SQL 标志行在前一个标志之后 2 小时出现答案

【问题标题】：SQL Flag Rows Occurring 2 Hours After Previous FlagSQL 标志行在前一个标志之后 2 小时出现
【发布时间】：2020-10-12 20:17:17
【问题描述】：

由于天线过于敏感，我有一些混乱的 RFID 数据。有一个物理过程可以跟踪 RFID 标签在一个周期中通过不同站点的移动。带有 RFID 标签的物品每天可以在循环中移动不止一次，但它不太可能在第一次读取后的两个小时内开始循环。

我正在尝试创建一个标志列来确定一个项目的新周期何时开始，或者返回一个项目经历了多少次周期的计数。

这里是一些示例数据：

CREATE TABLE [dbo].[samplerfiddata](
    [Item] [nvarchar](50) NOT NULL,
    [Station_Type] [nvarchar](50) NOT NULL,
    [Station_Name] [nvarchar](50) NOT NULL,
    [Timestamp] [datetime2](7) NOT NULL,
    [Trying_to_Create_this_Flag_Column] [nvarchar](50) NOT NULL
) ON [PRIMARY]
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'A', N'Decontamination', N'Decontamination', CAST(N'2020-10-10T06:30:00.0000000' AS DateTime2), N'1')
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'A', N'Decontamination', N'Decontamination', CAST(N'2020-10-11T14:30:00.0000000' AS DateTime2), N'1')
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'A', N'Washer', N'Washer', CAST(N'2020-10-11T14:45:00.0000000' AS DateTime2), N'0')
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'A', N'Decontamination', N'Decontamination', CAST(N'2020-10-11T15:15:00.0000000' AS DateTime2), N'0')
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'A', N'Other', N'Decontamination', CAST(N'2020-10-11T23:30:00.0000000' AS DateTime2), N'1')
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'A', N'Washer', N'Washer', CAST(N'2020-10-12T00:15:00.0000000' AS DateTime2), N'0')
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'A', N'Other', N'Decontamination', CAST(N'2020-10-12T00:45:00.0000000' AS DateTime2), N'0')
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'A', N'Other', N'Decontamination', CAST(N'2020-10-13T16:00:00.0000000' AS DateTime2), N'1')
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'A', N'Other', N'Decontamination', CAST(N'2020-10-13T16:30:00.0000000' AS DateTime2), N'0')
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'A', N'Decontamination', N'Decontamination', CAST(N'2020-10-14T13:30:00.0000000' AS DateTime2), N'1')
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'B', N'Other', N'Decontamination', CAST(N'2020-10-12T08:30:00.0000000' AS DateTime2), N'1')
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'B', N'Decontamination', N'Decontamination', CAST(N'2020-10-12T14:30:00.0000000' AS DateTime2), N'1')
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'B', N'Washer', N'Washer', CAST(N'2020-10-12T14:45:00.0000000' AS DateTime2), N'0')
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'B', N'Decontamination', N'Decontamination', CAST(N'2020-10-12T15:15:00.0000000' AS DateTime2), N'0')
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'B', N'Decontamination', N'Decontamination', CAST(N'2020-10-12T18:00:00.0000000' AS DateTime2), N'1')
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'B', N'Washer', N'Washer', CAST(N'2020-10-13T18:15:00.0000000' AS DateTime2), N'0')
GO
INSERT [dbo].[RFID SQL Problem Sample Data] ([Item], [Station_Type], [Station_Name], [Timestamp], [Trying_to_Create_this_Flag_Column]) VALUES (N'B', N'Decontamination', N'Decontamination', CAST(N'2020-10-13T19:00:00.0000000' AS DateTime2), N'1')
GO

在此数据中，我们有两个不同的项目在循环中移动，并捕获了所有场景。新周期的业务逻辑定义为当一个项目被一个 RFID 天线（站）读取时，其中站类型 = 去污或站名称 = 去污，并且距离上一次循环的第一次读取已经超过两个小时.

项目 | Station_Type |站名 |时间戳 | Trying_to_Create_this_Flag_Column :--- | :---------------- | :---------------- | :---------------------------- | :-------------------------------- 一个 |去污 |去污 | 2020-10-10 06:30:00.0000000 | 1 一个 |去污 |去污 | 2020-10-11 14:30:00.0000000 | 1 一个 |垫圈 |垫圈 | 2020-10-11 14:45:00.0000000 | 0 一个 |去污 |去污 | 2020-10-11 15:15:00.0000000 | 0 一个 |其他 |去污 | 2020-10-11 23:30:00.0000000 | 1 一个 |垫圈 |垫圈 | 2020-10-12 00:15:00.0000000 | 0 一个 |其他 |去污 | 2020-10-12 00:45:00.0000000 | 0 一个 |其他 |去污 | 2020-10-13 16:00:00.0000000 | 1 一个 |其他 |去污 | 2020-10-13 16:30:00.0000000 | 0 一个 |去污 |去污 | 2020-10-14 13:30:00.0000000 | 1 乙|其他 |去污 | 2020-10-12 08:30:00.0000000 | 1 乙|去污 |去污 | 2020-10-12 14:30:00.0000000 | 1 乙|垫圈 |垫圈 | 2020-10-12 14:45:00.0000000 | 0 乙|去污 |去污 | 2020-10-12 15:15:00.0000000 | 0 乙|去污 |去污 | 2020-10-12 18:00:00.0000000 | 1 乙|垫圈 |垫圈 | 2020-10-13 18:15:00.0000000 | 0 乙|去污 |去污 | 2020-10-13 19:00:00.0000000 | 1

最终，我试图返回如下所示的结果集：

项目 |循环 :--- | -----：一个 | 5 乙| 4

这怎么可能完成？

【问题讨论】：

标签： sql sql-server datetime recursive-query sql-server-2019

【解决方案1】：

如果我没听错，你需要递归查询：

with 
    data as (
        select item, station_type, station_name, timestamp, 
            row_number() over(partition by item order by timestamp) rn
        from samplerfiddata sf
    ),
    cte as (
        select d.*, timestamp first_timestamp 
        from data d 
        where rn = 1
        union all
        select d.*, 
            case when 'Decontamination' in (d.station_type, d.station_name) and d.timestamp > dateadd(hour, 2, c.first_timestamp)
                then d.timestamp
                else c.timestamp
            end
        from cte c
        inner join data d on d.item = c.item and d.rn = c.rn + 1
    )
select *
from cte
order by item, timestamp

这个想法是迭代遍历每个item的数据集，同时跟踪最后一个循环开始（这里别名为first_timestamp）；当遇到比前一次晚两个多小时的“去污”事件时，循环时间戳会重置。

您可以通过在外部查询中使用聚合来获取循环计数，例如：

select item, count(*) cnt_cycles
from cte
where timestamp = first_timestamp
group by item

Demo on DB Fiddle

【讨论】：

【解决方案2】：

试试这个，

with CTE as
(
select * ,
lead([Timestamp])over(partition by [Item] order by[Timestamp])lagtime
from [dbo].[samplerfiddata]
where Station_Type='Decontamination' or Station_Name='Decontamination'
)

select item
,sum(case when lagtime is null then 1
when datediff(MINUTE,[Timestamp],lagtime)>=60 then 1 else 0 end) as cycles

from CTE
group by item

【讨论】：