I need help with my query to calculate the time until midnight between two date and time columns
break down by day
This is the main table:
ID
Start_Time
End_time
DateDiff
32221
01-01-2022 13:10:00
01-03-2022 13:10:00
2880
My query:
SELECT
start_time.ID,
start_time.Date_Time AS Start_time,
end_time.Date_Time AS End_time,
DATEDIFF(minute, start_time.Date_Time, end_time.Date_Time) AS DateDiff
FROM
Main
what I need is similar to this:
ID
Date_start
End_time
DateDiff
32221
01-01-2022 13:10:00
01-01-2022 23:59:59
654
32221
01-02-2022 00:00:00
01-02-2022 23:59:59
1440
32221
01-03-2022 00:00:00
01-03-2022 13:10:00
781
how i can do that?
You can loop through the times, always adding the time untill midnight, untill your 'start_time + 1 day' is bigger than your end_time.
The below code can be run directly in SQL (mind the date notation, my SQL is in united states notation, so if yours is in Europe this will give you back results for 3 months instead of 3 days);
DECLARE #start_time datetime2 = '01/01/2022 13:00:00';
DECLARE #end_time datetime2 = '03/01/2022 14:00:00';
DECLARE #daily_end_time datetime2=NULL;
DECLARE #Table Table (start_time datetime2, end_time datetime2, diff nvarchar(8));
DECLARE #diff_minutes_start int = DATEDIFF(MINUTE,#start_time,DateDiff(day,0,dateadd(day,1,#start_time)));
DECLARE #diff_minutes_end int = DATEDIFF(minute,#end_time,DateDiff(day,0,dateadd(day,1,#end_time)))
SET #daily_end_time = DATEADD(mi,#diff_minutes_start,#start_time)
WHILE #daily_end_time < #end_time
BEGIN
INSERT INTO #Table (start_time,end_time,diff)
VALUES (
#start_time,
CASE WHEN DATEADD(day,1,#daily_end_time) > #end_time THEN #end_time ELSE
#daily_end_time END,
CASE WHEN DATEADD(day,1,#daily_end_time) > #end_time THEN #diff_minutes_end ELSE
#diff_minutes_start END )
SET #daily_end_time = DATEADD(mi,#diff_minutes_start,#start_time)
SET #start_time = DATEADD(mi,1,#daily_end_time);
select #diff_minutes_start =
DATEDIFF(MINUTE,#start_time,DateDiff(day,0,dateadd(day,1,#start_time)));
select #diff_minutes_end = DATEDIFF(minute,#end_time,DateDiff(day,0,dateadd(day,1,#end_time)))
END
SELECT * FROM #Table
And the results:
You may use a recursive CTE as the following:
With CTE As
(
Select ID, Start_Time, End_time, DATEADD(Second, -1, DATEADD(Day, DATEDIFF(Day,0, Start_Time), 1)) et
From main
Union All
Select C.ID, DATEADD(Second, 1, C.et), C.End_time, DATEADD(Day, 1, C.et)
From CTE C Join main T
On C.ID = T.ID
Where DATEADD(Second, 1, C.et) <= C.End_time
)
Select ID, Start_Time,
Case When End_Time <= et Then End_Time Else et End As End_Time,
DATEDIFF(Minute, Start_Time, DATEADD(Second, 1, Case When End_Time <= et Then End_Time Else et End)) As [DateDiff]
From CTE
Order By ID, Start_Time
See a demo with extended data sample from db<>fiddle.
You can also solve this with a tally table, using the expanded (to show different cases) sample data
ID
StartTime
EndTime
32221
2022-01-01 13:10:00
2022-01-03 13:10:00
32222
2022-02-02 10:10:00
2022-02-02 17:10:00
32223
2022-03-03 19:10:00
2022-03-04 08:10:00
32224
2022-04-04 19:10:00
2022-04-08 08:10:00
and the code
with cteSampleData as ( --Enter some sample data, include spans of 0, 1, and >1 days
SELECT * --Note that we need CONVERT to make sure the dates are treated as datetime not string!
FROM (VALUES(32221, CONVERT(datetime2(0), '01-01-2022 13:10:00'), CONVERT(datetime2(0), '01-03-2022 13:10:00') )
, (32222, '02-02-2022 10:10:00', '02-02-2022 17:10:00')
, (32223, '03-03-2022 19:10:00', '03-04-2022 08:10:00')
, (32224, '04-04-2022 19:10:00', '04-08-2022 08:10:00')
) as Samp(ID, StartTime, EndTime)
), cteWithControl as ( --Add some fields to make testing cledarer - you could do this as part of a subsequent step instead
SELECT *
, CONVERT(date, StartTime) as StartDate , CONVERT(date, EndTime) as EndDate
, DATEDIFF(day, StartTime , EndTime) as DiffDays
--, DATEDIFF(day, CONVERT(date, StartTime) , CONVERT(date, EndTime)) as DiffDays
FROM cteSampleData
), cteTally as ( --Get a list of integers to represent days, assume nothing lasts longer than a year
SELECT top 365 ROW_NUMBER() over (ORDER BY name) as Tally
FROM sys.objects --just a table we know has over 300 rows, look up tally tables for other generation methods
)--The real work begins below, partition the data into "same day" and "multi-day" spans
, cteSet as (
SELECT ID, StartTime, EndTime, DiffDays, 1 as DayNumber
FROM cteWithControl WHERE DiffDays = 0
UNION ALL
SELECT ID --For multi-day, cross with the tally table and treat first and last days special
, CASE WHEN T.Tally = 1 THEN StartTime --For the first day the start time is the real time
ELSE DATEADD (day, T.Tally - 1, startdate) END as StartTime --Otherwise it's the start of the day
, CASE WHEN T.Tally = DiffDays + 1 THEN EndTime --For the last day the end is the real end
ELSE DATEADD (second, -1, CONVERT(DATETIME2(0), DATEADD (day, T.Tally, startdate)))
END as EndTime --otherwise 1 second less than the next day
, DiffDays, Tally as DayNumber
FROM cteWithControl as D CROSS JOIN cteTally as T
WHERE DiffDays > 0 AND T.Tally <= D.DiffDays + 1
)--Now we display the results and calculate the length (in minutes) of each span
SELECT *
, DATEDIFF(MINUTE, StartTime, EndTime) as DateDiff
FROM cteSet
ORDER BY ID, DayNumber
we get the output
ID
StartTime
EndTime
DiffDays
DayNumber
DateDiff
32221
2022-01-01 13:10:00
2022-01-01 23:59:59
2
1
649
32221
2022-01-02 00:00:00
2022-01-02 23:59:59
2
2
1439
32221
2022-01-03 00:00:00
2022-01-03 13:10:00
2
3
790
32222
2022-02-02 10:10:00
2022-02-02 17:10:00
0
1
420
32223
2022-03-03 19:10:00
2022-03-03 23:59:59
1
1
289
32223
2022-03-04 00:00:00
2022-03-04 08:10:00
1
2
490
32224
2022-04-04 19:10:00
2022-04-04 23:59:59
4
1
289
32224
2022-04-05 00:00:00
2022-04-05 23:59:59
4
2
1439
32224
2022-04-06 00:00:00
2022-04-06 23:59:59
4
3
1439
32224
2022-04-07 00:00:00
2022-04-07 23:59:59
4
4
1439
32224
2022-04-08 00:00:00
2022-04-08 08:10:00
4
5
490
suppose this is my data in a table in the database
...
01/01/2016 00:00 367.2647688
01/06/2016 12:30 739.8067639 < INCLUDE THIS
01/01/2018 03:00 412.9686137
01/01/2018 03:30 150.6068046
01/01/2018 04:00 79.22204568
01/01/2018 04:30 648.702222
01/01/2018 09:00 75.41931365
01/01/2018 09:30 923.9435812
01/01/2018 10:00 342.9116004
02/01/2018 02:00 776.4855197 < INCLUDE THIS
08/04/2021 02:30 206.2066933
02/01/2022 03:00 852.9874735
02/01/2022 03:30 586.0818207
02/01/2022 04:00 363.5394613
02/01/2023 04:30 874.3073237
...
and this is my query to fetch data
SELECT * FROM MYTABLE WHERE [DATETIME] >= '2018/01/01 03:00' AND [DATETIME] < '2018/01/01 11:00'
I would also like the query to return one date before and after this range. so like the dates padded.
Now how can i do this efficiently. One way could be to get the dates in the ranges and then get all the data where they are less then min date and get the highest datetime of those and add to main range table also repeating this process for the max date. is there any other way?
You can use lead() and lag():
SELECT *
FROM (SELECT t.*,
LAG(DATETIME, 1, DATETIME) OVER (ORDER BY DATETIME) as PREV_DATETIME,
LEAD(DATETIME, 1, DATETIME) OVER (ORDER BY DATETIME) as NEXT_DATETIME
FROM MYTABLE t
) t
WHERE NEXT_DATETIME >= '2018-01-01 03:00:00' AND
PREV_DATETIME <= '2018-01-01 10:00:00'
Note: This uses default values to simplify the logic.
Here is a db<>fiddle. Based on the results you specified, I changed the last comparison to <= from <.
Lag and lead are window functions that are used to get preceding and succeeding value of any row within its partition. So when we want pad rows outside the range we can set offset parameter in both lag and lead functions.
