Question

假设我有一个表格，其中价格和日期为列，如：

Price       Date
3.20000000  2018-01-14
3.20000000  2018-01-18
3.20000000  2018-01-19
4.00000000  2018-02-10
4.00000000  2018-02-11
4.00000000  2018-02-12
5.10000000  2018-02-16
3.20000000  2018-03-11
4.00000000  2018-03-15

我希望每次出现新价格时都获得第一个日期，并希望在同一日期列中获得该价格的最后日期（加上一天）。请注意，价格可以上下波动，同一价格在整个表格中可能出现多次。因此上表的结果应为：

Price       Date
3.20000000  2018-01-14
3.20000000  2018-01-20 -- 19 + 1
4.00000000  2018-02-10
4.00000000  2018-02-13 -- 12 + 1
5.10000000  2018-02-16
5.10000000  2018-02-17 -- 16 + 1
3.20000000  2018-03-11
3.20000000  2018-03-12 -- 11 + 1
4.00000000  2018-03-15
4.00000000  2018-03-16 -- 15 + 1

我希望使用自联接，是否可以在没有row_number函数的情况下进行自联接？

Answer 1

这是一个变体。您可以对其进行修改以适合您的需求。

DECLARE @DataSource TABLE
(
    [Price] SMALLMONEY
   ,[Date] DATE
);

INSERT INTO @DataSource ([Price], [Date])
VALUES ('3.20000000', '2018-01-14')
      ,('3.20000000', '2018-01-18')
      ,('3.20000000', '2018-01-19')
      ,('4.00000000', '2018-02-10')
      ,('4.00000000', '2018-02-11')
      ,('4.00000000', '2018-02-12')
      ,('5.10000000', '2018-02-16')
      ,('3.20000000', '2018-03-11')
      ,('4.00000000', '2018-03-15');

-- we need to order the data, if you do not want to use ROW_NUMBER(), use another way to order it (for exmaple record ID of each row using IDENTITY column)
WITH DataSourceOrdered AS
(
    SELECT ROW_NUMBER() OVER(ORDER BY [Date]) AS [RowID]
         ,[Price]
         ,[Date]
    FROM @DataSource
), RecordsWhenPriceChanged AS
(
    -- records when price has changed
    SELECT DS1.[Price]
          ,DS1.[Date]
          ,DS1.[RowID]
    FROM DataSourceOrdered DS1
    INNER JOIN DataSourceOrdered DS2
        ON DS1.[RowID] -1 = DS2.[RowID]
        AND DS1.[Price] <> DS2.[Price]
    UNION ALL
    -- the initial record
    SELECT [Price]
          ,[Date]
          ,[RowID]
    FROM DataSourceOrdered
    WHERE [RowID] = 1
)
-- getting current records
SELECT [Price]
      ,[Date]
      ,[RowID]
FROM RecordsWhenPriceChanged
UNION ALL
-- getting previous records + 1 day for each
SELECT DS1.[Price]
      ,DATEADD(DAY, 1, DS1.[Date])
      ,DS1.[RowID]
FROM DataSourceOrdered DS1
INNER JOIN RecordsWhenPriceChanged DS2
    ON DS1.[RowID] + 1 = DS2.[RowID]
-- getting the last record
UNION ALL
SELECT *
FROM
(
    SELECT TOP 1 [Price]
                ,DATEADD(DAY, 1,[Date]) AS [Date]
                ,[RowID]
    FROM RecordsWhenPriceChanged
    ORDER BY [Date] DESC
) DS
ORDER BY [Date];

Answer 2

正如我在评论中所写，这是一个典型的空白和孤岛问题，解决此问题的简单方法是两次使用row_number。

首先，创建并填充示例表（请在您将来的问题中为我们保存此步骤）：

DECLARE @T AS TABLE
(
    Price decimal(15,5),
    [Date] Date
)

INSERT INTO @T (Price, [Date]) VALUES
(3.20000000, '2018-01-14'),
(3.20000000, '2018-01-18'),
(3.20000000, '2018-01-19'),
(4.00000000, '2018-02-10'),
(4.00000000, '2018-02-11'),
(4.00000000, '2018-02-12'),
(5.10000000, '2018-02-16'),
(3.20000000, '2018-03-11'),
(4.00000000, '2018-03-15')

现在，使用公共表格表达式标记岛：

;WITH CTE AS
(
SELECT  Price, 
        [Date], 
        ROW_NUMBER() OVER(ORDER BY [Date]) - 
        ROW_NUMBER() OVER(PARTITION BY Price ORDER BY [Date]) As Island
FROM @T
)

然后使用联合查询进行查询-一个使用min(date) over(partition by island)，第二个使用max(dateadd(day, 1, date)) over(partition by island)，按日期对结果进行排序：

SELECT Price, MIN(Date) OVER (PARTITION BY Island) As [Date]
FROM CTE
UNION
SELECT Price, DATEADD(DAY, 1, MAX(Date)  OVER (PARTITION BY Island))
FROM CTE
ORDER BY Date

结果：

Price       Date
3,20000     14.01.2018
3,20000     20.01.2018
4,00000     10.02.2018
4,00000     13.02.2018
5,10000     16.02.2018
5,10000     17.02.2018
3,20000     11.03.2018
3,20000     12.03.2018
4,00000     15.03.2018
4,00000     16.03.2018

You can see a live demo on rextester.

Answer 3

您可以尝试不使用row_number的此解决方案。

DECLARE @MyTable TABLE ([Price] DECIMAL(18,8), [Date] DATE);

INSERT INTO @MyTable ([Price], [Date])VALUES 
('3.20000000', '2018-01-14'),
('3.20000000', '2018-01-18'),
('3.20000000', '2018-01-19'),
('4.00000000', '2018-02-10'),
('4.00000000', '2018-02-11'),
('4.00000000', '2018-02-12'),
('5.10000000', '2018-02-16'),
('3.20000000', '2018-03-11'),
('4.00000000', '2018-03-15')

;WITH CTE AS 
(
    SELECT * FROM @MyTable 
    UNION ALL
    SELECT Price, DATEADD(DAY,1,[Date]) FROM @MyTable 
)
, CTE2 AS (
    SELECT T1.Price, MIN(T1.[Date]) MinDate, MAX(T1.[Date]) MaxDate
    FROM CTE T1
        OUTER APPLY( SELECT COUNT(DISTINCT Price) DCNT, MIN(Date) [Date] FROM CTE T2 WHERE T2.Date >= T1.Date ) TA1 
    GROUP BY T1.Price, TA1.DCNT
)
SELECT Price, [Date] FROM CTE2 UNPIVOT ( [Date] FOR Col IN (MinDate, MaxDate)) UNPVT
ORDER BY [Date]

结果：

Price                                   Date
--------------------------------------- ----------
3.20000000                              2018-01-14
3.20000000                              2018-01-20
4.00000000                              2018-02-10
4.00000000                              2018-02-13
5.10000000                              2018-02-16
5.10000000                              2018-02-17
3.20000000                              2018-03-11
3.20000000                              2018-03-12
4.00000000                              2018-03-15
4.00000000                              2018-03-16

每次在同一日期栏中更改价格后，如何选择首个日期和最后一个日期（加上一天）？

3 个答案: