为SQL Server 2012

时间：2017-03-23 10:08:36

标签： sql sql-server tsql sql-server-2012

背景：

我正在尝试确定客户在特定房间内花费的时间长度。每个客户都可以通过CustomerID识别，当他们访问时，会为他们分配VisitNumber。例如，如果客户今天访问，他们会得到VisitNumber说111111.然后他们会在下周离开并返回，并且会有VisitNumber 111112。

当客户第一次访问时，他们最初没有被分配一个房间，并且当他们最终被分配到他们的指定房间时，一个条目被写入数据库。 CurrentRoom会因为他们还没有房间而空白，而NewRoom是他们搬入的房间。

此条目将被记录为事件1（客户从无房间移动到房间），时间是交易发生的时间。如果客户将来在现有住宿期间被移动，将被记录为事件9（客户从一个房间移动到另一个房间），以及CurrentRoom＆amp; * NewRoom值也将被记录。

问题

我已设法使用 LAG 和 LEAD <从上一行和下一行的时间中获取时间/ strong>然后计算两次之间的差异，这给了我客户在那个房间里度过的时间。

使用LAG时的问题是获取之前的值，在某些情况下可能是来自完全不同的客户的值。我想得到 LAG ＆amp; LEAD 值仅适用于特定CustomerID＆amp;当前VisitNumber然后计算出值之间的差异，以找出该客户在一个房间内花了多长时间。

演示数据： 

CREATE TABLE #beds
(
    [id] [int] IDENTITY(1,1) NOT NULL PRIMARY KEY,
     [User] [nvarchar](50) NULL,
    [CustomerID] [nvarchar](50) NULL,
     [Area] [nchar](10) NULL,
    [Event] [nvarchar](50) NULL,
    [VisitNumber] [nvarchar](50) NULL,
    [Time] [datetime] NULL,
     [CurrentRoom] [nvarchar](50) NULL,
     [NewRoom] [nvarchar](50) NULL
)
GO

INSERT INTO #beds ([User],[CustomerID],[Area],[Event],[VisitNumber],[Time],[CurrentRoom],[NewRoom])
VALUES ('00001','C11111111','Area1',2,111111111,'2017-03-22 11:05:44.360','B22','B44'),
('00001','C11111111','Area1',1,111111111,'2017-03-22 11:05:15.517','','B22'),
('00001','C22222222','Area2',1,222222222,'2017-03-22 07:38:16.117','','POD3'),
('00001','C22222222','Area2',3,222222222,'2017-03-22 07:41:24.787','POD3','POD3'),
('00001','C22222222','Area2',9,222222222,'2017-03-22 09:10:49.697','POD3',''),
('00001','C22222222','Area2',1,222222222,'2017-03-22 10:05:19.130','','POD15'),
('00001','C22222222','Area2',2,222222222,'2017-03-22 10:13:43.057','POD15','A'),
('00001','C22222222','Area2',3,222222222,'2017-03-22 10:25:01.527','A','A'),
('00001','C22222222','Area2',3,222222222,'2017-03-22 10:46:03.960','A','A'),
('00001','C22222222','Area2',3,222222222,'2017-03-22 10:46:17.030','A','A'),
('00002','C33333333','Area3',1,333333333,'2017-03-22 09:20:23.660','','B46'),
('00001','C33333333','Area2',9,333333333,'2017-03-22 08:53:32.860','POD8','POD1'),
('00001','C33333333','Area2',1,333333333,'2017-03-22 07:34:58.810','POD7','POD8'),
('00001','C33333333','Area2',1,333333333,'2017-03-22 11:49:55.203','','BB4'),
('00001','C33333333','Area2',3,333333333,'2017-03-22 11:50:11.943','BB4','BB4'),
('00001','C33333333','Area2',3,333333333,'2017-03-22 08:42:56.157','POD8','POD8'),
('00001','C33333333','Area2',3,333333333,'2017-03-22 08:22:59.157','POD8','POD8'),
('00003','C33333333','Area3',1,333333333,'2017-03-23 06:41:12.753','','B46')

