I'm trying to write a SQL query (SQL Server) and part of it is determining the number of minutes per hour between two datetimes.
Example: 11/1/2018 09:05 - 11/1/2018 13:15
Hour 09: 55 minutes
Hour 10: 60 minutes
Hour 11: 60 minutes
Hour 12: 60 minutes
Hour 13: 15 minutes
These would then get put into a temp table and grouped by some other data which will then be used to calculate dollar amounts from these minutes.
Is there a way to accomplish something like this via SQL that isn't too slow or laborious?
Thanks!
I think a recursive CTE is possibly the best approach:
with cte as (
select startTime, endTime,
startTime_hour as hourStart,
(case when endTime < dateadd(hour, 1, startTime_hour) then endTime
else dateadd(hour, 1, startTime_hour)
end) as hourEnd
from (select t.*,
dateadd(hour, datediff(hour, 0, startTime), 0) as startTime_hour
from t
) t
union all
select startTime, endTime,
dateadd(hour, 1, hourStart) as hourStart,
(case when endTime < dateadd(hour, 2, hourStart) then endTime
else dateadd(hour, 2, hourStart)
end) as endHour
from cte
where hourEnd < endTime
)
select cte.hourStart,
(case when hourStart > startTime then datediff(minute, hourStart, hourEnd) else datediff(minute, startTime, hourEnd) end) as minutes
from cte
order by hourStart;
Here is a db<>fiddle.
Here is an alternative dynamic solution that you can work with two parameters (start/end dates) only:
create table #temp
([hour] int, [minutes] int)
declare #startTime datetime='11/1/2018 09:05'
declare #EndTime datetime='11/1/2018 13:15'
declare #tempStartTime datetime = #startTime
declare #nextTimeRounded datetime
declare #hourdiff int = DATEDIFF(HOUR,#startTime,#EndTime)
declare #counter int = DATEPART(HH,#startTime)
declare #limit int = #counter + #hourdiff + 1
while #counter < #limit
begin
insert into #temp ([hour]) values (#counter)
set #nextTimeRounded= (dateadd(hour,
1 + datepart(hour, #tempStartTime),
cast(convert(varchar(10),#tempStartTime, 112) as datetime))
)
if #nextTimeRounded > #EndTime
begin
set #nextTimeRounded = #EndTime
end
update #temp
set [minutes] = (case when DATEDIFF(MINUTE,#tempStartTime,#nextTimeRounded)=0 then 60 else DATEDIFF(MINUTE,#tempStartTime,#nextTimeRounded) end)
where [hour] = #counter
set #counter = #counter + 1
set #tempStartTime = DATEADD(MINUTE,DATEDIFF(MINUTE,#tempStartTime,#nextTimeRounded),#tempStartTime);
end
select * from #temp
Sample Data
Below, we pump four time ranges, with associated values, into a table. All time ranges are different, but the first two are 10h 30m apart. The second two are 9h 45m apart.
declare #times table (
startTime time,
endTime time,
val float
);
insert #times values
('2018-10-01 01:00:00', '2018-10-01 10:45:00', 7),
('2018-10-02 01:00:00', '2018-10-02 10:45:00', 8),
('2018-10-01 01:00:00', '2018-10-01 11:30:00', 1),
('2018-10-02 01:00:00', '2018-10-02 11:30:00', 3);
Solution
You can use the 'datediff' function to aggregate as you so desire. Use the modulo operator to convert your minutes into just the minutes that remain over when whole hours are discounted.
select ap.h,
ap.m,
sumVal = sum(val)
from #times
cross apply (select
h = datediff(hour, startTime, endTime),
m = datediff(minute, startTime, endTime) % 60
) ap
group by ap.h,
ap.m
Related
I've table in my schema => TimeSlot, which has several attributes with id, start time, end time.
