
----------------------------------------------------------
9 recursive calls
0 db block gets
15929 consistent gets
0 physical reads
0 redo size
1206 bytes sent via SQL*Net to client
503 bytes received via SQL*Net from client
3 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
23 rows processed

但是我们注意到上述标量子查询却存在一个问题，就是无法将子查询展开为表连接，换句话说无法采用灵活的hash join outer的关联方式。

关于标量子查询和表关联的性能简介：
如果主查询返回的数据较多，而子查询中又没有高效的索引，关联列对应的主查询表又没有较多的重复值，那么这个标量子查询的执行成本是很大的,如上面的标量子查询和外连接的sql语句中可以看出外连接IO成本要明显小于标量子查询。

但是标量子查询oracle内部确是有优化的，优化器cache了中间的结果,如果结果集不大，子查询中又有高效的索引，那么这个标量子查询可能会比常规的表关联更加高效。

小鱼列出几种常会涉及到的标量子查询和表连接的sql改写：
1 最简单的标量子查询
table :a(a1,a2),b(a1,b2)
select a2,(select b2 from b where b.a1=a.a1) from a

表连接：
select a2,b2 from a,b where a.a1=b.a1(+);

2 子查询中限制返回一行数据
table :a(a1,a2),b(a1,b2)
select a2,(select b2 from b where b.a1=a.a1 and rownum=1) from a

表连接：
SELECT a2, c.b2
FROM a,
(SELECT b2, ROW_NUMBER () OVER (PARTITION BY a1 ORDER BY a1) cn
FROM b
WHERE cn = 1) c
WHERE a.a1 = c.b2(+);

3 子查询带有聚合函数
table :a(a1,a2),b(a1,b2)
select a2,(select sum(b2) from b where b.a1 = a.a1) from a

表连接1：
SELECT a2, bb.sum_value
FROM a,
( SELECT SUM (b2) sum_value, a1
FROM b
GROUP BY a1) bb
WHERE a.a1 = b.a1(+);
表连接2：
SELECT a2, SUM (b2)
FROM a, b
WHERE a.a1 = b.a1(+)
GROUP BY a2;

4 查询中包括好几个标量子查询
table :a(a1,a2),b(a1,b2),c(a1,b2,c2)
SELECT a.a2,
(SELECT c2
FROM b, c
WHERE b.a1 = a.a1 AND b.b2 = c.b2 AND ROWNUM = 1),
(SELECT b2
FROM b
WHERE b.a1 = a.a1 AND ROWNUM = 1),
(SELECT c2
FROM c
WHERE c.a1 = a.a1 AND ROWNUM = 1)
FROM a

表连接：
SELECT a.a2,
bb.c2,
cc.b2,
dd.c2
FROM (SELECT c2,
b.a1,
ROW_NUMBER () OVER (PARTITION BY b.a1 ORDER BY b.a1) cnt
FROM b, c
WHERE b.b2 = c.b2 AND cnt = 1) bb,
(SELECT b2, ROW_NUMBER () OVER (PARTITION BY a1 ORDER BY a1) cnt
FROM b
WHERE cnt = 1) cc,
(SELECT c2, ROW_NUMBER () OVER (PARTITION BY a1 ORDER BY a1) cnt
FROM c
WHERE cnt = 1) dd,
a
WHERE a.a1 = bb.a1(+) AND a.a1 = cc.a1(+) AND a.a1 = dd.a1(+);

SELECT a2,
(SELECT SUM (c2)
FROM b, c
WHERE b.a1 = a.a1 AND b.b2 = c.b2),
(SELECT SUM (b2)
FROM b
WHERE b.a1 = a.a1),
(SELECT SUM (c2)
FROM c
WHERE c.a1 = a.a1)
FROM a
表连接：
SELECT a2,
bb.sum1,
cc.sum2,
dd.sum3
FROM ( SELECT SUM (c2) sum1, b.a1
FROM b, c
WHERE b.b2 = c.b2
GROUP BY b.a1) bb,
( SELECT SUM (b2) sum2, a1
FROM b
GROUP BY a1) cc,
( SELECT SUM (c2) sum3, a1
FROM c
GROUP BY a1) dd,
a
WHERE a.a1 = bb.a1(+) AND a.a1 = cc.a1(+) AND a.a1 = dd.a1(+);

原文地址：oracle标量子查询简介和表连接改写, 感谢原作者分享。