那些可以在数据库里做的事:分页与过滤

其实很多非业务逻辑的功能,比如分页,数据过滤,可以在程序上面节省很多内存和CPU时间,但往往又找不到一个比较通用有效的方法,花了点时间,终于把我想要的在数据库中分页和过滤的功能写了出来,在这里分享。

第一期望:我希望输入页码(pageIndex),每页显示的记录数(pageSize),然后这个存储过程就可以给我一个当前页的数据集。

首先,我假设有个表叫Configuration,里面就3个字段,Id, Key, Value,就是一个简单的数据表,定义如下:

CREATE TABLE [dbo].[Configuration]
(
    [Id] int NOT NULL IDENTITY (1,1), 
    [Key] varchar(20) NOT NULL UNIQUE,
    [Value] nvarchar(max) NULL
)

然后,就可以写存储过程了,这里想要说下,我对存储过程命名的一些习惯,我会以"sp_"开头,为了标明这个存储过程是CRUD中的哪个操作,我会再附加一个缩写,比如这个分页的主要操作时READ,所以我的名字应该是"sp_r_",然后为了说明它的主要功能是分页,就有了"sp_r_p_",“p”表示paginate,最后跟上操作的主要的表名,"sp_r_p_configuration"。很多同学都喜欢用比如bypage等等,也可以,但是我更喜欢这种命名法,在特定的约定下,这种简洁性的命名可可读性。
再者,一个良好的代码书写规范是非常必要的,该注释的地方千万别省,在多数情况下,代码的可读性是非常重要的,除非因性能因素做必须得让步。

/*
PROCEDURE:    [dbo].[sp_r_p_configuration]
OPERATION:    READ
FEATURES:        Paginate
PURPOSE:        provide a paginated list from the configuration table
CREATOR:        Jerry Weng
CREATETIME:    2014-4-9
UPDATETIME:    2014-4-9
VERSION:        1.0
*/

CREATE PROCEDURE [dbo].[sp_r_p_configuration]
    @pageIndex int = 1,                    /*which page of the total page*/
    @pageSize int=10,                    /*how many records show in one page*/
    @recordCount int OUTPUT,        /*output: return the count of the total records in all pages*/
    @pageCount int OUTPUT            /*output: return the count of the total pages*/
AS
    declare @startRow int;                -- the row number of the first record in the page
    declare @endRow int;                -- the row number of the last record in the page
    set @startRow = (@pageIndex - 1) * @pageSize + 1;
    set @endRow= @startRow + @pageSize - 1;

    ;with tmp as 
    (select (ROW_NUMBER() over(order by Id)) as row_id, [Id], [Key], [Value] 
    from dbo.Configuration)
    select [Id], [Key], [Value] from tmp
    where row_id between @startRow and @endRow
    
    -- calculate the record count in all pages
    select @recordCount=COUNT(*) from dbo.Configuration
    -- calculate the page count
    set @pageCount = CEILING(@recordCount/CAST(@pageSize as float))

RETURN 0

这里用的ROW_NUMBER()来实现分页,性能必拼SQL的方法好,WITH...AS可以省下一个表变量。最后计算总记录数和页数很简单,算下就好了。

第二期望:这样的,似乎差不多了,但是我还希望能够有一套规则,来过滤结果集,并且还要有类似AND和OR的运算功能,而且我仍旧不希望用inline-script去实现,能不用尽量不要用。

为了实现这个期望,我需要先定义怎么传入过滤规则,我最终决定用的是xml数据类型而不是普通字符串,因为普通字符串有长度限制(如果你说可以用text,那我只能说太奢侈,小弟用不起),定短了可能被截断出现错误输出,定长了,最多也就8000个字符,也浪费,我不喜欢不确定因素,然而xml不一样,既是结构化数据,又有索引支持,可长可短,客户端也可直接序列化成字符串传入,非常方便。于是剩下的是,约定一个规则,确定过滤规则的数据结构,在这里我专门写了一个function来将xml转换成table已被后用。P.S. 如果想传一个数组到存储过程,也可以用xml类型,比字符串加分隔符更安全。