So the question is: how can you make a script that will make 23/24 records, for which starttime - endtime, for example, 0.00-1.00, 1.00-2.00, etc.
i've tried this with some example, but can not figure out how to code it(not even know if it is correct):
SET NOCOUNT ON
TRUNCATE TABLE TimeSlot
DECLARE #InitHour TIME(0) = '1'
DECLARE #FinitHour TIME(0) = '24'
WHILE #InitHour < #FinitHour
BEGIN
INSERT INTO TimeSlot(
TimeSlotID, StartTime, EndTime
)
SELECT
The simplest solution is to generate a sequence of numbers, you can do this using any table that has the required number of rows and use row_number for an arbitrary number sequence.
with seq as (
select top(24) Row_Number() over (order by (select null))-1 n
from sys.syscomments
)
insert into TimeSlot (StartTime, EndTime)
select Convert(time,Concat(n,':00')) StartTime,
DateAdd(hour, 1, Convert(time,Concat(n,':00'))) EndTime
from seq
The basic answer to your question is that you need to generate the results somehow. There are lots of methods. If you have a table lying around with at least 24 rows, then you can use that. A "native" SQL method uses a recursive CTE>
Then, the TimeSlotID should be generated automatically (typically as an identity column):
WITH CTE as (
SELECT convert(datetime, '01:00:00') as StartTime
UNION ALL
SELECT DATEADD(HOUR, 1, StartTime)
FROM CTE
WHERE StartTime < convert(datetime, '23:00:00')
)
INSERT INTO TimeSlot (StartTime, EndTime)
SELECT StartTime, DATEADD(HOUR, 1, StartTime)
FROM CTE;
Fiddle here: http://sqlfiddle.com/#!18/9bac2/12
Please, check this solution: http://sqlfiddle.com/#!18/4cb97/17
set nocount on
declare #hour int = 0
while #hour <= 23
begin
declare #time datetime =
convert(
datetime,
convert(varchar, #hour) + ':00:00'
)
insert into TimeSlot(StartTime, EndTime)
values(
#time, dateadd(hour, 1, #time)
)
set #hour = #hour + 1
end
select *
from TimeSlot
I’ve a table which contains different time ranges:
Id Start Time End Time Points
1 0:00 3:00 10
2 3:01 6:00 20
3 6:01 23:59 30
Now I need to calculate the points achieved between two date ranges with respect to time specified.
Start date = 11/9/2016 18:17:00 and
End date = 11/10/2016 01:20:00
I need to calculate the sum of points gained between these two dates.
The time of start date that is 18:17 falls under Id 3, whose point is 30. So the calculation will be,
18:17 to 23:59 -> 6 hrs -> 6 * 30 = 180 points
The end time 01:20 falls under Id 1
0:00 to 1:20 -> 2 hrs
(if minute is greater than zero, it is rounded to next hour, ie; 2) -> 2 * 10 = 20 points
So the total points gained will be 200 points.
Taking the time difference, does not help me, if the start and end date difference is greater than one day.
Table Structure:
Id - int,
StartTime - time(7),
EndTime - time(7),
Points - int
How to write a query for this using SQL?
This question was good.
You can as the below:
DECLARE #Tbl TABLE (Id INT, StartTime TIME, EndTime TIME, Points INT)
INSERT INTO #Tbl
VALUES
(1, '0:00', '3:00' , 10),
(2, '3:01', '6:00' , 20),
(3, '6:01', '23:59', 30)
DECLARE #StartDate DATETIME = '2016.11.09 18:17:00'
DECLARE #EndDate DATETIME = '2016.11.10 01:20:00'
;WITH CTE
AS
(
SELECT 1 AS RowId, #StartDate CurrentDate, 0 Point, #StartDate DateVal UNION ALL
SELECT
A.RowId ,
IIF((A.CurrentDate + A.EndTime) > #EndDate, #EndDate, DATEADD(MINUTE, 1, (A.CurrentDate + A.EndTime))) AS CurrentDate,
A.Points,
IIF((A.CurrentDate + A.EndTime) > #EndDate, #EndDate, (A.CurrentDate + A.EndTime)) DateVal
FROM
(
SELECT
C.RowId + 1 AS RowId,
CAST(CAST(CurrentDate AS DATE) AS DATETIME) CurrentDate,
CAST((SELECT T.EndTime FROM #Tbl T WHERE CAST(CurrentDate AS TIME) BETWEEN T.StartTime AND T.EndTime) AS DATETIME) AS EndTime,
(SELECT T.Points FROM #Tbl T WHERE CAST(CurrentDate AS TIME) BETWEEN T.StartTime AND T.EndTime) AS Points,
C.CurrentDate AS TempDate
FROM CTE C
) A
WHERE
A.TempDate <> IIF((A.CurrentDate + A.EndTime) > #EndDate, #EndDate, DATEADD(MINUTE, 1, (A.CurrentDate + A.EndTime)))
), CTE2
AS
(
SELECT
C.RowId ,
C.CurrentDate ,
C.Point ,
C.DateVal,
DATEDIFF(MINUTE, LAG(C.DateVal) OVER (ORDER BY C.RowId), C.DateVal) MinuteOfDateDiff
FROM
CTE C
)
SELECT
SUM(CEILING(C.MinuteOfDateDiff * 1.0 / 60.0) * C.Point)
FROM
CTE2 C
Result: 200
i got a dataset in SQL Server Management Studio. The data looks like the following. i have a identifier for each people userID, date of the record, start timestartime and finish time endtime.
