Null 值对索引排序的影响案例一则
--原SQL 语句如下：
select *
from (select tmp_tb.*, ROWNUM row_id
from (select wpid,
customer_id,
customer_name,
gender,
to_char(wi.call_time, 'yyyy-mm-dd hh24:mi:ss') call_time,
call_num,
phoneno,
tel,
cardtype,
cardnum,
note,
sessionid,
(select c.nodevalue
from sys_code c
where c.auto_id = wi.wp_type) wp_type,
smalltype,
order_type,
priority,
is_callback,
is_hidden,
district,
address,
summary,
wp_title,
wp_remark,
(select c.nodename
from sys_code c
where c.auto_id = wi.wp_source) wp_source,
(select c.nodevalue
from sys_code c
where c.auto_id = wi.class1) class1,
(select c.nodevalue
from sys_code c
where c.auto_id = wi.class2) class2,
(select c.nodevalue
from sys_code c
where c.auto_id = wi.class3) class3,
(select c.nodevalue
from sys_code c
where c.auto_id = wi.class4) class4,
upload_acceptance,
upload_handle,
upload_back,
upload_supervise,
is_repeat,
to_char(wi.starttime, 'yyyy-mm-dd hh24:mi:ss') starttime,
sender,
to_char(wi.updatetime, 'yyyy-mm-dd hh24:mi:ss') updatetime,
lastmessage,
(select sf.statename
from sys_function sf
where sf.auto_id = wi.state) state,
(select sf.statename
from sys_function sf
where sf.auto_id = wi.next_state) next_state,
next_state next_state_id,
dept_center,
(select stu.stru_name
from sys_stru stu
where stu.auto_id = wi.dept_level1) dept_level1,
(select stu.stru_name
from sys_stru stu
where stu.auto_id = wi.dept_level2) dept_level2,
(select stu.stru_name
from sys_stru stu
where stu.auto_id = wi.dept_level3) dept_level3,
(select stu.stru_name
from sys_stru stu
where stu.auto_id = wi.dept_cooperate1) dept_cooperate1,
(select stu.stru_name
from sys_stru stu
where stu.auto_id = wi.dept_cooperate2) dept_cooperate2,
is_reminder,
is_supervise,
is_overtime,
is_review,
to_char(wi.send_time, 'yyyy-mm-dd hh24:mi:ss') send_time,
to_char(wi.time_boundry, 'yyyy-mm-dd hh24:mi:ss') time_boundry,
source_address,
ciid1,
ciid2,
complaintname,
complainttel,
compensate_money,
back_source,
donereport,
reply_point,
to_char(wi.reply_time, 'yyyy-mm-dd hh24:mi:ss') reply_time,
solved,
passoperator,
to_char(wi.donetime, 'yyyy-mm-dd hh24:mi:ss') donetime,
dept_workno,
specialSeat_type,
to_char(wi.acceptTime_boundry, 'yyyy-mm-dd hh24:mi:ss') acceptTime_boundry,
to_char(wi.acceptTime, 'yyyy-mm-dd hh24:mi:ss') acceptTime,
orderSymbol,
timeout_flag,
send_flag,
getLastWorknoByWpid(wi.wpid) operatorno,
is_relation,
(select count(1)
from wp_info wp
where wp.relation_wpid = wi.wpid) total,
ep_flag,
is_difficult,
(select count(1)
from wp_early_warning_info
where wpid = wi.wpid) lid,
endresult
from xxx_info wi
where 1 = 1
order by wi.starttime desc) tmp_tb
where ROWNUM <= 10)
where row_id > 0
--执行时间 147s
通过 explain plan for 上述语句,获取其执行计划
