[20191218]降序索引疑问4.txt
[20191218]降序索引疑问4.txt
--//前几天优化一个项目,我发现许多表里面有有隐含字段,一般开发很少建立函数索引.我自己检查发现里面存在大量的降序索引.
--//我感觉有点奇怪,为什么开发要建立大量降序索引有什么好处呢?
--//我在链接http://www.itpub.net/thread-2122088-1-1.html里面问这个问题,sqysl的解答给了我很好的提示,我通过例子说明使用降
--//序索引的一点点好处,通过例子说明:
1.环境:
SCOTT@book> @ ver1
PORT_STRING VERSION BANNER
------------------- -------------- ----------------------------------------------------------------------------
x86_64/Linux 2.4.xx 11.2.0.4.0 Oracle Database 11g Enterprise Edition Release 11.2.0.4.0 - 64bit Production
SCOTT@book> alter system set pga_aggregate_target=4G;
System altered.
SCOTT@book> create table t (id1 number,id2 number,vc varchar2(50));
Table created.
create index i_t_id1_id2 on t (id1,id2);
create index i_t_id1desc_id2desc on t (id1 desc,id2 desc);
create index i_t_id1desc_id2 on t (id1 desc,id2 );
create index i_t_id1_id2desc on t (id1 ,id2 desc );
SCOTT@book> insert into t select rownum,rownum,lpad('a',50,'a') from dual connect by level<=1e6;
1000000 rows created.
SCOTT@book> commit ;
Commit complete.
--//分析略.
SCOTT@book> select index_name,index_type,blevel,leaf_blocks,distinct_keys from dba_indexes where owner=user and table_name='T';
INDEX_NAME INDEX_TYPE BLEVEL LEAF_BLOCKS DISTINCT_KEYS
------------------------------ --------------------------- ---------- ----------- -------------
I_T_ID1_ID2 NORMAL 2 2623 1000000
I_T_ID1DESC_ID2DESC FUNCTION-BASED NORMAL 2 5877 1000000
I_T_ID1DESC_ID2 FUNCTION-BASED NORMAL 2 5618 1000000
I_T_ID1_ID2DESC FUNCTION-BASED NORMAL 2 2753 1000000
--//无论那种倒序索引,索引都比正序索引大,特别是第1个字段选择desc的情况.因为我插入的字段是自增字段.
--//倒序索引分裂都是50-50分裂.
2.测试例子:
SCOTT@book> alter session set statistics_level = all;
Session altered.
SCOTT@book> select * from (select * from t where id1<=1e6 order by id1 desc,id2 ) where rownum<=10;
ID1 ID2 VC
---------- ---------- --------------------------------------------------
1000000 1000000 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
999999 999999 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
999998 999998 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
999997 999997 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
999996 999996 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
999995 999995 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
999994 999994 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
999993 999993 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
999992 999992 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
999991 999991 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
10 rows selected.
SCOTT@book> @ dpc '' ''
PLAN_TABLE_OUTPUT
-------------------------------------
SQL_ID 20hb1uyyqx4tf, child number 0
-------------------------------------
select * from (select * from t where id1<=1e6 order by id1 desc,id2 )
where rownum<=10
Plan hash value: 404407004
-------------------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Starts | E-Rows |E-Bytes| Cost (%CPU)| E-Time | A-Rows | A-Time | Buffers |
-------------------------------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | | | 4 (100)| | 10 |00:00:00.01 | 6 |
|* 1 | COUNT STOPKEY | | 1 | | | | | 10 |00:00:00.01 | 6 |
| 2 | VIEW | | 1 | 10 | 530 | 4 (0)| 00:00:01 | 10 |00:00:00.01 | 6 |
| 3 | TABLE ACCESS BY INDEX ROWID| T | 1 | 1000K| 58M| 4 (0)| 00:00:01 | 10 |00:00:00.01 | 6 |
|* 4 | INDEX RANGE SCAN | I_T_ID1DESC_ID2 | 1 | 10 | | 3 (0)| 00:00:01 | 10 |00:00:00.01 | 4 |
-------------------------------------------------------------------------------------------------------------------------------------------
Query Block Name / Object Alias (identified by operation id):
-------------------------------------------------------------
1 - SEL$1
2 - SEL$2 / from$_subquery$_001@SEL$1
3 - SEL$2 / T@SEL$2
4 - SEL$2 / T@SEL$2
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter(ROWNUM<=10)
4 - access("T"."SYS_NC00004$">=HEXTORAW('3BFDFF') AND "T"."SYS_NC00004$" IS NOT NULL)
filter(SYS_OP_UNDESCEND("T"."SYS_NC00004$")<=1000000)
32 rows selected.