UserID date startime endtime
1 20110203 09:30 09:35
1 20110203 09:31 09:38
1 20110203 10:03 10:05
1 20110203 10:04:00 10:35:00
2 20110203 11:02 11:05
For each people, i want check if there is any overlapping time. If there is, I want to keep the smallest startime and largest endtime. if no overlapping time, I keep the original data. In addition, I want to calculate the duration of maxi endtime and smallest startime.
The result I want should looks like the following. Can anyone teach me how to code this please.
UserID date startime endtime diff
1 20110203 09:30 09:38 00:08
1 20110203 10:03 10:35 00:02
2 20110203 11:02 11:05 00:03
It seems that SELECT with CTE needs to recursively merge undetermined number of rows. In that case I would prefer safe CURSOR based solution:
DECLARE #t TABLE
(
UserId int,
[Date] date,
StartTime time,
EndTime time
);
INSERT INTO #t VALUES
(1, '2011-02-03', '09:30:00', '09:35:00'),
(1, '2011-02-03', '09:31:00', '09:38:00'),
(1, '2011-02-03', '09:36:00', '09:41:00'),
(1, '2011-02-03', '09:40:00', '09:45:00'),
(1, '2011-02-03', '09:42:00', '09:43:00'),
(1, '2011-02-03', '10:03:00', '10:05:00'),
(2, '2011-02-03', '11:02:00', '11:05:00'),
(1, '2011-02-03', '12:00:00', '12:05:00'),
(1, '2011-02-03', '12:04:00', '12:06:00');
------------------
DECLARE #result TABLE
(
UserId int,
[Date] date,
StartTime time,
EndTime time
)
DECLARE cur CURSOR FOR
SELECT UserId, [Date], StartTime, EndTime
FROM #t
ORDER BY UserId, [Date], StartTime;
DECLARE #UserId int
DECLARE #Date date
DECLARE #StartTime time
DECLARE #EndTime time
DECLARE #LastUserId int
DECLARE #LastDate date
DECLARE #LastStartTime time
DECLARE #LastEndTime time
OPEN cur
FETCH NEXT FROM cur INTO #UserId, #Date, #StartTime, #EndTime
SET #LastUserId = #UserId
SET #LastDate = #Date
SET #LastStartTime = #StartTime
SET #LastEndTime = #EndTime
WHILE ##FETCH_STATUS = 0
BEGIN
IF #UserId = #LastUserId AND #Date = #LastDate AND #StartTime <= #LastEndTime
SET #LastEndTime = CASE WHEN #LastEndTime > #EndTime THEN #LastEndTime ELSE #EndTime END
ELSE
BEGIN
INSERT #result(UserId, [Date], StartTime, EndTime) VALUES (#LastUserId, #LastDate, #LastStartTime, #LastEndTime)
SET #LastUserId = #UserId
SET #LastDate = #Date
SET #LastStartTime = #StartTime
SET #LastEndTime = #EndTime
END
FETCH NEXT FROM cur INTO #UserId, #Date, #StartTime, #EndTime
END
INSERT #result(UserId, [Date], StartTime, EndTime) VALUES (#LastUserId, #LastDate, #LastStartTime, #LastEndTime)
CLOSE cur
DEALLOCATE cur
SELECT UserId,
[Date],
StartTime,
EndTime,
CAST(DATEADD(second,DATEDIFF(second,StartTime,EndTime),'2000-01-01') AS time) Diff
FROM #result
which returns
1 2011-02-03 09:30:00.0000000 09:45:00.0000000 00:15:00.0000000
1 2011-02-03 10:03:00.0000000 10:05:00.0000000 00:02:00.0000000
1 2011-02-03 12:00:00.0000000 12:06:00.0000000 00:06:00.0000000
2 2011-02-03 11:02:00.0000000 11:05:00.0000000 00:03:00.0000000
Following a redesigned Version of my previous cte Approach. However, it will still have Problems if there are multiple records for the same user with identical start time... didn't have time to fix that one, but as far as I understood this is not possible in the described process!?
