Computed Column（计算列）是自SQL Server 2005开始就有的特性。计算列的定义是一个表达式。表达式可以是非计算列，常量，函数间的组合。但是不可以是子查询。

计算列数据固化

默认情况下计算列的数据是存储在磁盘上，仅当计算列被查询引用是才进行实时计算。只在计算列在定义是添加了PERSISTED关键词是才将数据固化。

计算列上创建索引或者作为分区函数的引用列

计算列上是运行创建索引和作为分区函数的引用列。但是必须指定PERSISTED关键词。

用法其实很简单。那么这里有些问题。使用计算列的代价到底有多大？

INSERT发生时使用计算列和非计算列的性能区别

这里做一个测试。首先创建好两张表

IF EXISTS(SELECT * FROM sys.tables WHERE name = 'computed_column_test_computed')

BEGIN

    DROP TABLE dbo.computed_column_test_computed

END

IF EXISTS(SELECT * FROM sys.tables WHERE name = 'computed_column_test_noncomputed')

BEGIN

    DROP TABLE dbo.computed_column_test_noncomputed

END

CREATE TABLE dbo.computed_column_test_computed

(

    dttm DATETIME,

    dttm_year AS YEAR(dttm) PERSISTED,

    dttm_month AS MONTH(dttm) PERSISTED,

    dttm_nextday AS DATEADD(DAY,1,dttm) PERSISTED,

    dttm_previousday AS DATEADD(DAY,-1,dttm) PERSISTED,

    dttm_week AS DATEPART(ww,dttm),

    dttm_monthname AS CASE DATEPART(mm, dttm)

                        WHEN 1 THEN 'January'

                        WHEN 2 THEN 'February'

                        WHEN 3 THEN 'March'

                        WHEN 4 THEN 'April'

                        WHEN 5 THEN 'May'

                        WHEN 6 THEN 'June'

                        WHEN 7 THEN 'July'

                        WHEN 8 THEN 'August'

                        WHEN 9 THEN 'September'

                        WHEN 10 THEN 'October'

                        WHEN 11 THEN 'November'

                        WHEN 12 THEN 'December'

                      END PERSISTED,

    dttm_quarter AS DATEPART(qq,dttm) PERSISTED

)

CREATE TABLE dbo.computed_column_test_noncomputed

(

    dttm DATETIME,

    dttm_year SMALLINT,

    dttm_month SMALLINT,

    dttm_nextday DATETIME,

    dttm_previousday DATETIME,

    dttm_week INT,

    dttm_monthname VARCHAR(30),

    dttm_quarter VARCHAR(30)

)

然后开启IO和TIME的统计信息开关，然后分别对两张表进行数据插入。

SET STATISTICS TIME ON

SET STATISTICS IO ON

INSERT dbo.computed_column_test_computed(

dttm

)

SELECT DATEADD(SECOND, Num, GETDATE())

FROM dbo.Numbers

WHERE Num <= 1000000

INSERT dbo.computed_column_test_noncomputed(dttm ,

    dttm_year ,

    dttm_month ,

    dttm_nextday ,

    dttm_previousday ,

    dttm_week ,

    dttm_monthname ,

    dttm_quarter)

SELECT     GETDATE(),

        YEAR(GETDATE()) PERSISTED,

        MONTH(GETDATE()) PERSISTED,

        DATEADD(DAY,1,GETDATE()) ,

        DATEADD(DAY,-1,GETDATE()),

        DATEPART(ww,GETDATE()),

        CASE DATEPART(mm, GETDATE())

        WHEN 1 THEN 'January'

        WHEN 2 THEN 'February'

        WHEN 3 THEN 'March'

        WHEN 4 THEN 'April'

        WHEN 5 THEN 'May'

        WHEN 6 THEN 'June'

        WHEN 7 THEN 'July'

        WHEN 8 THEN 'August'

        WHEN 9 THEN 'September'

        WHEN 10 THEN 'October'

        WHEN 11 THEN 'November'

        WHEN 12 THEN 'December'

        END,

    DATEPART(qq,GETDATE())

FROM dbo.Numbers

WHERE Num <= 1000000

我的例子里面分别进行10万行、30万行、50万行和100万行数据的插入测试。一共有7个计算列。每个例子测试两次。

 SQL Server Execution Times:

   CPU time = 0 ms,  elapsed time = 0 ms.

