BLOG文档结构图

在上一篇中http://blog.itpub.net/26736162/viewspace-1652985/，我们主要分析了一些单表查询的时候需要注意的内容，今天第二章也很简单，主要是关于排序方面的内容，以下贴出第二章的内容：

第 2 章 给查询结果排序

2.1 以指定的次序返回查询结果

2.2 按多个字段排序

2.3 按子串排序

2.4 TRANSLATE

2.5 按数字和字母混合字符串中的字母排序

2.6 处理排序空值

2.7 根据条件取不同列中的值来排序

排序基本上没有什么可以讲的，不过书中着重介绍了下translate的用法。

一.1translate用法

语法：TRANSLATE(char, from, to)

用法：

1.返回将出现在from中的每个字符替换为to中的相应字符以后的字符串。

2.若from比to字符串长，那么在from中比to中多出的字符将会被删除，或者认为from中多出的字符在to中与空对应

3.三个参数中有一个是空，返回值也将是空值。

09:43:50 SQL>  select translate('abcdefga','abc','wo')  from dual;

TRANSLA

-------

wodefgw

Elapsed: 00:00:00.14

09:43:57 SQL>  select translate('abcdefga','abc','')  from dual;

T

-

Elapsed: 00:00:00.00

SELECTtranslate('ab 你好 bcadefg','abcdefg','1234567'),translate('ab 你好 bcadefg','1abcdefg','1')FROMdual;

一.2按数字和字母混合字符串中的字母排序，采用translate函数来实现

09:52:01 SQL> create or replace view v as select empno || ' '||ename as data from scott.emp;

View created.

Elapsed: 00:00:00.54

09:52:07 SQL> select * from V

09:52:15   2  ;

DATA

---------------------------------------------------

9000 lastwiner

9001 lastwiner

7369 SMITH

7499 ALLEN

7521 WARD

7566 JONES

7654 MARTIN

7698 BLAKE

7782 CLARK

7788 SCOTT

7839 KING

7844 TURNER

7876 ADAMS

7900 JAMES

7902 FORD

7934 MILLER

16 rows selected.

Elapsed: 00:00:00.20

09:55:07 SQL>select data,translate(data,'- 0123456789','-') from V order by 2;

DATA                                                TRANSLATE(DATA,'-0123456789','-')

--------------------------------------------------- ------------------------------------------------------------------------------------------------------

7876 ADAMSADAMS

7499 ALLENALLEN

7698 BLAKEBLAKE

7782 CLARKCLARK

7902 FORDFORD

7900 JAMESJAMES

7566 JONESJONES

7839 KINGKING

7654 MARTINMARTIN

7934 MILLERMILLER

7788 SCOTTSCOTT

7369 SMITHSMITH

7844 TURNERTURNER

7521 WARDWARD

9001 lastwinerlastwiner

9000 lastwinerlastwiner

16 rows selected.

Elapsed: 00:00:00.10

09:55:33 SQL>

一.3关于order by排序的优化

关于SQL优化中有一个原则叫：避免使用耗费资源的操作(DISTINCT、UNION、MINUS、INTERSECT、ORDER BY、group by、SMJ、created index)

带有DISTINCT,UNION,MINUS,INTERSECT,ORDER BY的SQL语句会启动SQL引擎执行耗费资源的排序(SORT)功能. DISTINCT需要一次排序操作, 而其他的至少需要执行两次排序.

例如,一个UNION查询,其中每个查询都带有GROUP BY子句, GROUP BY会触发嵌入排序(NESTED SORT) ; 这样, 每个查询需要执行一次排序, 然后在执行UNION时, 又一个唯一排序(SORT UNIQUE)操作被执行而且它只能在前面的嵌入排序结束后才能开始执行. 嵌入的排序的深度会大大影响查询的效率.

通常, 带有UNION, MINUS , INTERSECT的SQL语句都可以用其他方式重写.

