处理SQL Datetime中缺少登录/注销值

Question

我的Sql查询：

CREATE TABLE TimeLog (
    [User] NVARCHAR(6),
    [Event] NVARCHAR(3),
    [Time] DATETIME
);
INSERT INTO TimeLog VALUES
(N'Jibran',N'IN','2015-04-15 00:31:00'),
(N'Jibran',N'IN','2015-04-16 20:10:00'),
(N'Jibran',N'IN','2015-04-21 14:59:00'),
(N'Jibran',N'OUT','2015-04-22 01:01:00'),
(N'Jibran',N'IN','2015-04-22 10:46:00'),
(N'Jibran',N'OUT','2015-04-23 00:58:00'),
(N'Jibran',N'IN','2015-04-23 14:50:00'),
(N'Jibran',N'OUT','2015-04-24 01:37:00'),
(N'Jibran',N'OUT','2015-04-25 01:01:00'),
(N'Jibran',N'OUT','2015-04-27 00:57:00'),
(N'Jibran',N'IN','2015-04-17 10:32:00'),
(N'Jibran',N'IN','2015-04-29 15:03:00'),
(N'Jibran',N'OUT','2015-05-01 00:44:00'),
(N'Jibran',N'OUT','2015-05-02 01:19:00'),
(N'Jibran',N'IN','2015-05-02 15:08:00'),
(N'Jibran',N'OUT','2015-05-03 01:08:00'),
(N'Jibran',N'IN','2015-05-03 15:06:00'),
(N'Jibran',N'OUT','2015-05-04 01:01:00'),
(N'Jibran',N'IN','2015-05-04 15:11:00'),
(N'Jibran',N'OUT','2015-05-05 01:08:00');

SELECT TOP (30) UserName, EventName, EventTime
From AttendanceEvents
Where UserName = 'Jibran'

结果是 ：

有没有办法处理没有OUT日期时间值的每个IN值的缺失值？

我一直在读一篇文章：https://www.red-gate.com/simple-talk/sql/t-sql-programming/calculating-gaps-between-overlapping-time-intervals-in-sql/

但无法理解它的一半。

如何获取日期时间平均值以将其用于缺失值？

期望输出将具有每个IN的OUT值。

谢谢。

Answer 1

我想出了一个解决方案，将缺少的行添加到表中 - 两者都缺少in行和缺少out行。 我所做的一件事与你提出的不同的是，缺少的行在他们的源行中完成了8小时的一天。

请注意，每次只能与单个用户一起使用。

所以这是如何：

首先，我创建了一个公用表表达式，其中包含表中属于特定用户的所有行。在那个cte我用lag和lead来获取下一个事件，前一个事件和下一个事件时间，以及一个row_number列。

然后我使用了一个基于cte的三个查询的联合 - 一个用于原始行，一个用于具有in事件的新行，一个用于具有out事件的新行。

该脚本基于您的示例数据，您可以看到online demo on rextester.

DECLARE @User nvarchar(6) = N'Jibran';

WITH CTE AS
(
    SELECT  [User],
            [Event],
            [Time],
            ROW_NUMBER() OVER(ORDER BY [Time]) + 0.0 As rn,
            LAG([Event]) OVER(ORDER BY [Time])  As PrevEvent,
            LEAD([Event]) OVER(ORDER BY [Time])  As NextEvent,
            LEAD([Time]) OVER(ORDER BY [Time]) As NextEventTime
    FROM TimeLog
    WHERE [User] = @User
)

SELECT  [User],
        'OUT' As [Event],
        DATEADD(HOUR, 8, [Time]) As [Time],
        rn + 0.5 As rn
FROM CTE
WHERE NextEvent = [Event]
AND [Event] = 'IN'

UNION ALL

SELECT  [User],
        'IN' As [Event],
        DATEADD(HOUR, -8, [NextEventTime]) As [Time],
        rn - 0.3 As rn
FROM CTE
WHERE PrevEvent = [Event]
AND [Event] = 'OUT'