--
-- This part is temporary and has to be replaced by your tables
-- There several more records included now
-- There is still a glitch if the starttime is identical for two records - but as far as I understood, this is not possible in the described case?
--
declare #t table (userid int, date int, starttime time, endtime time);
insert into #t values (1, 20110203, '09:30:00', '09:35:00'), (1, 20110203, '09:31:00', '09:38:00'), (1, 20110203, '09:36:00', '09:41:00'), (1, 20110203, '10:03:00', '10:05:00'),(1, 20110203, '10:04:00', '10:35:00'),
(2, 20110203, '11:02:00', '11:05:00'), (2, 20110203, '11:03:00', '11:20:00'), (2, 20110203, '11:04:00', '11:35:00'), (2, 20110203, '13:02:00', '13:05:00'), (2, 20110203, '13:04:00', '13:15:00');
--
-- First cte: selects all start and endtimes and their - if existing - "overlaps"; recursive cte
--
WITH cte AS(
SELECT 1 lvl, a.userid
,CASE WHEN a.starttime <= ISNULL(b.starttime, a.starttime) THEN a.starttime ELSE b.starttime END AS starttime
,CASE WHEN a.endtime >= ISNULL(b.endtime, a.endtime) THEN a.endtime ELSE b.endtime END AS endtime
FROM #t as a
LEFT OUTER JOIN #t AS b ON b.userid = a.userid
AND b.starttime < a.starttime
AND b.endtime > a.starttime
UNION ALL
select a.lvl+1, a.userid
,CASE WHEN a.starttime <= ISNULL(b.starttime, a.starttime) THEN a.starttime ELSE b.starttime END AS xStart
,CASE WHEN a.endtime >= ISNULL(b.endtime, a.endtime) THEN a.endtime ELSE b.endtime END AS xEnd
from cte as a
INNER JOIN #t AS b ON b.userid = a.userid
AND b.starttime < a.starttime
AND b.endtime > a.starttime
),
--
-- Second cte: get the max. lvl result per user from the recursive cte
--
cteUserMaxLvl AS (
SELECT userid, max(lvl) AS MaxLvl
FROM cte
GROUP BY userid
),
--
-- third cte: get the rows matching the max.lvl; their timespan should cover the desired min. start and max. end
--
cteNoMoreOverlap AS(
SELECT a.userid, starttime, endtime
FROM cte AS a
JOIN cteUserMaxLvl AS b ON a.userid = b.userid AND a.lvl = b.MaxLvl
)
--
-- Select the rows from the "No more overlap" cte
--
SELECT userid, starttime, endtime
FROM cteNoMoreOverlap
UNION ALL
--
-- And finally select all rows, which are not covered by the previously selected timespan
--
SELECT a.userid, min(a.starttime) AS starttime, max(a.endtime) AS endtime
FROM cte AS a
JOIN cteNoMoreOverlap AS b ON a.userid = b.userid AND a.starttime NOT BETWEEN b.starttime AND b.endtime
GROUP BY a.userid
order by userid, starttime, endtime
I believe when you say overlapping time, you are saying within the same hour on the same day. If that is what you mean, following solution might work. Attached is the screenshot of my results.