select * from table(dbms_xplan.display()); PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Plan hash value: 964360475 ---------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes |TempSpc| Cost (%CPU)| Time |
---------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 10 | 116K| | 205K (1)| 00:41:04 |
| 1 | TABLE ACCESS BY INDEX ROWID| SYS_CODE | 1 | 17 | | 2 (0)| 00:00:01 |
|* 2 | INDEX UNIQUE SCAN | SYS_C005980 | 1 | | | 1 (0)| 00:00:01 |
| 3 | TABLE ACCESS BY INDEX ROWID| SYS_CODE | 1 | 17 | | 2 (0)| 00:00:01 |
|* 4 | INDEX UNIQUE SCAN | SYS_C005980 | 1 | | | 1 (0)| 00:00:01 |
| 5 | TABLE ACCESS BY INDEX ROWID| SYS_CODE | 1 | 17 | | 2 (0)| 00:00:01 |
|* 6 | INDEX UNIQUE SCAN | SYS_C005980 | 1 | | | 1 (0)| 00:00:01 |
| 7 | TABLE ACCESS BY INDEX ROWID| SYS_CODE | 1 | 17 | | 2 (0)| 00:00:01 |
|* 8 | INDEX UNIQUE SCAN | SYS_C005980 | 1 | | | 1 (0)| 00:00:01 |
| 9 | TABLE ACCESS BY INDEX ROWID| SYS_CODE | 1 | 17 | | 2 (0)| 00:00:01 |
|* 10 | INDEX UNIQUE SCAN | SYS_C005980 | 1 | | | 1 (0)| 00:00:01 |
| 11 | TABLE ACCESS BY INDEX ROWID| SYS_CODE | 1 | 17 | | 2 (0)| 00:00:01 |
|* 12 | INDEX UNIQUE SCAN | SYS_C005980 | 1 | | | 1 (0)| 00:00:01 |
| 13 | TABLE ACCESS BY INDEX ROWID| SYS_FUNCTION | 1 | 15 | | 1 (0)| 00:00:01 |
|* 14 | INDEX UNIQUE SCAN | SYS_FUNC_ID | 1 | | | 0 (0)| 00:00:01 |
| 15 | TABLE ACCESS BY INDEX ROWID| SYS_FUNCTION | 1 | 15 | | 1 (0)| 00:00:01 |
|* 16 | INDEX UNIQUE SCAN | SYS_FUNC_ID | 1 | | | 0 (0)| 00:00:01 |
| 17 | TABLE ACCESS BY INDEX ROWID| SYS_STRU | 1 | 21 | | 2 (0)| 00:00:01 |
|* 18 | INDEX UNIQUE SCAN | STRU_ID_PK | 1 | | | 1 (0)| 00:00:01 |
| 19 | TABLE ACCESS BY INDEX ROWID| SYS_STRU | 1 | 21 | | 2 (0)| 00:00:01 |
|* 20 | INDEX UNIQUE SCAN | STRU_ID_PK | 1 | | | 1 (0)| 00:00:01 |
| 21 | TABLE ACCESS BY INDEX ROWID| SYS_STRU | 1 | 21 | | 2 (0)| 00:00:01 |
|* 22 | INDEX UNIQUE SCAN | STRU_ID_PK | 1 | | | 1 (0)| 00:00:01 |
| 23 | TABLE ACCESS BY INDEX ROWID| SYS_STRU | 1 | 21 | | 2 (0)| 00:00:01 |
|* 24 | INDEX UNIQUE SCAN | STRU_ID_PK | 1 | | | 1 (0)| 00:00:01 |
| 25 | TABLE ACCESS BY INDEX ROWID| SYS_STRU | 1 | 21 | | 2 (0)| 00:00:01 |
|* 26 | INDEX UNIQUE SCAN | STRU_ID_PK | 1 | | | 1 (0)| 00:00:01 |
| 27 | SORT AGGREGATE | | 1 | 2 | | | |
|* 28 | INDEX RANGE SCAN | IDX_WPINFO_RELATIONWPID | 32 | 64 | | 1 (0)| 00:00:01 |
| 29 | SORT AGGREGATE | | 1 | 15 | | | |
|* 30 | INDEX RANGE SCAN | IDX_WEWI_WPID | 1 | 15 | | 1 (0)| 00:00:01 |
|* 31 | VIEW | | 10 | 116K| | 205K (1)| 00:41:04 |
|* 32 | COUNT STOPKEY | | | | | | |
| 33 | VIEW | | 1051K| 11G| | 205K (1)| 00:41:04 |
|* 34 | SORT ORDER BY STOPKEY | | 1051K| 778M| 1026M| 205K (1)| 00:41:04 |
| 35 | TABLE ACCESS FULL | WP_INFO | 1051K| 778M| | 33545 (1)| 00:06:43 |
