Question

我希望找到一种从日期范围中找到可能连续或不连续的日期的最佳方法（我试图避免使用光标，或者尽可能避免使用繁重的功能）。

假设我有酒店客人来来往往（办理登机手续，退房）。我想找一个某位客人和我们一起度过 45th 晚上的日期。我们使用的数据库记录数据：

Create Table #GuestLog(
    ClientId int, 
    StartDate DateTime, 
    EndDate DateTime)

以下是一些数据

Insert Into #GuestLog Values(1, '01/01/2010', '01/10/2010')
Insert Into #GuestLog Values(1, '01/16/2010', '01/29/2010')
Insert Into #GuestLog Values(1, '02/13/2010', '02/26/2010')
Insert Into #GuestLog Values(1, '04/05/2010', '06/01/2010')
Insert Into #GuestLog Values(1, '07/01/2010', '07/21/2010')

到目前为止，我只能想到涉及临时表功能的解决方案以及类似的疯狂东西，我觉得我已经过度思考了。

提前谢谢。

编辑： @Andriy M解决方案的轻微模式。

DECLARE @ClientID int, @NightNo int;
SET @ClientID = 1;
SET @NightNo = 45;

SELECT *
FROM ( SELECT  gl.ClientId
        , Date = gl.StartDate + v.number - 1
        , rownum = ROW_NUMBER() OVER ( PARTITION BY gl.ClientId ORDER BY gl.StartDate, v.Number)
    FROM #GuestLog gl
    INNER JOIN master..spt_values v ON v.type = 'P'
    AND v.number BETWEEN 1  AND gl.EndDate - gl.StartDate + 1) as s //--added "+ 1"
WHERE ClientId = @ClientId
AND rownum = @NightNo

Answer 1

按照Jeremy Pridemore的好榜样，我也参考了我的解决方案（为什么不呢？）。

一个注意事项：因为你已经说过“第45天晚上”，我认为这意味着应该采取之前之前的日期。如果我在那里错了，那么只需删除计算- 1的{{1}}部分。

Date

Answer 2

试试这个（我讨厌对列使用内联查询，但却想不到任何其他路径。）：

WITH logd 
     AS (SELECT a.*, 
                (SELECT SUM(Datediff(d, startdate, enddate)) 
                 FROM   #guestlog b 
                 WHERE  b.clientid = a.clientid 
                        AND b.startdate <= a.startdate) dayssofar 
         FROM   #guestlog a) 
SELECT a.*, 
       Dateadd(d, ( 45 - dayssofar ), enddate) 
FROM   (SELECT b.*, 
               Row_number() OVER(PARTITION BY clientid ORDER BY dayssofar)rn 
        FROM   logd b 
        WHERE  dayssofar > 44) a 
WHERE  rn = 1

Answer 3

我无法在我坐的地方测试我的代码，但以下情况应该有效：

首先，生成tally table（如果可以的话，将其设为永久物）。

然后使用它来平整日期范围，如下所示：

SELECT DATEADD(d,n.number,'01/01/2000') AS StayedDate
FROM numbers n
INNER JOIN #GuestLog g ON DATEADD(d,n.number,'01/01/2000') BETWEEN g.StartDate AND g.EndDate)
ORDER BY n.number

然后添加一个带有ROW_NUMBER（）的CTE来访问第45行。

如果您经常使用这些查询类型，请创建一个额外的日期表（就像数字表，但有日期），以摆脱丑陋的DATEADD。

Answer 4

我在具有90兼容性的数据库（SQL Server 2005）上测试了我的SQL Server 2008 R2中的解决方案，所以我相信这会做你想要的：

-- PLEASE NOTE: MAXRECURSION at the bottom needs to have a number that is higher than the
--  number of stored stays that any guest this will run on will have. Otherwise you'll need
--  to find a way to do this without recursion.

-- Parameterized because...why not? :)
DECLARE @CustomerID INT = 1
    , @NthStayDay INT = 45;

-- This does nothing but get the rows out of GuestLog that we care about. From my experience
--  it's a good idea to do a simple data grab from a physical table or indexed view using
--  a seek, then play with that smaller subset of data in other CTE's. Though I'm sure that
--  those with more performance knowledge could give better answers. RowNumber is added for
--  recursion in the next CTE.
WITH OrderedStays(RowNumber, StartDate, EndDate) AS
(
    SELECT
        ROW_NUMBER() OVER(ORDER BY StartDate) AS RowNumber
        , StartDate
        , EndDate
    FROM @GuestLog GuestLog
    WHERE GuestLog.ClientId = @CustomerID
)
-- This is a recusive CTE, but I don't imagine it will preform to badly because there is no IO
--  at this point, simply processing the previous CTE. You'll have to be the judge of that.
--  The purpose of this CTE is to be able to limit down to the start date that we care about.
, StayRanges(RowNumber, StartDate, EndDate, FirstDayCount, LastDayCount) AS
(
    -- This is our anchor row. It is the first date range at which the guest stayed with you.
    --  The DATEDIFF returns 9 with dates of 20100101 - 20100110, but since I don't think the 
    --  0th day stayed makes sense, I'm making it return 10 in that case since we're starting
    --  at 1.
    SELECT
        RowNumber
        , StartDate
        , EndDate
        , 1 AS FirstDayCount
        , DATEDIFF(DAY, StartDate, EndDate) + 1 AS LastDayCount
    FROM OrderedStays
    WHERE RowNumber = 1

    UNION ALL

    -- This is the recursion. This joins the first CTE on what we have where the first CTE's
    --  RowNumber is 1 more than whatever is in our StayRanges CTE. The column logic is the
    --  same as above, but now we need to add in the LastDayCount from our previous iteration.
    SELECT
        OrderedStays.RowNumber
        , OrderedStays.StartDate
        , OrderedStays.EndDate
        , StayRanges.LastDayCount + 1 AS FirstDayCount
        , DATEDIFF(DAY, OrderedStays.StartDate, OrderedStays.EndDate) + StayRanges.LastDayCount + 1 AS LastDayCount
    FROM OrderedStays
    INNER JOIN StayRanges
        ON (StayRanges.RowNumber + 1) = OrderedStays.RowNumber
)
-- Now that we have our ranges, we can select the range that has the day we want in it with a
--  simple between. Once that's done, take out the FirstDayCount from the day we care about so
--  that you're left with the difference from the StartDate and the date we want, and add that
--  to the StartDate. Done!
SELECT
    DATEADD(DAY, @NthStayDay - FirstDayCount, StartDate) AS DateOfNthStayDate
FROM StayRanges
WHERE @NthStayDay BETWEEN FirstDayCount AND LastDayCount
OPTION(MAXRECURSION 5000)

从非连续范围中查找日期

4 个答案: