自加入可变数据

Question

我有下列数据的表：

Fiscal Year | Fiscal Quarter | Fiscal Week | Data
2009        | 2              | 22          | 9.5
2009        | 2              | 24          | 8.8
2009        | 2              | 26          | 8.8
2009        | 3              | 28          | 8.8
2009        | 3              | 31          | 9.1
2009        | 3              | 33          | 8.8

我想写入可能会产生如下的查询：

Fiscal Year | Fiscal Quarter | Fiscal Week | Data | Trend
2009        | 2              | 22          | 9.5  | NULL
2009        | 2              | 24          | 8.8  | -0.7
2009        | 2              | 26          | 8.8  | 0
2009        | 3              | 28          | 8.8  | 0
2009        | 3              | 31          | 9.1  | 0.3
2009        | 3              | 33          | 8.8  | -0.3 

我知道这可以通过执行表的简单连接到自身与前财政周可以轻松实现，但是这不会永远是一个简单的t1.[Fiscal Week] = t2.[Fiscal Week] - 2因为有时候差为3周。

我可以像这样轻松地将最大记录：

SELECT
    MAX(t1.[Fiscal Week]) "LastWeek"
FROM t1
WHERE t1.[Fiscal Year] = 2009
    AND t1.[Fiscal Week] < 31

但我不知所措，当谈到抽象它使参加工作。

我该怎么办的“最大规模的财政一周比当前记录更小的”？

自连接

Answer 1

和Sql 2005+做到这一点的最简单方法是使用排名及窗口功能 - 只是分区/订单数据和分配适当的顺序给每个记录（在这种情况下，你是通过fiscalYear和排序在该窗口被划分fiscalWeek） - 看起来是这样的：

with data as
(
    select  row_number() over (partition by a.fiscalYear order by a.fiscalWeek) as rn,
            a.fiscalYear, a.fiscalQuarter, a.fiscalWeek, a.Data
    from    #TableName a
)
select  a.fiscalYear, a.fiscalQuarter, a.fiscalWeek, a.Data, 
        a.Data - b.Data as Trend
from    data a
left join data b
on      a.fiscalYear = b.fiscalYear
and     a.rn = b.rn + 1
order by a.fiscalYear, a.rn

这个特定的查询将不允许延伸跨越fiscalYear边界 - 如果你想扩展到越过你只是想“通过分区”条款删除这些年的边界，“fiscalYear”的连接条件，而不是为了通过组合设置

：所述fiscalYear和fiscalWeek，像这样的

with data as
(
    select  row_number() over (order by a.fiscalYear + a.fiscalWeek) as rn,
            a.fiscalYear, a.fiscalQuarter, a.fiscalWeek, a.Data
    from    #TableName a
)
select  a.fiscalYear, a.fiscalQuarter, a.fiscalWeek, a.Data, 
        a.Data - b.Data as Trend
from    data a
left join data b
on      a.rn = b.rn + 1
order by a.rn

Answer 2

工作的呢？该联查询有点拗口，可能有这样做的更好的办法...

select
  [fiscal year]
, [fiscal quarter]
, [fiscal week]
, [data] 
, (
    select top 1 (i.data - t1.data) from t1 as i
    where i.[fiscal year] >= t1.[fiscal year]
    and i.[fiscal quarter] >= t1.[fiscal quarter]
    and i.[fiscal week] >= t1.[fiscal week]
    and (
          i.[fiscal year] <> t1.[fiscal year]
       or i.[fiscal quarter] <> t1.[fiscal quarter]
       or i.[fiscal week] <> t1.[fiscal week]
    )
    order by [fiscal year] asc, [fiscal quarter] asc, [fiscal week] asc
) as trend
from t1

Answer 3

如果您使用的是SQL Server 2005中，你可以使用CROSS APPLY条款。

在这里，你可以有一个表值函数，该函数将返回该行本财一周，一个当前行之前。

CREATE FUNCTION dbo.fn_GetFiscalWeekBeforeThis(@Year AS int, 
  @CurrentWeek as int)
  RETURNS TABLE
AS
RETURN
  SELECT TOP 1 *
  FROM t1
  WHERE [Fiscal Year] = @Year 
  AND [Fiscal Week] < @CurrentWeek
  ORDER BY [Fiscal Week] DESC
GO

然后，

SELECT A.*,
A.Data - B.Data
FROM
t1 A
CROSS APPLY 
   dbo.fn_GetFiscalWeekBeforeThis(A.[Fiscal Year],  
   A.[Fiscal Week]) AS B

编辑：请注意，我已经看的文章和没有IDE适应的SQL结果。 另外，我不知道 - 怎么会在第一行的工作

所以，请类型：）

EDIT2：让我知道，如果它不工作在所有结果。 这将帮助我知道 - 不回答的东西，而不检查结果

EDIT3：我已删除用于从所述功能区参数的需要

Answer 4

DECLARE @Fiscal TABLE
  ( 
   FiscalYear int
  ,FiscalQuarter int
  ,FiscalWeek int
  ,[Data] decimal(4,1) 
  )

INSERT INTO @Fiscal
          ( FiscalYear, FiscalQuarter, FiscalWeek, [Data] )
SELECT 2009, 2, 22, 9.5 UNION
SELECT 2009, 2, 24, 8.8 UNION
SELECT 2009, 2, 26, 8.8 UNION
SELECT 2009, 3, 28, 8.8 UNION
SELECT 2009, 3, 31, 9.1 UNION
SELECT 2009, 3, 33, 8.8 UNION
SELECT 2010, 1, 1, 9.0 UNION
SELECT 2010, 1, 2, 9.2

;
WITH  abcd
        AS ( SELECT FiscalYear
                   ,FiscalQuarter
                   ,FiscalWeek
                   ,[Data]
                   ,row_number() OVER ( ORDER BY FiscalYear, FiscalWeek ) AS rn
             FROM   @Fiscal
           )
  SELECT  a.FiscalYear
         ,a.FiscalQuarter
         ,a.FiscalWeek
         ,a.[Data]
         ,a.[Data] - b.[Data] AS [Trend]
  FROM    abcd AS a
          LEFT JOIN abcd AS b ON b.rn = ( a.rn - 1 )
  ORDER BY a.FiscalYear
         ,a.FiscalWeek