CREATE TABLE #OverlappingDates
(
UserID INT
, [date] DATE
, starttime VARCHAR(5)
, endtime VARCHAR(5)
);
INSERT INTO #OverlappingDates
( UserID, date, starttime, endtime )
VALUES ( 1 -- UserID - int
, '20110203' -- date - date
, '09:30' -- starttime - time
, '09:35' -- endtime - time
),
( 1 -- UserID - int
, '20110203' -- date - date
, '09:31' -- starttime - time
, '09:38' -- endtime - time
),
( 1 -- UserID - int
, '20110203' -- date - date
, '10:03' -- starttime - time
, '10:05' -- endtime - time
),
( 2 -- UserID - int
, '20110203' -- date - date
, '11:02' -- starttime - time
, '11:05' -- endtime - time
),
( 2 -- UserID - int
, '20110203' -- date - date
, '11:05' -- starttime - time
, '11:15' -- endtime - time
),
( 2 -- UserID - int
, '20110203' -- date - date
, '11:05' -- starttime - time
, '12:00' -- endtime - time
);
WITH cte
AS ( SELECT UserID
, date
, MIN(starttime) AS StartTime
, MAX(endtime) AS EndTime
FROM #OverlappingDates
GROUP BY UserID
, date
, LEFT(starttime, 2)
, LEFT(endtime, 2)
)
SELECT cte.UserID
, cte.date
, cte.StartTime
, cte.EndTime
, ( RIGHT('0'
+ CAST(( DATEDIFF(SECOND,
( CAST(CONCAT(( CAST(cte.[date] AS VARCHAR(10)) ),
' ', cte.StartTime) AS DATETIME) ),
( CAST(CONCAT(( CAST(cte.[date] AS VARCHAR(10)) ),
' ', cte.EndTime) AS DATETIME) )) )
/ 3600 AS VARCHAR(2)), 2) + ':' + RIGHT('0'
+ CAST(( ( DATEDIFF(SECOND,
( CAST(CONCAT(( CAST(cte.[date] AS VARCHAR(10)) ),
' ',
cte.StartTime) AS DATETIME) ),
( CAST(CONCAT(( CAST(cte.[date] AS VARCHAR(10)) ),
' ', cte.EndTime) AS DATETIME) )) )
/ 60 ) % 60 AS VARCHAR(2)),
2) ) AS Diff
FROM cte;
I have a table with a DATETIME column 'Start' and a DATETIME column 'End'. I want to return the number of minutes between the start and the end (End is always after than Start). Usually I'd just use 'DateDiff()' but this time I need to exclude another date range. For example - From Tuesday at 9am until Wednesday at 6pm, of each week, should be ignored.
If a row has a Start of Tuesday at 8am and an End of Wednesday at 7pm - the elapsed time should be two hours (120 minutes) - because of the ignored date range.
I'm having trouble coming up with a decent way of doing this and my searching online hasn't found quite what I'm looking for. Can someone help me along?
Try This:
--total time span to calculate difference
DECLARE #StartDate DATETIME = '2015-11-10 8:00:00AM',
#EndDate DATETIME = '2015-11-11 7:00:00PM'
--get the day of week (-1 because sunday is counted as first weekday)
DECLARE #StartDayOfWeek INT = (SELECT DATEPART(WEEKDAY, #StartDate)) -1
DECLARE #EndDayOfWeek INT = (SELECT DATEPART(WEEKDAY, #EndDate)) -1
--set the time span to exclude
DECLARE #InitialDOWToExclude TINYINT = 2
DECLARE #InitialTODToExclude VARCHAR(100) = '9:00:00 AM'
DECLARE #EndDOWToExclude TINYINT = 3
DECLARE #EndTODToExclude VARCHAR(100) = '6:00:00 PM'
--this will be the final output in hours
DECLARE #ElapsedHours INT = (SELECT DATEDIFF(HOUR, #StartDate, #EndDate))
DECLARE #WeeksBetween INT = (SELECT DATEDIFF(WEEK, #StartDate, #EndDate))
DECLARE #Iterator INT = 0
WHILE (#Iterator <= #WeeksBetween)
BEGIN
DECLARE #InitialDaysBetween INT = #StartDayOfWeek - #InitialDOWToExclude
DECLARE #StartDateToExclude DATETIME = (SELECT DATEADD(DAY, #InitialDaysBetween, DATEADD(WEEK, #Iterator, #StartDate)))
SET #StartDateToExclude =CAST(DATEPART(YEAR, #StartDateToExclude) AS VARCHAR(100))
+ CAST(DATEPART(MONTH, #StartDateToExclude) AS VARCHAR(100))
+ CAST(DATEPART(DAY, #StartDateToExclude) AS VARCHAR(100))
+ ' '
+ CAST(#InitialTODToExclude AS VARCHAR(100))
DECLARE #EndDaysBetween INT = #EndDayOfWeek - #EndDOWToExclude
DECLARE #EndDateToExclude DATETIME = (SELECT DATEADD(DAY, #EndDaysBetween, DATEADD(WEEK, #Iterator, #EndDate)))
SET #EndDateToExclude =CAST(DATEPART(YEAR, #EndDateToExclude) AS VARCHAR(100))
+ CAST(DATEPART(MONTH, #EndDateToExclude) AS VARCHAR(100))
+ CAST(DATEPART(DAY, #EndDateToExclude) AS VARCHAR(100))
+ ' '
+ CAST(#EndTODToExclude AS VARCHAR(100))
SET #ElapsedHours = #ElapsedHours - DATEDIFF(HOUR, #StartDateToExclude, #EndDateToExclude)
SET #Iterator = #Iterator + 1
END
SELECT #ElapsedHours
This might get you pretty close..