Table 'computed_column_test_computed'. Scan count 0, logical reads 100840, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 164, physical reads 2, read-ahead reads 162, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 515 ms,  elapsed time = 1564 ms.

(100000 row(s) affected)

SQL Server parse and compile time:

   CPU time = 0 ms, elapsed time = 1327 ms.

Table 'computed_column_test_noncomputed'. Scan count 0, logical reads 100833, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 164, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 265 ms,  elapsed time = 759 ms.

(100000 row(s) affected)

SQL Server parse and compile time:

   CPU time = 0 ms, elapsed time = 5 ms.

 SQL Server Execution Times:

   CPU time = 0 ms,  elapsed time = 0 ms.

 SQL Server Execution Times:

   CPU time = 0 ms,  elapsed time = 0 ms.

Table 'computed_column_test_computed'. Scan count 0, logical reads 100840, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 164, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 391 ms,  elapsed time = 411 ms.

(100000 row(s) affected)

Table 'computed_column_test_noncomputed'. Scan count 0, logical reads 100833, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 164, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 234 ms,  elapsed time = 273 ms.

(100000 row(s) affected)

 SQL Server Execution Times:

   CPU time = 0 ms,  elapsed time = 0 ms.

SQL Server parse and compile time:

   CPU time = 0 ms, elapsed time = 6 ms.

Table 'computed_column_test_computed'. Scan count 0, logical reads 302521, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 487, physical reads 1, read-ahead reads 330, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 1250 ms,  elapsed time = 1986 ms.

(300000 row(s) affected)

SQL Server parse and compile time:

   CPU time = 0 ms, elapsed time = 74 ms.

Table 'computed_column_test_noncomputed'. Scan count 0, logical reads 302499, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 487, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 813 ms,  elapsed time = 966 ms.

(300000 row(s) affected)

SQL Server parse and compile time:

   CPU time = 0 ms, elapsed time = 14 ms.

Table 'computed_column_test_computed'. Scan count 0, logical reads 302521, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 487, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 1156 ms,  elapsed time = 1709 ms.

(300000 row(s) affected)

Table 'computed_column_test_noncomputed'. Scan count 0, logical reads 302499, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 487, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 734 ms,  elapsed time = 3089 ms.

(300000 row(s) affected)

SQL Server parse and compile time:

   CPU time = 8 ms, elapsed time = 8 ms.

Table 'computed_column_test_computed'. Scan count 0, logical reads 504201, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 809, physical reads 0, read-ahead reads 315, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 2031 ms,  elapsed time = 2080 ms.

(500000 row(s) affected)

Table 'computed_column_test_noncomputed'. Scan count 0, logical reads 504166, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 809, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 1297 ms,  elapsed time = 2915 ms.

(500000 row(s) affected)

SQL Server parse and compile time:

   CPU time = 0 ms, elapsed time = 0 ms.

SQL Server parse and compile time:

   CPU time = 0 ms, elapsed time = 4 ms.

Table 'computed_column_test_computed'. Scan count 0, logical reads 504201, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 809, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 1984 ms,  elapsed time = 3540 ms.

(500000 row(s) affected)

SQL Server parse and compile time:

   CPU time = 0 ms, elapsed time = 2 ms.

Table 'computed_column_test_noncomputed'. Scan count 0, logical reads 504166, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 809, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 1266 ms,  elapsed time = 1341 ms.

(500000 row(s) affected)

SQL Server parse and compile time:

   CPU time = 13 ms, elapsed time = 13 ms.

Table 'computed_column_test_computed'. Scan count 0, logical reads 1008359, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 1615, physical reads 0, read-ahead reads 791, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 4500 ms,  elapsed time = 9110 ms.