ORDER BY语句决定了Oracle如何将返回的查询结果排序。Order by语句对要排序的列没有什么特别的限制，也可以将函数加入列中(象联接或者附加等)。任何在Order by语句的非索引项或者有计算表达式都将降低查询速度。

仔细检查order by语句以找出非索引项或者表达式，它们会降低性能。解决这个问题的办法就是重写order by语句以使用索引，也可以为所使用的列建立另外一个索引，同时应绝对避免在order by子句中使用表达式。

●在频繁进行排序或分组(即进行group by或order by操作)的列上建立索引。

●如果待排序的列有多个，可以在这些列上建立复合索引(compound index)。

磁盘排序的开销是很大的，有几个方面的原因。首先，和内存排序相比较，它们特别慢；而且磁盘排序会消耗临时表空间中的资源。Oracle还必须分配缓冲池块来保持临时表空间中的块。无论什么时候，内存排序都比磁盘排序好，磁盘排序将会令任务变慢，并且会影响Oracle实例的当前任务的执行。还有，过多的磁盘排序将会令free buffer waits的值变高，从而令其它任务的数据块由缓冲中移走。

一.3.1总结

(1)采用索引避免排序：排序数据较多时

(2)去掉不必要的distinct，很多distinct是由于程序员对数据的了解不自信而多加的。

总而言之，排序是非常耗费CPU资源的，能不排序就不要排序，如果非得排序，可以考虑在排序列上建立合适的索引。

记得之前有个SQL，不加排序的话，秒级可以出结果，即响应速度很快，但是加上排序后得5或6分钟才可以，看了下是结果集很大，又得排序造成的。

这里简单举个例子吧：

一.3.2例子

[oracle@rhel6_lhr ~]$ sqlplus / as sysdba

SQL*Plus: Release 11.2.0.3.0 Production on Thu May 14 10:55:26 2015

Copyright (c) 1982, 2011, Oracle.  All rights reserved.

Connected to:

Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production

With the Partitioning, Automatic Storage Management, OLAP, Data Mining

and Real Application Testing options

10:55:26 SQL> conn lhr/lhr

Connected.

12:08:08 SQL> create table test_index_lhr as select * from dba_objects;

Table created.

Elapsed: 00:00:03.70

12:08:27 SQL> insert into test_index_lhr select * from test_index_lhr;

77241 rows created.

12:08:39 SQL> commit;

Commit complete.

Elapsed: 00:00:00.00

12:08:41 SQL>set autot traceonly explain stat

12:08:41 SQL>select  object_name from test_index_lhr where object_name is not null order by object_name;

154482 rows selected.

Elapsed: 00:00:01.18

Execution Plan

----------------------------------------------------------

Plan hash value: 1466335622

---------------------------------------------------------------------------------------------

| Id  | Operation          | Name           | Rows  | Bytes |TempSpc| Cost (%CPU)| Time     |

---------------------------------------------------------------------------------------------

|   0 | SELECT STATEMENT   |                |   155K|     9M||  3078   (1)| 00:00:37 |

|   1 |SORT ORDER BY|                |   155K|     9M|10M|  3078   (1)| 00:00:37 |

|*  2 |   TABLE ACCESS FULL| TEST_INDEX_LHR |   155K|     9M||   623   (1)| 00:00:08 |

---------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):

---------------------------------------------------

2 - filter("OBJECT_NAME" IS NOT NULL)

Note

-----

- dynamic sampling used for this statement (level=2)

Statistics

----------------------------------------------------------

10  recursive calls

6  db block gets

2808  consistent gets

614  physical reads

34996  redo size

3787521  bytes sent via SQL*Net to client

113801  bytes received via SQL*Net from client

10300  SQL*Net roundtrips to/from client

1  sorts (memory)

0  sorts (disk)

154482  rows processed

12:08:48 SQL> create index ind_test_inde on test_index_lhr(object_name) ;

Index created.

Elapsed: 00:00:02.45

12:08:58 SQL> EXEC dbms_stats.gather_table_stats(ownname => 'LHR', tabname=> 'test_index_lhr',   cascade => TRUE );

PL/SQL procedure successfully completed.