DECLARE #Table1 TABLE ([Id] INT, [Start] DATETIME, [End] DATETIME)
INSERT INTO #Table1 VALUES
(1, '2015-11-08 00:00:00', '2015-11-10 21:45:38'),
(2, '2015-11-09 00:00:00', '2015-11-11 21:45:38')
;
-- hours to exclude
WITH excludeCTE AS
(
SELECT *
FROM (VALUES('Tuesday', 9, 0), ('Wednesday', 0, 0)) AS T([Day], [Hour], [Amount])
UNION ALL
SELECT [Day], [Hour] + 1, [Amount]
FROM excludeCTE
WHERE ([Day] = 'Tuesday' AND [Hour] < 23) OR ([Day] = 'Wednesday' AND [Hour] < 18)
),
-- all hours between start and end
dateCTE AS
(
SELECT [Id],
[Start],
[End],
DATENAME(weekday, [Start])[Day],
DATENAME(hour, [Start])[Hour]
FROM #Table1 t
UNION ALL
SELECT cte.[Id],
DATEADD(HOUR, 1, cte.[Start]),
cte.[End],
DATENAME(weekday, DATEADD(HOUR, 1, cte.[Start]))[Day],
DATENAME(hour, DATEADD(HOUR, 1, cte.[Start]))[Hour]
FROM #Table1 t
JOIN dateCTE cte ON t.Id = cte.Id
WHERE DATEADD(HOUR, 1, cte.[Start]) <= t.[End]
)
SELECT t.[Id],
t.[Start],
t.[End],
SUM(COALESCE(e.[Amount], 1)) [Hours]
FROM #Table1 t
INNER JOIN dateCTE d ON t.[Id] = d.[Id]
LEFT JOIN excludeCTE e ON d.[Day] = e.[Day] AND d.[Hour] = e.[Hour]
GROUP BY t.[Id],
t.[Start],
t.[End]
OPTION (MAXRECURSION 0) -- allow more than 100 hours
Putting the additional constraint that there can only be one excluded range between any two date
CREATE TABLE worktable (
_Id INT
, _Start DATETIME
, _End DATETIME
);
INSERT INTO worktable VALUES
(1, '2015-11-09 00:00:00', '2015-11-09 00:45:00') -- Start and End before excluded range
, (2, '2015-11-09 00:00:00', '2015-11-11 21:45:00') -- Start before, End after
, (3, '2015-11-09 00:00:00', '2015-11-10 21:00:00') -- Start before, End between
, (4, '2015-11-10 10:00:00', '2015-11-11 10:00:00') -- Start between, End between
, (5, '2015-11-10 10:00:00', '2015-11-11 21:45:00') -- Start between, End after
With getDates As (
SELECT _Id
, a = _Start
, b = _End
, c = DATEADD(hh, 9
, DATEADD(DAY,DATEDIFF(DAY, 0, _Start) / 7 * 7
+ 7 * Cast(Sign(1 - DatePart(dw, _Start)) + 1 as bit), 1))
, d = DATEADD(hh, 18
, DATEADD(DAY,DATEDIFF(DAY, 0, _Start) / 7 * 7
+ 7 * Cast(Sign(1 - DatePart(dw, _Start)) + 1 as bit), 2))
FROM worktable
), getDiff As (
SELECT c_a = DATEDIFF(mi, a, c)
, c_b = DATEDIFF(mi, b, c)
, b_d = DATEDIFF(mi, d, b)
, a, b, c, d, _id
FROM getDates
)
Select _id
, (c_a + ABS(c_a)) / 2
- (c_b + ABS(c_b)) / 2
+ (b_d + ABS(b_d)) / 2
FROM getDiff;
c is the date of the first Tuesday after the start date (Find the next occurance of a day of the week in SQL) you may need to adjust the last value depending from DATEFIRST
d is the date of the first Wednesday after the start date in the same week of c
Cast(Sign(a - b) + 1 as bit) is 1 if a is more than or equal b, 0 otherwise
(x + ABS(x)) / 2 is x if not negative, otherwise 0
Given that the formula to get the elapsed time with the excluded range is:
+ (Exclusion Start - Start) If (Start < Exclusion Start)
- (Exclusion Start - End) If (End < Exclusion Start)
+ (End - Exclusion End) If (Exclusion End < End)
-- excluded range (weekday numbers run from 1 to 7)
declare #x datetime = /*ignore*/ '1900012' + /*start day # and time*/ '3 09:00am';
declare #y datetime = /*ignore*/ '1900012' + /* end day # and time*/ '4 06:00pm';
-- normalize date to 1900-01-21, which was a Sunday
declare #s datetime =
dateadd(day, 19 + datepart(weekday, #start), cast(cast(#start as time) as datetime));
declare #e datetime =
dateadd(day, 19 + datepart(weekday, #end), cast(cast(#end as time) as datetime));