--//逻辑读仅仅6个.而其它2个索引都无法达到这样的效果.
select * from (select /*+ index(t I_T_ID1DESC_ID2DESC) */ * from t where id1<=1e6 order by id1 desc,id2 ) where rownum<=10;
select * from (select /*+ index_desc(t I_T_ID1DESC_ID2DESC) */ * from t where id1<=1e6 order by id1 desc,id2 ) where rownum<=10;
select * from (select /*+ index(t I_T_ID1_ID2) */ * from t where id1<=1e6 order by id1 desc,id2 ) where rownum<=10;
select * from (select /*+ index_desc(t I_T_ID1_ID2) */ * from t where id1<=1e6 order by id1 desc,id2 ) where rownum<=10;
--//结果不在贴出.大家可以自行测试.
3.继续测试:
SCOTT@book> alter index I_T_ID1DESC_ID2 invisible;
Index altered.
SCOTT@book> select * from (select * from t where id1<=1e6 order by id1 desc,id2 ) where rownum<=1;
ID1 ID2 VC
---------- ---------- --------------------------------------------------
1000000 1000000 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
SCOTT@book> @ dpc '' ''
PLAN_TABLE_OUTPUT
-------------------------------------
SQL_ID 2r7j4a5gdhpnj, child number 0
-------------------------------------
select * from (select * from t where id1<=1e6 order by id1 desc,id2 )
where rownum<=1
Plan hash value: 3299198703
----------------------------------------------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Starts | E-Rows |E-Bytes|E-Temp | Cost (%CPU)| E-Time | A-Rows | A-Time | Buffers | Reads | OMem | 1Mem | Used-Mem |
----------------------------------------------------------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | | | | 17500 (100)| | 1 |00:00:00.38 | 9285 | 74 | | | |
|* 1 | COUNT STOPKEY | | 1 | | | | | | 1 |00:00:00.38 | 9285 | 74 | | | |
| 2 | VIEW | | 1 | 1000K| 50M| | 17500 (1)| 00:03:30 | 1 |00:00:00.38 | 9285 | 74 | | | |
|* 3 | SORT ORDER BY STOPKEY| | 1 | 1000K| 58M| 69M| 17500 (1)| 00:03:30 | 1 |00:00:00.38 | 9285 | 74 | 65M| 2806K| |
|* 4 | TABLE ACCESS FULL | T | 1 | 1000K| 58M| | 2744 (1)| 00:00:33 | 1000K|00:00:00.12 | 9285 | 74 | | | |
----------------------------------------------------------------------------------------------------------------------------------------------------------------------
Query Block Name / Object Alias (identified by operation id):
-------------------------------------------------------------
1 - SEL$1
2 - SEL$2 / from$_subquery$_001@SEL$1
3 - SEL$2
4 - SEL$2 / T@SEL$2
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter(ROWNUM<=1)
3 - filter(ROWNUM<=1)
4 - filter("ID1"<=1000000)
32 rows selected.
--//oracle选择全表扫描.
SCOTT@book> alter index I_T_ID1DESC_ID2 visible;
Index altered.
4.再继续测试:
--//测试第1个字段正序,第2个字段倒序的情况.
