在 MSSQL 中选择 2 个不同列的行之间的 timediff答案

【问题标题】：Select timediff between rows of 2 different columns in MSSQL在 MSSQL 中选择 2 个不同列的行之间的 timediff
【发布时间】：2021-06-23 10:53:42
【问题描述】：

我正在尝试在 SQL 中查找电话之间的时差。我的示例数据看起来像

name	call_start	call_finished
Johnny	2021-05-21 10:52:22.000	2021-05-21 10:52:25.000
Peter	2021-05-21 10:52:43.000	2021-05-21 10:53:10.000
Peter	2021-05-21 10:54:01.000	2021-05-21 10:54:01.000
Johnny	2021-05-21 10:53:48.000	2021-05-21 10:53:51.000
Peter	2021-05-21 10:54:14.000	2021-05-21 10:56:31.000
Johnny	2021-05-21 10:56:33.000	2021-05-21 10:56:38.000
Johnny	2021-05-21 10:58:13.000	2021-05-21 10:58:52.000
Johnny	2021-05-21 11:00:37.000	2021-05-21 11:00:37.000
Johnny	2021-05-21 11:03:14.000	2021-05-21 11:04:06.000
Peter	2021-05-21 11:05:20.000	2021-05-21 11:05:20.000

我想为每个name 找出call_finished 和call_start 之间的区别，以便生成一个Time_since_last_call 列。

name	call start	call_finished	Time_since_last_call
Johnny	2021-05-21 10:52:22.000	2021-05-21 10:52:25.000	Null
Peter	2021-05-21 10:52:43.000	2021-05-21 10:53:10.000	Null
Peter	2021-05-21 10:54:01.000	2021-05-21 10:54:01.000	51
Johnny	2021-05-21 10:53:48.000	2021-05-21 10:53:51.000	83
Peter	2021-05-21 10:54:14.000	2021-05-21 10:56:31.000	13
Johnny	2021-05-21 10:56:33.000	2021-05-21 10:56:38.000	162
Johnny	2021-05-21 10:58:13.000	2021-05-21 10:58:52.000	95
Johnny	2021-05-21 11:00:37.000	2021-05-21 11:00:37.000	105
Johnny	2021-05-21 11:03:14.000	2021-05-21 11:04:06.000	157
Peter	2021-05-21 11:05:20.000	2021-05-21 11:05:20.000	529

问题

如何为每个名称动态执行此操作？
是否可以只查找每天的时差？那么Time_since_last_call 不是在天数之间计算的吗？

【问题讨论】：

标签： sql-server azure-sql-database

【解决方案1】：

您可以使用DATEDIFF 和LAG 轻松实现此目的：

USE Sandbox;
GO
CREATE TABLE dbo.YourTable ([Name] varchar(10),
                            CallStart datetime,
                            CallFinished datetime);
GO
INSERT INTO dbo.YourTable
VALUES('Johnny','2021-05-21T10:52:22.000','2021-05-21T10:52:25.000'),
      ('Peter','2021-05-21T10:52:43.000','2021-05-21T10:53:10.000'),
      ('Peter','2021-05-21T10:54:01.000','2021-05-21T10:54:01.000'),
      ('Johnny','2021-05-21T10:53:48.000','2021-05-21T10:53:51.000'),
      ('Peter','2021-05-21T10:54:14.000','2021-05-21T10:56:31.000'),
      ('Johnny','2021-05-21T10:56:33.000','2021-05-21T10:56:38.000'),
      ('Johnny','2021-05-21T10:58:13.000','2021-05-21T10:58:52.000'),
      ('Johnny','2021-05-21T11:00:37.000','2021-05-21T11:00:37.000'),
      ('Johnny','2021-05-21T11:03:14.000','2021-05-21T11:04:06.000'),
      ('Peter','2021-05-21T11:05:20.000','2021-05-21T11:05:20.000');
GO
SELECT Name,
       CallStart,
       CallFinished,
       DATEDIFF(SECOND,LAG(CallFinished) OVER (PARTITION BY [Name] ORDER BY CallStart),CallStart) AS TimeSinceLastCall
FROM dbo.YourTable
ORDER BY CallStart;

GO

DROP TABLE dbo.YourTable;

如果您不希望它分散几天（此方案没有示例数据），这意味着每天的第一个呼叫（对于每个人）将具有值 NULL 和 TimeSinceLastCall，然后将CONVERT(date,CallStart) 或CAST(CallStart AS date) 添加到PARTITION BY 子句中。

【讨论】：

喜欢DATEDIFF(SECOND,LAG(Call_Finished) OVER (PARTITION BY CAST([Call Start] AS date), [Name] ORDER BY [Call Start]),[Call Start]) AS TimeSinceLastCall ?