(1000000 row(s) affected)

Table 'computed_column_test_noncomputed'. Scan count 0, logical reads 1008333, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 1615, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 2531 ms,  elapsed time = 3903 ms.

(1000000 row(s) affected)

SQL Server parse and compile time:

   CPU time = 0 ms, elapsed time = 0 ms.

SQL Server parse and compile time:

   CPU time = 0 ms, elapsed time = 3 ms.

Table 'computed_column_test_computed'. Scan count 0, logical reads 1008359, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 1615, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 3797 ms,  elapsed time = 6559 ms.

(1000000 row(s) affected)

SQL Server parse and compile time:

   CPU time = 0 ms, elapsed time = 3 ms.

Table 'computed_column_test_noncomputed'. Scan count 0, logical reads 1008333, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 1615, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 2422 ms,  elapsed time = 3895 ms.

(1000000 row(s) affected)

可以看到确实使用computed column会对性能有一定的影响。当计算列数量越多的情况下性能的影响越大。但是当计算列数量很少的情况下，影响或者说差别其实很小很小。以我做的实验为例，讲计算列数量减少到只有3个，数据量依旧停留在100万行的情况，两者的性能差异其实已经很小了。

SQL Server parse and compile time:

   CPU time = 0 ms, elapsed time = 4 ms.

 SQL Server Execution Times:

   CPU time = 0 ms,  elapsed time = 0 ms.

 SQL Server Execution Times:

   CPU time = 0 ms,  elapsed time = 0 ms.

Table 'computed_column_test_computed'. Scan count 0, logical reads 1005813, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 1615, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 3203 ms,  elapsed time = 5058 ms.

(1000000 row(s) affected)

Table 'computed_column_test_noncomputed'. Scan count 0, logical reads 1005319, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 1615, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 2281 ms,  elapsed time = 4747 ms.

(1000000 row(s) affected)

SQL Server parse and compile time:

   CPU time = 16 ms, elapsed time = 28 ms.

Table 'computed_column_test_computed'. Scan count 0, logical reads 1005813, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 1615, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 2984 ms,  elapsed time = 3512 ms.

(1000000 row(s) affected)

Table 'computed_column_test_noncomputed'. Scan count 0, logical reads 1005319, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Table 'Numbers'. Scan count 1, logical reads 1615, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

 SQL Server Execution Times:

   CPU time = 2672 ms,  elapsed time = 2859 ms.

(1000000 row(s) affected)

那么总结下，计算列的使用原则我认为是在表中计算列的数量本身不多，而且一次性数据行插入量不大，计算逻辑固定，计算复杂度大的情况下，推荐使用计算列。

比如像DimDate这种表，表中可能有非常多的属性列用于表示当前日期的一些额外属性，比如下一天的日期，前一天的日期等等。用计算列是一个很好的选择。

参考：

Computed Columns

SQL Server ->> Computed Column（计算列）的更多相关文章

  1. SQL Server中的标识列

    一.标识列的定义以及特点 SQL Server中的标识列又称标识符列,习惯上又叫自增列. 该种列具有以下三种特点: .列的数据类型为不带小数的数值类型 .在进行插入(Insert)操作时,该列的值是由 ...

  2. SQL Server自动化运维系列——关于邮件通知那点事（.Net开发人员的福利）

    需求描述 在我们的生产环境中,大部分情况下需要有自己的运维体制,包括自己健康状态的检测等.如果发生异常,需要提前预警的,通知形式一般为发邮件告知. 邮件作为一种非常便利的预警实现方式,在及时性和易用性 ...

  3. SQL Server自动化运维系列——监控跑批Job运行状态（Power Shell）

    需求描述 在我们的生产环境中,大部分情况下需要有自己的运维体制,包括自己健康状态的检测等.如果发生异常,需要提前预警的,通知形式一般为发邮件告知. 在上一篇文章中已经分析了SQL SERVER中关于邮 ...

  4. SQL SERVER将某一列字段中的某个值替换为其他的值 分类： MSSQL 2014-11-05 13:11 67人阅读 评论(0) 收藏

    SQL SERVER将某一列字段中的某个值替换为其他的值 UPDATE 表名 SET 列名 = REPLACE(列名 ,'贷','袋') SQL SERVER"函数 replace 的参数 ...