/*
FUNCTION:        [dbo].[fun_xmlfilter_parse]
OPERATION:    READ
FEATURES:        
PURPOSE:        parse a table from a xml with the specific column name
CREATOR:        Jerry Weng
CREATETIME:    2014-4-9
UPDATETIME:    2014-4-9
VERSION:        1.0
*/

CREATE FUNCTION [dbo].[fun_xmlfilter_parse]
(
    @filter xml,                    /*    The format should be:
                                        <filter>
                                            <rule 
                                                column="COLUMN_NAME"    -- the column name which the rule should be applied
                                                match="xxx"                    -- the key which the rule defines to filter and the result should match
                                                required="0|1"                    -- if 1, the result collection has to meet this rule or return nothing
                                             />
                                             <rule ....
                                        </filter>
                                    */
    @column varchar(100)
)
RETURNS @returntable TABLE 
(
    [Match] nvarchar(100),
    [Required] bit
)
AS
BEGIN
    ;with tmp as
    (
    select
            [T].[RULE].value('@column','varchar(100)') as [Column],
            [T].[RULE].value('@match','nvarchar(100)') as [Match],
            [T].[RULE].value('@required','bit') as [Required]
    from @filter.nodes('filter//rule') [T]([RULE])
    )
    INSERT @returntable
    select [Match],[Required] from tmp where [Column] = @column
        
    RETURN 
END

xml的结构由一个根节点<filter>开始,里面是多个<rule>节点,具体的规则定义在rule的属性列表中。column表示具体要过滤哪列的数据,match表示要匹配的数据值,比如我想要过滤列"Value"下,值为"sample"的数据,这样的话我只需要把xml写成<filter><rule column="value" match="sample" required="1" /></filter>,最后的哪个required就是来实现AND和OR的操作的,如果是1表示必须匹配,类似AND,0的话可选匹配,类似OR,输出的表示由COLUMN的值过滤过的,因为比较是以COLUMN为维度的。

最后,新建一个存储过程,叫"sp_r_fp_configuration",这里比上一个多了一个"f",表示带有过滤功能。

 

/*
PROCEDURE:    [dbo].[sp_r_fp_configuration]
OPERATION:    READ
FEATURES:        Paginate, Filter
PURPOSE:        provide a paginated & filterable list from the configuration table
CREATOR:        Jerry Weng
CREATETIME:    2014-4-9
UPDATETIME:    2014-4-9
VERSION:        1.0
*/

CREATE PROCEDURE [dbo].[sp_r_fp_configuration]
    @pageIndex    int = 1,                /*which page of the total page*/
    @pageSize        int=10,                /*how many records show in one page*/
    @filter            xml,                    /*filter string for the records*/
    @recordCount int OUTPUT,        /*output: return the count of the total records in all pages*/
    @pageCount    int OUTPUT            /*output: return the count of the total pages*/
AS
    if(@filter is null)
    begin
        exec dbo.sp_r_p_configuration 
                @pageIndex,
                @pageSize,
                @pageCount=@pageCount OUTPUT, 
                @recordCount=@recordCount OUTPUT
    end
    else
    begin
        
        declare @tmptbl table (row_id int, Id int, [Key] varchar(100), [Value] varchar(200))
    
        declare @startRow int;                -- the row number of the first record in the page
        declare @endRow int;                -- the row number of the last record in the page
        set @startRow = (@pageIndex - 1) * @pageSize + 1;
        set @endRow= @startRow + @pageSize - 1;