UNION ALL
SELECT  [User],
        [Event],
        [Time],
        rn
FROM CTE

ORDER BY rn

结果：

User        Event   Time                    rn
Jibran      IN      15.04.2015 00:31:00     1,0
Jibran      OUT     15.04.2015 08:31:00     1,5
Jibran      IN      16.04.2015 20:10:00     2,0
Jibran      OUT     17.04.2015 04:10:00     2,5
Jibran      IN      17.04.2015 10:32:00     3,0
Jibran      OUT     17.04.2015 18:32:00     3,5
Jibran      IN      21.04.2015 14:59:00     4,0
Jibran      OUT     22.04.2015 01:01:00     5,0
Jibran      IN      22.04.2015 10:46:00     6,0
Jibran      OUT     23.04.2015 00:58:00     7,0
Jibran      IN      23.04.2015 14:50:00     8,0
Jibran      OUT     24.04.2015 01:37:00     9,0
Jibran      IN      26.04.2015 16:57:00     9,7
Jibran      OUT     25.04.2015 01:01:00     10,0
Jibran      IN      29.04.2015 07:03:00     10,7
Jibran      OUT     27.04.2015 00:57:00     11,0
Jibran      IN      29.04.2015 15:03:00     12,0
Jibran      OUT     01.05.2015 00:44:00     13,0
Jibran      IN      02.05.2015 07:08:00     13,7
Jibran      OUT     02.05.2015 01:19:00     14,0
Jibran      IN      02.05.2015 15:08:00     15,0
Jibran      OUT     03.05.2015 01:08:00     16,0
Jibran      IN      03.05.2015 15:06:00     17,0
Jibran      OUT     04.05.2015 01:01:00     18,0
Jibran      IN      04.05.2015 15:11:00     19,0
Jibran      OUT     05.05.2015 01:08:00     20,0

旁注：您可以通过rn告诉哪一行添加了哪一部分联合所有查询。

更新

在我们在问题的评论中进行对话之后 - 为了获得不同列中的详细信息，我将联合所有查询包装在另一个cte中，并在其上加上一个cte来获取bigint行号。然后，这只是一个带有扭曲的条件聚合问题 - 使用交叉应用将偶数和奇数行数组合成组。

这是完整的脚本 - 当然，an online demo：

DECLARE @User nvarchar(6) = N'Jibran';

WITH CTE AS
(
    SELECT  [User],
            [Event],
            [Time],
            ROW_NUMBER() OVER(ORDER BY [Time]) + 0.0 As rn,
            LAG([Event]) OVER(ORDER BY [Time])  As PrevEvent,
            LEAD([Event]) OVER(ORDER BY [Time])  As NextEvent,
            LEAD([Time]) OVER(ORDER BY [Time]) As NextEventTime
    FROM TimeLog
    WHERE [User] = @User
), CTERows AS
(
    -- Added out rows
    SELECT  [User],
            'OUT' As [Event],
            DATEADD(HOUR, 8, [Time]) As [Time],
            rn + 0.5 As rn
    FROM CTE
    WHERE NextEvent = [Event]
    AND [Event] = 'IN'

    UNION ALL

    -- Added in rows
    SELECT  [User],
            'IN' As [Event],
            DATEADD(HOUR, -8, [NextEventTime]) As [Time],
            rn - 0.3 As rn
    FROM CTE
    WHERE PrevEvent = [Event]
    AND [Event] = 'OUT'

    UNION ALL

    -- Existing rows
    SELECT  [User],
            [Event],
            [Time],
            rn
    FROM CTE
), CTEIntNumberedRows AS
(
    SELECT  [User],
            [Event],
            [Time],
            ROW_NUMBER() OVER (ORDER BY rn) As rn
    FROM CteRows
)

SELECT  [User],
        MAX(CASE WHEN [Event] = 'IN' THEN [Time] END) As 'IN',
        MAX(CASE WHEN [Event] = 'OUT' THEN [Time] END) As 'OUT',
        Pairs
FROM CTEIntNumberedRows
CROSS APPLY 
(
    SELECT CASE WHEN rn % 2 = 0 THEN rn-1 ELSE rn END As Pairs
) x
GROUP BY [User], Pairs
ORDER BY Pairs