GO

这是我到目前为止的查询;这将给我前一行值和下一行值，但我不认为它会考虑客户。

SELECT
    T1.[User], T1.[CustomerID],
    T1.[Area], T1.[Event],
    T1.[VisitNumber],
    T1.[CurrentRoom], T1.[NewRoom],
    T1.[Time],
    LAG(T1.TIME) OVER (ORDER BY T1.VisitNumber) PreviousTime,
    LEAD(T1.TIME) OVER (ORDER BY T1.VisitNumber) NextTime
FROM
    #beds t1
WHERE
    T1.[Area] = 'Area2'
    AND T1.[CurrentRoom] IS NOT NULL
    AND T1.[NewRoom] IS NOT NULL
    AND T1.[CustomerID] IS NOT NULL
    AND T1.[CustomerID] <> ' '
    AND T1.Event IN (1,9)
ORDER BY
    VisitNumber DESC

预期输出：这是我期待的输出。我只想要TimeInRoom（不包括当时的日期字段）：

+------------+-------+-------------+-------------+---------+------------+
| CustomerID | Area  | VisitNumber | CurrentRoom | NewRoom | TimeInRoom |
+------------+-------+-------------+-------------+---------+------------+
|C33333333   |Area2  | 333333333   |             | BB4     | 00:10      |
|C33333333   |Area2  | 333333333   |             | POD8    | 00:20      |
|C33333333   |Area2  | 333333333   | POD8        |         | 00:30      |
+------------+-------+-------------+-------------+---------+------------+

3 个答案:

答案 0 :(得分：4)

可能我不太了解您的问题，但尝试在LAG / LEAD函数中使用PARTITION BY子句：

,LAG(T1.TIME) OVER (PARTITION BY CustomerID ORDER BY T1.VisitNumber) PreviousTime
,LEAD(T1.TIME) OVER (PARTITION BY CustomerID ORDER BY T1.VisitNumber) NextTime

答案 1 :(得分：4)

我希望这会有所帮助：

;WITH cte_Result AS
(
    SELECT
        [CustomerID],
        [Area],
        [VisitNumber],
        [CurrentRoom],
        [NewRoom],
        [Time],
        LAG([TIME]) OVER (partition by [CustomerID],[VisitNumber] ORDER BY ID DESC) PreviousTime,
        LEAD([TIME]) OVER (partition by [CustomerID],[VisitNumber] ORDER BY ID DESC) NextTime
    FROM #beds
    WHERE   [Area] = 'Area2'
        AND [CurrentRoom] IS NOT NULL
        AND [NewRoom] IS NOT NULL
        AND [CustomerID] IS NOT NULL
        AND [CustomerID] <> ' '
        AND [Event] IN (1,9)
        --AND [CustomerID] = 'C33333333'
),
cte_BuildStayPeriod
AS (
    SELECT CustomerID,
        Area,
        VisitNumber,
        CurrentRoom,
        NewRoom,
        DATEDIFF(SECOND, COALESCE([NextTime], PreviousTime), COALESCE(PreviousTime, [time])) AS StayDuration
    FROM cte_Result
)
SELECT CustomerID,
    Area,
    VisitNumber,
    CurrentRoom,
    NewRoom,
    StayDuration,
    CAST(DATEADD(SECOND, StayDuration, '1900-01-01') AS TIME) AS StayDuration
FROM cte_BuildStayPeriod

答案 2 :(得分：0)

在您的示例中，当客户访问一次时会出现问题，并且使用滞后/潜在客户，您将获得另一位客户的访问信息。

试一试：

SELECT
    T1.[User], T1.[CustomerID],
    T1.[Area], T1.[Event],
    T1.[VisitNumber],
    T1.[CurrentRoom], T1.[NewRoom],
    T1.[Time],
    (select TOP (1) t.Time from #beds t where t.[CustomerID] = T1.[CustomerID] and t.Time<T1.Time order by t.Time desc) PreviousTime,
    (select TOP (1) t.Time from #beds t where t.[CustomerID] = T1.[CustomerID] and t.Time>T1.Time order by t.Time) NextTime
FROM
    #beds t1
WHERE
    T1.[Area] = 'Area2'
    AND T1.[CurrentRoom] IS NOT NULL
    AND T1.[NewRoom] IS NOT NULL
    AND T1.[CustomerID] IS NOT NULL
    AND T1.[CustomerID] <> ' '
    AND T1.Event IN (1,9)
ORDER BY
    VisitNumber DESC
相关问题
最新问题