--------------------------------------------------------------------------------------------------------------- Predicate Information (identified by operation id):
--------------------------------------------------- 2 - access("C"."AUTO_ID"=TO_NUMBER(:B1))
4 - access("C"."AUTO_ID"=TO_NUMBER(:B1))
6 - access("C"."AUTO_ID"=TO_NUMBER(:B1))
8 - access("C"."AUTO_ID"=TO_NUMBER(:B1))
10 - access("C"."AUTO_ID"=TO_NUMBER(:B1))
12 - access("C"."AUTO_ID"=TO_NUMBER(:B1))
14 - access("SF"."AUTO_ID"=TO_NUMBER(:B1))
16 - access("SF"."AUTO_ID"=TO_NUMBER(:B1))
18 - access("STU"."AUTO_ID"=:B1)
20 - access("STU"."AUTO_ID"=:B1)
22 - access("STU"."AUTO_ID"=:B1)
24 - access("STU"."AUTO_ID"=:B1)
26 - access("STU"."AUTO_ID"=:B1)
28 - access("WP"."RELATION_WPID"=:B1)
30 - access("WPID"=:B1)
31 - filter("ROW_ID">0)
32 - filter(ROWNUM<=10)
34 - filter(ROWNUM<=10) 该语句看起来挺长，但主要还是从WP_INFO 表取数据，无其他where条件限制，根据starttime 对全表倒序排序，取前十
执行计划看出WP_INFO表 行数1051K ，全表扫描，排序操作用了1026M空间。那么该如何优化呢？
既然存在排序行为，那么是否可以考虑在排序列建索引呢？索引存储了已排序好的非空列值，那么是否可以通过建索引的方式避免排序操作呢？ 检查该表发现starttime列已经存在索引。
那么为什么不走索引呢？难道一定要有限制条件？还是说空值限制了索引的使用？ starttime 列确实可为空。
针对空值有两种处理办法：
1. 修改表的列 starttime 属性非空,但生产库岂敢随意修改。
alter table xxx modify starttime not null;
2. Where 条件中添加限制，只取非空值，如下：
where 1 = 1
and wi.starttime is not null
order by wi.starttime desc) tmp_tb
查看新的执行计划： PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Plan hash value: 793409466 ----------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
----------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 10 | 116K| 12 (0)| 00:00:01 |
| 1 | TABLE ACCESS BY INDEX ROWID | SYS_CODE | 1 | 17 | 2 (0)| 00:00:01 |
|* 2 | INDEX UNIQUE SCAN | SYS_C005980 | 1 | | 1 (0)| 00:00:01 |
| 3 | TABLE ACCESS BY INDEX ROWID | SYS_CODE | 1 | 17 | 2 (0)| 00:00:01 |
|* 4 | INDEX UNIQUE SCAN | SYS_C005980 | 1 | | 1 (0)| 00:00:01 |
| 5 | TABLE ACCESS BY INDEX ROWID | SYS_CODE | 1 | 17 | 2 (0)| 00:00:01 |
|* 6 | INDEX UNIQUE SCAN | SYS_C005980 | 1 | | 1 (0)| 00:00:01 |
| 7 | TABLE ACCESS BY INDEX ROWID | SYS_CODE | 1 | 17 | 2 (0)| 00:00:01 |
|* 8 | INDEX UNIQUE SCAN | SYS_C005980 | 1 | | 1 (0)| 00:00:01 |
| 9 | TABLE ACCESS BY INDEX ROWID | SYS_CODE | 1 | 17 | 2 (0)| 00:00:01 |
|* 10 | INDEX UNIQUE SCAN | SYS_C005980 | 1 | | 1 (0)| 00:00:01 |
| 11 | TABLE ACCESS BY INDEX ROWID | SYS_CODE | 1 | 17 | 2 (0)| 00:00:01 |
|* 12 | INDEX UNIQUE SCAN | SYS_C005980 | 1 | | 1 (0)| 00:00:01 |
| 13 | TABLE ACCESS BY INDEX ROWID | SYS_FUNCTION | 1 | 15 | 1 (0)| 00:00:01 |
|* 14 | INDEX UNIQUE SCAN | SYS_FUNC_ID | 1 | | 0 (0)| 00:00:01 |
| 15 | TABLE ACCESS BY INDEX ROWID | SYS_FUNCTION | 1 | 15 | 1 (0)| 00:00:01 |
|* 16 | INDEX UNIQUE SCAN | SYS_FUNC_ID | 1 | | 0 (0)| 00:00:01 |