SCOTT@book> select * from (select * from t where id1<=1e6 order by id1 ,id2 desc ) where rownum<=5;
ID1 ID2 VC
---------- ---------- --------------------------------------------------
1 1 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
2 2 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
3 3 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
4 4 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
5 5 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
SCOTT@book> @ dpc '' ''
PLAN_TABLE_OUTPUT
-------------------------------------
SQL_ID 8f5yj8c3frpnc, child number 0
-------------------------------------
select * from (select * from t where id1<=1e6 order by id1 ,id2 desc )
where rownum<=5
Plan hash value: 2787951352
-------------------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Starts | E-Rows |E-Bytes| Cost (%CPU)| E-Time | A-Rows | A-Time | Buffers |
-------------------------------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | | | 4 (100)| | 5 |00:00:00.01 | 6 |
|* 1 | COUNT STOPKEY | | 1 | | | | | 5 |00:00:00.01 | 6 |
| 2 | VIEW | | 1 | 5 | 265 | 4 (0)| 00:00:01 | 5 |00:00:00.01 | 6 |
| 3 | TABLE ACCESS BY INDEX ROWID| T | 1 | 1000K| 58M| 4 (0)| 00:00:01 | 5 |00:00:00.01 | 6 |
|* 4 | INDEX RANGE SCAN | I_T_ID1_ID2DESC | 1 | 5 | | 3 (0)| 00:00:01 | 5 |00:00:00.01 | 4 |
-------------------------------------------------------------------------------------------------------------------------------------------
Query Block Name / Object Alias (identified by operation id):
-------------------------------------------------------------
1 - SEL$1
2 - SEL$2 / from$_subquery$_001@SEL$1
3 - SEL$2 / T@SEL$2
4 - SEL$2 / T@SEL$2
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter(ROWNUM<=5)
4 - access("ID1"<=1000000)
31 rows selected.
--//可以发现很好第使用I_T_ID1_ID2DESC索引.逻辑读也很小,仅仅6.
SCOTT@book> alter index I_T_ID1_ID2DESC invisible;
Index altered.
select * from (select * from t where id1<=1e6 order by id1 ,id2 desc ) where rownum<=1;
SCOTT@book> select * from (select * from t where id1<=1e6 order by id1 ,id2 desc ) where rownum<=5;
ID1 ID2 VC
---------- ---------- --------------------------------------------------
1 1 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
2 2 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
3 3 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
4 4 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
5 5 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
SCOTT@book> @ dpc '' ''
PLAN_TABLE_OUTPUT
-------------------------------------
SQL_ID 8f5yj8c3frpnc, child number 0
-------------------------------------
select * from (select * from t where id1<=1e6 order by id1 ,id2 desc )
where rownum<=5
Plan hash value: 2145689175
-----------------------------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Starts | E-Rows |E-Bytes| Cost (%CPU)| E-Time | A-Rows | A-Time | Buffers | Reads |
-----------------------------------------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | | | 4 (100)| | 5 |00:00:00.01 | 6 | 1 |
|* 1 | COUNT STOPKEY | | 1 | | | | | 5 |00:00:00.01 | 6 | 1 |
| 2 | VIEW | | 1 | 5 | 265 | 4 (0)| 00:00:01 | 5 |00:00:00.01 | 6 | 1 |
| 3 | TABLE ACCESS BY INDEX ROWID | T | 1 | 1000K| 58M| 4 (0)| 00:00:01 | 5 |00:00:00.01 | 6 | 1 |
|* 4 | INDEX RANGE SCAN DESCENDING| I_T_ID1DESC_ID2 | 1 | 5 | | 3 (0)| 00:00:01 | 5 |00:00:00.01 | 4 | 1 |
-----------------------------------------------------------------------------------------------------------------------------------------------------
Query Block Name / Object Alias (identified by operation id):
-------------------------------------------------------------
1 - SEL$1
2 - SEL$2 / from$_subquery$_001@SEL$1
3 - SEL$2 / T@SEL$2
4 - SEL$2 / T@SEL$2
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter(ROWNUM<=5)
4 - access("T"."SYS_NC00004$" IS NOT NULL AND "T"."SYS_NC00004$">=HEXTORAW('3BFDFF') )
filter(SYS_OP_UNDESCEND("T"."SYS_NC00004$")<=1000000)
32 rows selected.
--//你会发现一个奇特现象,oracle会使用"倒过来"的一个索引,也就是id1 desc,id2 asc的索引.
SCOTT@book> alter index I_T_ID1DESC_ID2 invisible;
Index altered.