  5. SQL Server 2016：内存列存储索引

    作者 Jonathan Allen,译者 谢丽 SQL Server 2016的一项新特性是可以在“内存优化表(Memory Optimized Table)”上添加“列存储索引(Columnstor ...

  6. SQL Server ->> ColumnStore Index（列存储索引）

    Columnstored index是SQL Server 2012后加入的重大特性,数据不再以heap或者B Tree的形式存储(row level)存储在每一个数据库文件的页里面,而是以列为单位存 ...

  7. Sql Server中的标识列(自增长字段)

    一.标识列的定义以及特点 SQL Server中的标识列又称标识符列,习惯上又叫自增列.该种列具有以下三种特点: 1.列的数据类型为不带小数的数值类型2.在进行插入(Insert)操作时,该列的值是由 ...

  8. SQL Server 2012 自动增长列，值跳跃问题

    介绍 从 SQL Server 2012 版本开始, 当SQL Server 实例重启之后,表格的自动增长列的值会发生跳跃,而具体的跳跃值的大小是根据增长列的数据类型而定的.如果数据类型是 整型(in ...

  9. 为SQL Server表中的列添加/修改/删除注释属性（sp_addextendedproperty、sp_updateextendedproperty、sp_dropextendedproperty）

    本篇基本完全参考:sql--sp_addextendedproperty和sp_updateextendedproperty (Transact-SQL) 三个存储过程用法一样,以sp_addexte ...

随机推荐

  1. 人机交互print

    python两种输出类型: 1· 表达式语句,python理解的 #  repr() 2· print() 结果是给用户看的      # str()  >>> "Hell ...

  2. CDH集群安装配置（二）- 公共环境的配置和虚拟机的克隆

    1. 配置网络-ip地址设置静态 vi /etc/sysconfig/network-scripts/ifcfg-eth33 增加如下配置 ONBOOT=yes BOOTPROTO=static IP ...

  3. Dubbo---初识

    1.概述 1.1 Dubbo是一款高性能.轻量级的java RPC框架: 1.2 Dubbo提供的功能: 面向接口的远程调用: 智能容错.负载均衡: 服务注册.发现: 1.3 Dubbo架构 Prov ...

  4. Testlink 机器重启后Access denied for user 'admin '@'localhost' (using password: YES)解决

    问题表现: 装完Testlink,重启系统后,在testlink权限未分配会出现如下提示: 1045 - Access denied for user 'Testlink '@'localhost' ...

  5. List与IList的区别

    在我看一个源程序的时候看到这个例子使用了IList<T>返回类型,因为上午刚刚总结过List<T>的详细用法,突然出现了IList<T>,感觉很奇怪,于是上网搜集了 ...

  6. 浅谈flex布局

    Flex布局,俗称弹性布局,有了这个布局,咱们做的事情很多,以前那些很难实现比如说垂直居中之类都不存在了. 盒模型布局依赖于float,display,定位之类的方式来布局,这种的布局对一些特殊布局来 ...

  7. SSH框架学习步骤

    Hibernate 对象状态 关系映射 SQL语句 缓存抓取 struts action的分发配置 参数传递  文件上传 spring IOC AOP

  8. 【wordpress】wordpress自定义主题

    wordpress每个主题至少要有这两个文件 – style.css 和 index.php. index.php 告诉主题中所有的元素如何布局; style.css 则告诉主题中所有的元素该如何展示 ...

  9. How to Install Eclipse C/C++ Development Tool--转

    http://www3.ntu.edu.sg/home/ehchua/programming/howto/EclipseCpp_HowTo.html Eclipse 4.3 (Kepler) for ...

  10. kd-tree 小结

    核心思想 是一种分割 \(k\) 维数据空间的数据结构 一维情况下就是平衡树,以 \(key\) 为标准判断插入左儿子还是右儿子 \(kdtree\) 就是平衡树在多维空间的扩展 因为有多维,我们按不 ...