Elapsed: 00:00:02.62

12:09:04 SQL> select  object_name from test_index_lhr where object_name is not null order by object_name;

154482 rows selected.

Elapsed: 00:00:01.35

Execution Plan

----------------------------------------------------------

Plan hash value: 712275200

----------------------------------------------------------------------------------

| Id  | Operation        | Name          | Rows  | Bytes | Cost (%CPU)| Time     |

----------------------------------------------------------------------------------

|   0 | SELECT STATEMENT |               |   154K|  3771K|   766   (1)| 00:00:10 |

|*  1 |  INDEX FULL SCAN | IND_TEST_INDE |   154K|  3771K|   766   (1)| 00:00:10 |

----------------------------------------------------------------------------------

Predicate Information (identified by operation id):

---------------------------------------------------

1 - filter("OBJECT_NAME" IS NOT NULL)

Note

-----

- SQL plan baseline "SQL_PLAN_8kcy12j8f3s5n2a6f2b1f" used for this statement

Statistics

----------------------------------------------------------

704  recursive calls

64  db block gets

11715  consistent gets

37  physical reads

33236  redo size

3787521  bytes sent via SQL*Net to client

113801  bytes received via SQL*Net from client

10300  SQL*Net roundtrips to/from client

0sorts (memory)

0  sorts (disk)

154482  rows processed

12:09:09 SQL> select  owner, object_name from test_index_lhr where object_name is not null order by object_name;

154482 rows selected.

Elapsed: 00:00:01.28

Execution Plan

----------------------------------------------------------

Plan hash value: 1466335622

---------------------------------------------------------------------------------------------

| Id  | Operation          | Name           | Rows  | Bytes |TempSpc| Cost (%CPU)| Time     |

---------------------------------------------------------------------------------------------

|   0 | SELECT STATEMENT   |                |   154K|  4676K|       |  1947   (1)| 00:00:24 |

|   1 |  SORT ORDER BY     |                |   154K|  4676K|  6072K|  1947   (1)| 00:00:24 |

|*  2 |   TABLE ACCESS FULL| TEST_INDEX_LHR |   154K|  4676K|       |   623   (1)| 00:00:08 |

---------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):

---------------------------------------------------

2 - filter("OBJECT_NAME" IS NOT NULL)

Statistics

----------------------------------------------------------

6  recursive calls

6  db block gets

2241  consistent gets

895  physical reads

680  redo size

4232382  bytes sent via SQL*Net to client

113801  bytes received via SQL*Net from client

10300  SQL*Net roundtrips to/from client

1  sorts (memory)

0  sorts (disk)

154482  rows processed

12:09:22 SQL>select /*+index(a,IND_TEST_INDE)*/ owner,  object_name from test_index_lhrawhere object_name is not null order by object_name;

154482 rows selected.

Elapsed:00:00:09.59

Execution Plan

----------------------------------------------------------

Plan hash value: 880046030

----------------------------------------------------------------------------------------------

| Id  | Operation                   | Name           | Rows  | Bytes | Cost (%CPU)| Time     |

----------------------------------------------------------------------------------------------

|   0 | SELECT STATEMENT            |                |   154K|  4676K|   109K  (1)| 00:21:57 |

|   1 |  TABLE ACCESS BY INDEX ROWID| TEST_INDEX_LHR |   154K|  4676K|   109K  (1)| 00:21:57 |

|*  2 |   INDEX FULL SCAN           | IND_TEST_INDE  |   154K|       |   766   (1)| 00:00:10 |

----------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):

---------------------------------------------------

2 - filter("OBJECT_NAME" IS NOT NULL)

Statistics

----------------------------------------------------------

6  recursive calls

4  db block gets

122955  consistent gets

2198  physical reads

724  redo size

4392715  bytes sent via SQL*Net to client

113801  bytes received via SQL*Net from client

10300  SQL*Net roundtrips to/from client

0  sorts (memory)

0  sorts (disk)

154482  rows processed

12:14:23 SQL>