SCOTT@book> select * from (select * from t where id1<=1e6 order by id1 ,id2 desc ) where rownum<=5;
ID1 ID2 VC
---------- ---------- --------------------------------------------------
1 1 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
2 2 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
3 3 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
4 4 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
5 5 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
SCOTT@book> @ dpc '' ''
PLAN_TABLE_OUTPUT
-------------------------------------
SQL_ID 8f5yj8c3frpnc, child number 0
-------------------------------------
select * from (select * from t where id1<=1e6 order by id1 ,id2 desc )
where rownum<=5
Plan hash value: 3299198703
----------------------------------------------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Starts | E-Rows |E-Bytes|E-Temp | Cost (%CPU)| E-Time | A-Rows | A-Time | Buffers | Reads | OMem | 1Mem | Used-Mem |
----------------------------------------------------------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | | | | 17500 (100)| | 5 |00:00:00.41 | 9279 | 9269 | | | |
|* 1 | COUNT STOPKEY | | 1 | | | | | | 5 |00:00:00.41 | 9279 | 9269 | | | |
| 2 | VIEW | | 1 | 1000K| 50M| | 17500 (1)| 00:03:30 | 5 |00:00:00.41 | 9279 | 9269 | | | |
|* 3 | SORT ORDER BY STOPKEY| | 1 | 1000K| 58M| 69M| 17500 (1)| 00:03:30 | 5 |00:00:00.41 | 9279 | 9269 | 2048 | 2048 | 2048 (0)|
|* 4 | TABLE ACCESS FULL | T | 1 | 1000K| 58M| | 2744 (1)| 00:00:33 | 1000K|00:00:00.15 | 9279 | 9269 | | | |
----------------------------------------------------------------------------------------------------------------------------------------------------------------------
Query Block Name / Object Alias (identified by operation id):
-------------------------------------------------------------
1 - SEL$1
2 - SEL$2 / from$_subquery$_001@SEL$1
3 - SEL$2
4 - SEL$2 / T@SEL$2
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter(ROWNUM<=5)
3 - filter(ROWNUM<=5)
4 - filter("ID1"<=1000000)
32 rows selected.
--//选择的是全表扫描.
SCOTT@book> alter index i_t_id1desc_id2desc invisible;
Index altered.
SCOTT@book> select * from (select * from t where id1<=1e6 order by id1 desc ,id2 desc ) where rownum<=1;
ID1 ID2 VC
---------- ---------- --------------------------------------------------
1000000 1000000 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
SCOTT@book> @ dpc '' ''
PLAN_TABLE_OUTPUT
-------------------------------------
SQL_ID 66z6w3qc77jwp, child number 0
-------------------------------------
select * from (select * from t where id1<=1e6 order by id1 desc ,id2
desc ) where rownum<=1
Plan hash value: 3873686303
----------------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Starts | E-Rows |E-Bytes| Cost (%CPU)| E-Time | A-Rows | A-Time | Buffers |
----------------------------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | | | 4 (100)| | 1 |00:00:00.01 | 4 |
|* 1 | COUNT STOPKEY | | 1 | | | | | 1 |00:00:00.01 | 4 |
| 2 | VIEW | | 1 | 1 | 53 | 4 (0)| 00:00:01 | 1 |00:00:00.01 | 4 |
| 3 | TABLE ACCESS BY INDEX ROWID | T | 1 | 1000K| 58M| 4 (0)| 00:00:01 | 1 |00:00:00.01 | 4 |
|* 4 | INDEX RANGE SCAN DESCENDING| I_T_ID1_ID2 | 1 | 1 | | 3 (0)| 00:00:01 | 1 |00:00:00.01 | 3 |
----------------------------------------------------------------------------------------------------------------------------------------
--//也能很好的选择正序索引.
5.总结:
--//通过以上测试,仅仅一种特殊的情况才需要建立倒序索引.存在id1 desc,id2 asc的情况下,建议完全可反过来建立(特别对于自增序列字段),
--//就是反过来建立索引id1 ,id2 desc,这样的索引占用磁盘空间更小.
--//其它情况我看不到降序索引的优势.
[20191218]降序索引疑问4.txt的更多相关文章
- [20181124]关于降序索引问题3.txt
[20181124]关于降序索引问题3.txt --//链接:blog.itpub.net/267265/viewspace-2221425/,探讨降序索引中索引的键值.--//实际上使用函数sys_ ...