| 17 | TABLE ACCESS BY INDEX ROWID | SYS_STRU | 1 | 21 | 2 (0)| 00:00:01 |
|* 18 | INDEX UNIQUE SCAN | STRU_ID_PK | 1 | | 1 (0)| 00:00:01 |
| 19 | TABLE ACCESS BY INDEX ROWID | SYS_STRU | 1 | 21 | 2 (0)| 00:00:01 |
|* 20 | INDEX UNIQUE SCAN | STRU_ID_PK | 1 | | 1 (0)| 00:00:01 |
| 21 | TABLE ACCESS BY INDEX ROWID | SYS_STRU | 1 | 21 | 2 (0)| 00:00:01 |
|* 22 | INDEX UNIQUE SCAN | STRU_ID_PK | 1 | | 1 (0)| 00:00:01 |
| 23 | TABLE ACCESS BY INDEX ROWID | SYS_STRU | 1 | 21 | 2 (0)| 00:00:01 |
|* 24 | INDEX UNIQUE SCAN | STRU_ID_PK | 1 | | 1 (0)| 00:00:01 |
| 25 | TABLE ACCESS BY INDEX ROWID | SYS_STRU | 1 | 21 | 2 (0)| 00:00:01 |
|* 26 | INDEX UNIQUE SCAN | STRU_ID_PK | 1 | | 1 (0)| 00:00:01 |
| 27 | SORT AGGREGATE | | 1 | 2 | | |
|* 28 | INDEX RANGE SCAN | IDX_WPINFO_RELATIONWPID | 32 | 64 | 1 (0)| 00:00:01 |
| 29 | SORT AGGREGATE | | 1 | 15 | | |
|* 30 | INDEX RANGE SCAN | IDX_WEWI_WPID | 1 | 15 | 1 (0)| 00:00:01 |
|* 31 | VIEW | | 10 | 116K| 12 (0)| 00:00:01 |
|* 32 | COUNT STOPKEY | | | | | |
| 33 | VIEW | | 10 | 116K| 12 (0)| 00:00:01 |
| 34 | TABLE ACCESS BY INDEX ROWID| WP_INFO | 1051K| 778M| 12 (0)| 00:00:01 |
|* 35 | INDEX FULL SCAN DESCENDING| IDX_WPINFO_STIME | 10 | | 3 (0)| 00:00:01 |
---------------------------------------------------------------------------------------------------------- Predicate Information (identified by operation id):
--------------------------------------------------- 2 - access("C"."AUTO_ID"=TO_NUMBER(:B1))
4 - access("C"."AUTO_ID"=TO_NUMBER(:B1))
6 - access("C"."AUTO_ID"=TO_NUMBER(:B1))
8 - access("C"."AUTO_ID"=TO_NUMBER(:B1))
10 - access("C"."AUTO_ID"=TO_NUMBER(:B1))
12 - access("C"."AUTO_ID"=TO_NUMBER(:B1))
14 - access("SF"."AUTO_ID"=TO_NUMBER(:B1))
16 - access("SF"."AUTO_ID"=TO_NUMBER(:B1))
18 - access("STU"."AUTO_ID"=:B1)
20 - access("STU"."AUTO_ID"=:B1)
22 - access("STU"."AUTO_ID"=:B1)
24 - access("STU"."AUTO_ID"=:B1)
26 - access("STU"."AUTO_ID"=:B1)
28 - access("WP"."RELATION_WPID"=:B1)
30 - access("WPID"=:B1)
31 - filter("ROW_ID">0)
32 - filter(ROWNUM<=10)
35 - filter("WI"."STARTTIME" IS NOT NULL) cost 从205k 下降到了12 ，评估执行时间从41min 下降到0.01 s， 效果很好有木有。看下面执行计划走了索引全扫描，没了排序操作。 实际执行 0.06s ，从147s 下降到0.06s 效率提升了2000多倍。
要等很久的一个查询秒出了。想想就有点小激动，所以记录一下。 总结：索引只保存确定值，而 Null 值不确定所以并不被索引记录，如果根据索引排序，那么这个排序并没有Null值，排序结果对于原表来说不准确，数据库不能返回给你一个假的结果，所以不能走索引。
如果索引列根本没有空值呢？所有的值都被索引所记录？ 事实是不论这个列是否有空值，但既然可以为空，那么数据库就认为可能存在空值,坚决不能走索引。
通过修改列属性为非空，或在条件中限制只取非空值，数据库就知道索引中确实包含了你需要的所有值，数据库就可以放心大胆的走索引了。
同理类似,select max/min/count(1)/count(*) 都会因为可能存在空值导致返回结果不准确而不走索引。
Null 值对索引排序的影响案例一则的更多相关文章
- Oracle中NULL值与索引
NULL值是关系数据库系统布尔型(true,false,unknown)中比较特殊类型的一种值,通常称为UNKNOWN或空值,即是未知的,不确定的.由于NULL存在着无数的可能,因此NULL值也不等于 ...
