原文:SQL Server表分区的NULL值问题

SQL Server表分区的NULL值问题

SQL Server表分区只支持range分区这一种类型，往往会被大家吐槽

人家MySQL支持四种类型：RANGE分区、LIST分区、HASH分区、KEY分区

共同点是MySQL跟SQL Server也有分区对齐的问题，都是水平切分，大家都允许分区列存在NULL值

这次我们测试SQL Server表分区的分区列的NULL值，究竟NULL值是被存放在哪个区间，以前一直没有注意

测试脚本

--1.创建文件组
ALTER DATABASE [sss]
ADD FILEGROUP [FG_TinyBlog_Id_01]

ALTER DATABASE [sss]
ADD FILEGROUP [FG_TinyBlog_Id_02]

ALTER DATABASE [sss]
ADD FILEGROUP [FG_TinyBlog_Id_03]

ALTER DATABASE [sss]
ADD FILEGROUP [FG_TinyBlog_Id_04]

ALTER DATABASE [sss]
ADD FILEGROUP [FG_TinyBlog_Id_05]

--2.创建文件
ALTER DATABASE [sss]
ADD FILE
(NAME = N'FG_TinyBlog_Id_01_data',FILENAME = N'E:\DataBase\sss\FG_TinyBlog_Id_01_data.ndf',SIZE = 96MB, FILEGROWTH = 24MB )
TO FILEGROUP [FG_TinyBlog_Id_01];

ALTER DATABASE [sss]
ADD FILE
(NAME = N'FG_TinyBlog_Id_02_data',FILENAME = N'E:\DataBase\sss\FG_TinyBlog_Id_02_data.ndf',SIZE = 96MB, FILEGROWTH = 24MB )
TO FILEGROUP [FG_TinyBlog_Id_02];

ALTER DATABASE [sss]
ADD FILE
(NAME = N'FG_TinyBlog_Id_03_data',FILENAME = N'E:\DataBase\sss\FG_TinyBlog_Id_03_data.ndf',SIZE = 96MB, FILEGROWTH = 24MB )
TO FILEGROUP [FG_TinyBlog_Id_03];

ALTER DATABASE [sss]
ADD FILE
(NAME = N'FG_TinyBlog_Id_04_data',FILENAME = N'E:\DataBase\sss\FG_TinyBlog_Id_04_data.ndf',SIZE = 96MB, FILEGROWTH = 24MB )
TO FILEGROUP [FG_TinyBlog_Id_04];

ALTER DATABASE [sss]
ADD FILE
(NAME = N'FG_TinyBlog_Id_05_data',FILENAME = N'E:\DataBase\sss\FG_TinyBlog_Id_05_data.ndf',SIZE = 96MB, FILEGROWTH = 24MB )
TO FILEGROUP [FG_TinyBlog_Id_05];

--3.创建分区函数
CREATE PARTITION FUNCTION
Fun_TinyBlog_Id(INT) AS
RANGE LEFT
FOR VALUES(-10,0,1,6)

--4.创建分区方案
CREATE PARTITION SCHEME
[Sch_TinyBlog_Id] AS
PARTITION [Fun_TinyBlog_Id]
TO([FG_TinyBlog_Id_01],[FG_TinyBlog_Id_02],[FG_TinyBlog_Id_03],[FG_TinyBlog_Id_04],[FG_TinyBlog_Id_05])

View Code

插入测试数据

USE [sss]

CREATE TABLE TinyBlog(id INT  NULL,NAME VARCHAR(100))
 ON [Sch_TinyBlog_Id](id)

SELECT * FROM TinyBlog ORDER BY id 

INSERT INTO [dbo].[TinyBlog]
        ( [id], [NAME] )
VALUES  ( NULL, -- id - int
          '3232'  -- NAME - varchar(100)
          )

INSERT INTO [dbo].[TinyBlog]
        ( [id], [NAME] )
VALUES  ( -2, -- id - int
          '-2'  -- NAME - varchar(100)
          )

INSERT INTO [dbo].[TinyBlog]
        ( [id], [NAME] )
VALUES  ( 66, -- id - int
          '66'  -- NAME - varchar(100)
          )

INSERT INTO [dbo].[TinyBlog]
        ( [id], [NAME] )
VALUES  ( 0, -- id - int
          '0'  -- NAME - varchar(100)
          )

INSERT INTO [dbo].[TinyBlog]
        ( [id], [NAME] )
VALUES  ( -30, -- id - int
          '-30'  -- NAME - varchar(100)
          )

表数据如下

SELECT * FROM TinyBlog ORDER BY id 

分区分布

--查看分区架构文件组分布
SELECT  CONVERT(VARCHAR(MAX), ps.name) AS partition_scheme ,
        p.partition_number ,
        CONVERT(VARCHAR(MAX), ds2.name) AS filegroup ,
        CONVERT(VARCHAR(MAX), ISNULL(v.value, ''), 120) AS range_boundary ,
        STR(p.rows, 9) AS rows
FROM    sys.indexes i
        JOIN sys.partition_schemes ps ON i.data_space_id = ps.data_space_id
        JOIN sys.destination_data_spaces dds ON ps.data_space_id = dds.partition_scheme_id
        JOIN sys.data_spaces ds2 ON dds.data_space_id = ds2.data_space_id
        JOIN sys.partitions p ON dds.destination_id = p.partition_number
                                 AND p.object_id = i.object_id
                                 AND p.index_id = i.index_id
        JOIN sys.partition_functions pf ON ps.function_id = pf.function_id
        LEFT JOIN sys.Partition_Range_values v ON pf.function_id = v.function_id
                                                  AND v.boundary_id = p.partition_number
                                                  - pf.boundary_value_on_right
WHERE   i.object_id = OBJECT_ID('TinyBlog')
        AND i.index_id IN ( 0, 1 )
ORDER BY p.partition_number

--分区区间
--SELECT  *  FROM    sys.partition_range_values

View Code

分区情况

--分区情况
SELECT  c.* ,
        b.[groupname] AS '分区方案对应的文件组名称' ,
        d.name '当前分区函数对应的分区方案'
FROM    sys.destination_data_spaces AS a
        INNER JOIN sysfilegroups AS b ON a.[data_space_id] = b.[groupid]
        INNER JOIN ( SELECT $PARTITION.Fun_TinyBlog_Id(id) AS 分区编号 ,
                            MIN(id) AS Min_value ,
                            MAX(id) AS Max_value ,
                            COUNT(id) AS 记录数
                     FROM   [dbo].[TinyBlog]
                     GROUP BY $PARTITION.Fun_TinyBlog_Id(id)
                   ) AS c ON c.[分区编号] = a.[destination_id]
        INNER JOIN sys.partition_schemes AS d ON a.[partition_scheme_id] = d.data_space_id
ORDER BY c.[分区编号]

View Code

第二个视图直接把NULL值忽略了

根据第一个视图，我们画出一个总结图

分析一下

结论其实很清晰，分区列不要允许NULL，如果允许列，那么我们交换分区归档数据的时候就会有麻烦了，因为无法判断分区列是NULL值的那些数据行是老数据还是新数据

相关文章：

http://www.cnblogs.com/gaizai/archive/2010/11/05/1870071.html

https://msdn.microsoft.com/zh-cn/library/ms187802.aspx

如有不对的地方，欢迎大家拍砖o(∩_∩)o