        insert into @tmptbl
            select (ROW_NUMBER() over (order by m.Id)) as row_id, m.[Id], m.[Key], m.[Value] 
            from dbo.Configuration as m
            outer apply fun_xmlfilter_parse(@filter,'key') as k
            outer apply fun_xmlfilter_parse(@filter,'value') as v
            where 
            (
                not (isnull(k.[Required],0)=0 and isnull(v.[Required],0)=0) and
                (isnull(k.[Required],0)=0 or (k.[Required] = 1 and isnull(m.[Key],'') = isnull(k.[Match],''))) and
                (isnull(v.[Required],0)=0 or (v.[Required] = 1 and isnull(m.[Value],'') = isnull(v.[Match],'')))
            ) or
            k.[Required] = 0 and isnull(m.[Key],'') = isnull(k.[Match],'') or
            v.[Required] = 0 and isnull(m.[Value],'') = isnull(v.[Match],'')
        
        select [Id], [Key], [Value] from @tmptbl
        where row_id between @startRow and @endRow
    
        -- calculate the record count in all pages
        select @recordCount=COUNT(*) from @tmptbl
        -- calculate the page count
        set @pageCount = CEILING(@recordCount/CAST(@pageSize as float))
    end
RETURN 0

这里用了一个表变量,注意表变量是可被缓存的,临时表没有哦,用表变量不用with...as是因为后面在计算过滤后的总记录数的时候,还需用一次过滤后的集合。这里我希望列Key和列Value都可以被过滤,于是用了两个outer apply来将刚才那个函数的输出表附加到每行上,在where比较的地方,先比较Required是1的必要条件,当然如果针对某个column没有rule的话,也就是required is null的情况,是需要排除的,最后再附加连个required=0的可选条件,这样的话,一个select就可以把各种条件过滤出来了。不过,有个问题是,如果需要过滤的列比较多,那比较语句也要一条条加上去。

最后,来测试下:

我给表填充了这么些数据:

SET IDENTITY_INSERT [dbo].[Configuration] ON
INSERT [dbo].[Configuration] ([Id], [Key], [Value]) VALUES (1, N'test1', N'123')
INSERT [dbo].[Configuration] ([Id], [Key], [Value]) VALUES (2, N'test2', N'123')
INSERT [dbo].[Configuration] ([Id], [Key], [Value]) VALUES (3, N'test3', N'567')
INSERT [dbo].[Configuration] ([Id], [Key], [Value]) VALUES (4, N'test4', NULL)
INSERT [dbo].[Configuration] ([Id], [Key], [Value]) VALUES (5, N'test5', NULL)
INSERT [dbo].[Configuration] ([Id], [Key], [Value]) VALUES (6, N'test6', NULL)
INSERT [dbo].[Configuration] ([Id], [Key], [Value]) VALUES (7, N'test7', NULL)
INSERT [dbo].[Configuration] ([Id], [Key], [Value]) VALUES (8, N'test8', NULL)
INSERT [dbo].[Configuration] ([Id], [Key], [Value]) VALUES (9, N'test9', NULL)
INSERT [dbo].[Configuration] ([Id], [Key], [Value]) VALUES (10, N'test10', NULL)
INSERT [dbo].[Configuration] ([Id], [Key], [Value]) VALUES (11, N'test11', NULL)
SET IDENTITY_INSERT [dbo].[Configuration] OFF

测试脚本:

declare @pageIndex int; set @pageIndex=1;
declare @pageSize int;    set @pageSize=5;
declare @filter xml; 
set @filter='<filter><rule column="key" match="test3" required="1"/><rule column="value" match="123" required="0"/></filter>'
declare @pageCount int;
declare @recordCount int;

exec dbo.sp_r_fp_configuration @pageIndex,@pageSize,@filter,@pageCount=@pageCount OUTPUT, @recordCount=@recordCount OUTPUT
select @pageCount as [PAGECOUNT], @recordCount as [RECORDCOUNT]

输出,如果需要找null值的话,直接把match属性去掉就可以了<rule column="value" required="1"/>:

 

posted @ 2014-04-09 16:38  万力王  阅读(4451)  评论(40编辑  收藏  举报