- [20181124]关于降序索引问题2.txt
[20181124]关于降序索引问题2.txt --//链接:blog.itpub.net/267265/viewspace-2221425/,探讨降序索引中索引的键值.--//实际上使用函数sys_ ...
- [20190910]关于降序索引问题5.txt
[20190910]关于降序索引问题5.txt --//测试了索引TERM使用0xfe表示,回想到以前遇到降序索引的特殊字符编码问题,现在可是忘得一干二净.--//现在想想当时自己怎么猜测出来的,^_ ...
- MySQL 8 新特性之降序索引
MySQL 8.0终于支持降序索引了.其实,从语法上,MySQL 4就支持了,但正如官方文档所言,"they are parsed but ignored",实际创建的还是升序索引 ...
- MySQL8新增降序索引
MySQL8新增降序索引 桃花坞里桃花庵,桃花庵里桃花仙.桃花仙人种桃树,又摘桃花卖酒钱. 一.MySQL5.7 降序索引 MySQL在语法上很早就已经支持降序索引,但实际上创建的却仍然是升序索引,如 ...
- 谈谈MySQL中的降序索引 order by id DESC / ASC
今天这篇主要讲order by 语句中的多个字段asc desc的问题.mysql5中,索引存储的排序方式是ASC的,没有DESC的索引.现在能够理解为啥order by 默认是按照ASC来排序的了吧 ...
- Shell练习 统计单词个数，降序排列
原文:https://leetcode.com/problems/word-frequency/ Write a bash script to calculate the frequency of e ...
- 选取两个有序数组中最大的K个值，降序存入另一个数组中
原题: 假设有两个有序的整型数组int *a1, int *a2,长度分别为m和n.试用C语言写出一个函数选取两个数组中最大的K个值(K可能大于m+n)写到int *a3中,保持a3降序,并返回a3实 ...
- [20190423]那个更快的疑问3.txt
[20190423]那个更快的疑问3.txt --//前一阵子,做了11g在单表单条记录唯一索引扫描的测试,摘要如下:--//参考链接:http://blog.itpub.net/267265/vie ...
随机推荐
- k8s 上部署 Redis 三主三从 集群
目录 介绍 为什么要使用Redis? 什么是Redis群集? 在Kubernetes中部署Redis集群 从 GitHub 上下载: 创建pv 创建statefulset 创建service 初始化 ...
- window.open()打开新窗口 及参数
在jsp页面中需要使用到弹出窗口,想到js的window对象有一个open方法可以弹出窗口,于是对open方法进行记录. 首先是open方法的语法及定义: 定义: open() 方法用于打开一个新的浏 ...
- hadoop环境的安装 和 spark环境的安装
hadoop环境的安装1.前提:安装了java spark环境的安装1.前提:安装了java,python2.直接pip install pyspark就可以安装完成.(pip是python的软件安装 ...
- NIO-Channel
目录 NIO-Channel 目录 前言 什么是Channel 通道类型 如何使用 ServerSocketChannel SocketChannel FileChannel 总结 相关文献 NIO- ...
- 使用Docker测试静态网站
参考书籍 :第一本docker书[澳]James Turnbull 1.Sample网站的初始Dockerfile 文件目录如下: Dockerfile文件代码: 安装nginx 在容器中创建一个目 ...
- luogu P3152 正整数序列
题目描述 kkk制造了一个序列,这个序列里的数全是由正整数构成的.你别认为她的数列很神奇--其实就是1, 2, -, n而已.当然,n是给定的.kkk的同学lzn认为0是一个好数字(看上去很饱满有木有 ...
- SpringBoot中JdbcTemplate
步骤如下: 依赖: <dependency> <groupId>org.springframework.boot</groupId> <artifactId& ...
- LightOJ 1253 Misere NIM(反NIM博弈)
Alice and Bob are playing game of Misère Nim. Misère Nim is a game playing on k piles of stones, eac ...
- ACM-ICPC 2018 焦作赛区网络预赛 K题 Transport Ship
There are NN different kinds of transport ships on the port. The i^{th}ith kind of ship can carry th ...
- 2018HDU多校训练一 K - Time Zone
Chiaki often participates in international competitive programming contests. The time zone becomes a ...