- 网页页面NULL值对浏览器兼容性的影响
网页页面NULL值对浏览器兼容性的影响 近期做项目中一个页面中的input radio出现浏览器兼容性问题. 主要问题: 在谷歌浏览器,360急速模式和搜狗急速模式中给radio初始动态赋 ...
- PostgreSQL 数据库NULL值的默认排序行为与查询、索引定义规范 - nulls first\last, asc\desc
背景 在数据库中NULL值是指UNKNOWN的值,不存储任何值,在排序时,它排在有值的行前面还是后面通过语法来指定. 例如 -- 表示null排在有值行的前面 select * from tbl or ...
- mysq对存在null值的字段排序
1.建立学生表,建表sql如下: ),age int); 2.插入几条数据,包括id字段值为null的 ,),(,),(,),(),(); 3.我们查询下,可以看到存在id字段为空的值: 4.对学生表 ...
- 让索引包含null值的两种方法
1. 把有NULL值的列与一个常数,或者一个带有not null约束的列一同索引 create index ind_01 on t01(col01,1); 或者 create index ind_01 ...
- 【mysql】mysql null值
在数据表我们有时候有些表字段会为null,表示空.其实在mysql中null值是占用空间的. mysql手册如下解释 NULL columns require additional space in ...
- oracle 关于null值排序
在oracle中根据字段来desc排序的话null值可能会在数据的最前面.然而有时候我们查看数据的时候并不希望能够在前面看到这些null值的排序数据. 因此我查了一下: 1.排序的时候运用nvl(). ...
- 关于null值的排序
关于空值null的排序问题 Oracle排序中NULL值处理的五种常用方法: 1.缺省Oracle在Order by 时缺省认为null是最大值,所以如果是ASC升序则排在最后,DESC降序则排在 ...
- 使用复合索引代替单键索引，来避免单键有null值的情况
查看原表: SQL> select count(*) from t1; COUNT(*) ---------- 3229088 SQL> select count(*) from t1 w ...
随机推荐
- React Native之React速学教程(中)
概述 本篇为<React Native之React速学教程>的第一篇.本篇将从React的特点.如何使用React.JSX语法.组件(Component)以及组件的属性,状态等方面进行讲解 ...
- 关于安卓开发的学习一：webview
在网上看到几篇不错的博客,分享和学习一下! Android使用WebView加载网页 https://blog.csdn.net/tuke_tuke/article/details/51684254 ...
- vue2.0中的ref.html
前置知识请戳这里 获取DOM对象以及组件对象 <!DOCTYPE html> <html> <head> <meta charset="UTF-8& ...
- c windows控制台输出颜色文字
#include <windows.h> //设置文字颜色void SetColor(int ForgC){ WORD wColor; //We will need this handle ...
- According to TLD, tag fmt:formatDate must be empty, but is not 问题的解决
在执行jsp格式化后报错,检查下代码,发现变成如下的样式: <fmt:formatDate value="${cur.sa_date}" pattern="yyyy ...
- SpringMVC学习（五）——拦截器示例
部分内容摘自开涛的<跟我学SpringMVC.PDF> 拦截器,本质类似于AOP,主要的应用场景: 1.日志记录:记录请求信息的日志,以便进行信息监控.信息统计.计算PV等. 2.权限检查 ...
- Python unittest模块心得(一)
关于unittest模块的一些心得,主要是看官网的例子,加上一点自己的理解,官网地址:https://docs.python.org/3.6/library/unittest.html 基础概念介绍: ...
- SQL Server ->> SQL Server 2016新特性之 -- Dynamic Data Masking
Dynamic Data Masking是为了防止敏感数据暴露给未经授权的用户,以一种最小开销和维护成本的形式.Dynamic Data Masking用于表的字段,相当于盖住字段数据的一部分.比如一 ...
- 使用GDI技术创建ASP.NET验证码
using System; using System.Collections.Generic; using System.Linq; using System.Web; using System.Dr ...
- Angular4 微信的坑
1.不要重置对象的引用!(重置只应该在组件或服务的初始化时) why:会使页面产生闪烁 2.不要给图片绑定一个空的值或空的图片引用(如果值从服务器异步过来,那应该在初始化时给它一个默认值) why:会 ...