SELECT t.Datetime,t.val FROM (SELECT T.*,LAG(Datetime,1,Datetime) over (order by Datetime) as lagdate,lead(Datetime,1,Datetime)over (order by Datetime) as leaddate FROM Mytable T)t
WHERE leaddate>= '2018/01/01 03:00' and lagdate<='2018/01/01 11:00'
I have a MSSQL table where the raw data is formatted as this:
date1 time1
2008-01-20 00:00:00 654
2008-01-20 00:00:00 659
2008-01-20 00:00:00 1759
and I need to join both of them together so I can query for example all date_time that happened in the last 15 hours. what I did was
in the select statement:
combined = CONVERT(VARCHAR(10), Date1, 103) +' ' + (left((replace((CONVERT(dec(7, 2), time1) / 100 ),'.',':')),4) + ':00') ,
This helped me with getting results for
date1 time1 combined1
2008-01-20 00:00:00 654 20/01/2008 6:54:00
2008-01-20 00:00:00 659 20/01/2008 6:59:00
2008-01-20 00:00:00 1759 20/01/2008 17:5:00
I cant change the table data & I cant get the right syntax to convert it fully (including taking in consideration the 24h hour format - 1759 for example)
And in the end I need to be able to do a where statement on the combined1 column to see only the rows that happened in the last 15 hours
DATEADD(hour, - 15, GETDATE())
Thanks in advance
Try This
select date1,time1, DATEADD(MINUTE, time1%100, DATEADD(HOUR, time1/100, convert(varchar(10),date1,101))) as Combined
from Table
Where DATEADD(MINUTE, time1%100, DATEADD(HOUR, time1/100, convert(varchar(10),date1,101)))>(DATEADD(hour,-15,GETDATE()))
Try like this:
DECLARE #date DATETIME = '2008-01-20 00:00:00'
, #Time INT = 654
SELECT DATEADD(MINUTE, #Time%100, DATEADD(HOUR, #Time/100, #date))
;WITH cte
AS (SELECT CAST('2008-01-20 00:00:00' AS DATETIME) AS date1, 654 AS Time1
UNION ALL
SELECT CAST('2008-01-20 00:00:00' AS DATETIME) AS date1, 659 AS Time1
UNION ALL
SELECT CAST('2008-01-20 00:00:00' AS DATETIME) AS date1, 1759 AS Time1
)
SELECT
DATEADD(ms, DATEDIFF(ms, '00:00:00', CAST(FORMAT(Time1, '##:##') AS TIME)), date1) AS [CombinedDateTime]
FROM cte;
--Results to:
CombinedDateTime
2008-01-20 06:54:00.000
2008-01-20 06:59:00.000
2008-01-20 17:59:00.000
I have a table in a SQL Server 2012 database that logs events, with columns StartDate and EndDate. I need to aggregate all of the records within a certain time period, and determine the duration of time where any events were active, not counting overlapping durations. For example, if my table looks like this:
id StartDate EndDate
=======================================================
1 2017-08-28 12:00:00 PM 2017-08-28 12:01:00 PM
2 2017-08-28 1:15:00 PM 2017-08-28 1:17:00 PM
3 2017-08-28 1:16:00 PM 2017-08-28 1:20:00 PM
4 2017-08-28 1:30:00 PM 2017-08-28 1:35:00 PM
And my time period to search was from 2017-08-28 12:00:00 PM to 2017-08-28 2:00:00 PM, then my desired output should be:
Duration of Events Active = 00:11:00
I have been able to aggregate the records and get a total duration (basically just EndDate - StartDate), but I cannot figure out how to exclude overlapping time. Any help would be appreciated. Thanks!
How about a CTE and aggregation? This can be done with a sub-query too.
declare #table table (id int, StartDate datetime, EndDate datetime)
insert into #table
values
( 1,'2017-08-28 12:00:00 PM','2017-08-28 12:01:00 PM'),
(2,'2017-08-28 1:15:00 PM','2017-08-28 1:17:00 PM'),
(3,'2017-08-28 1:16:00 PM','2017-08-28 1:20:00 PM'),
(4,'2017-08-28 1:30:00 PM','2017-08-28 1:35:00 PM')
declare #StartDate datetime = '2017-08-28 12:00:00'
declare #EndDate datetime = '2017-08-28 14:00:00'
;with cte as(
select
id
,StartDate = case when StartDate < lag(EndDate) over (order by id) then lag(EndDate) over (order by id) else StartDate end
,EndDate
from
#table
where
StartDate >= #StartDate
and EndDate <= #EndDate),
cte2 as(
select
Dur = datediff(second, StartDate, EndDate)
from cte)
select
Dur = convert(varchar, dateadd(ms, sum(Dur) * 1000, 0), 114)
from
cte2
How can I count the number of rows per hour in SQL Server with full date-time as result.
I've already tried this, but it returns only the hours
SELECT DATEPART(HOUR,TimeStamp), Count(*)
FROM [TEST].[dbo].[data]
GROUP BY DATEPART(HOUR,TimeStamp)
ORDER BY DATEPART(HOUR,TimeStamp)
Now the result is:
Hour Occurrence
---- ----------
10 2157
11 60740
12 66189
13 77096
14 90039
But I need this:
Timestamp Occurrence
------------------- ----------
2013-12-21 10:00:00 2157
2013-12-21 11:00:00 60740
2013-12-21 12:00:00 66189
2013-12-21 13:00:00 77096
2013-12-21 14:00:00 90039
2013-12-22 09:00:00 84838
2013-12-22 10:00:00 64238
You actually need to round the TimeStamp to the hour. In SQL Server, this is a bit ugly, but easy to do:
SELECT dateadd(hour, datediff(hour, 0, TimeStamp), 0) as TimeStampHour, Count(*)
FROM [TEST].[dbo].[data]
GROUP BY dateadd(hour, datediff(hour, 0, TimeStamp), 0)
ORDER BY dateadd(hour, datediff(hour, 0, TimeStamp), 0);