-- split range into two parts, one before #x and the other after #y
-- each part collapses to zero when #s and #e respectively fall between #x and #y
select (
datediff(second, -- diff in minutes would truncate so count seconds
case when #s < #x then #s else #x end, -- minimum of #s, #x
case when #e < #x then #e else #x end -- minimum of #e, #x
) +
datediff(second,
case when #s > #y then #s else #y end, -- maximum of #s, #y
case when #e > #y then #e else #y end -- maximum of #e, #y
)
) / 60; -- convert seconds to minutes, truncating with integer division
I glanced at the earlier answers and I thought that surely there was something more straightforward and elegant. Perhaps this is easier to understand and one clear advantage over some solutions is that it's trivial to change the excluded range and that range doesn't have to be limited to a single day.
I'm assuming that your dates never span more than one regular calendar week. It wouldn't be too difficult to extend it to handle more though. One approach would be to handle starting and ending partial weeks plus the full weeks in the middle.
Imagine that your start time is 8:59:30am and your end time is 6:00:30pm. In such a case I'm figuring that you'd want to accumulate the half minutes on each side to get a full minute in total after subtracting the 9-6 block. If you use datediff(minute, ...) you would be truncating the partial minutes and never get the chance to add them together: so that's why I count seconds and then divide by sixty at the end. Of course, if you're only dealing in whole minutes then you won't need to do it that way.
I've chosen my reference date somewhat arbitrarily. At first I thought it might possibly be handy to look at a real and convenient date on the calendar but ultimately it only really matters that it falls on a Sunday. So I settled on the first Sunday falling on a date ending in the digit 1.
Note that the solution also relies on datefirst being set to Sunday. That could be tweaked or made more portable if necessary.
I have a table with a datetime column in it, consider it an event log for simple, analogous purposes.
I want to produce a report detailing the average number of events that occur at each time of day, to 30 min accuracy.
so the logic is,
get just the time component of each date
round the time to the nearest 30 min window (it can be floored, i.e. 00:29 -> 00:00)
count these (grouped by date)
average all these counts over all days
I also don't want to have any time holes in my data, for example, if nothing occurred in the 00:00 - 00:30 range, i want to report a 0, rather than having a missing row.
How can I achieve this?
WITH TestDates (date) AS (
SELECT CONVERT(DATETIME, '2011-11-15 10:00') UNION ALL
SELECT CONVERT(DATETIME, '2011-11-15 11:31') UNION ALL
SELECT CONVERT(DATETIME, '2011-11-16 10:00')
-- CTE to generate 4 million rows with a sequential integer starting at 0
), GeneratedRows (seq) AS (
SELECT ROW_NUMBER() OVER (ORDER BY N1.number) - 1
FROM master..spt_values AS N1
CROSS JOIN master..spt_values AS N2
WHERE N1.name IS NULL
AND N2.name IS NULL
), RoundedTestDates (date) AS (
SELECT CASE
-- Subtract the minute part
WHEN DATEPART(MINUTE, date) < 25 THEN DATEADD(MINUTE, -1 * DATEPART(MINUTE, date), date)
-- Subtract the minute part, then add an hour
WHEN DATEPART(MINUTE, date) >= 45 THEN DATEADD(HOUR, 1, DATEADD(MINUTE, -1 * DATEPART(MINUTE, date), date))
-- Subtract the minute part, then add an half-hour
ELSE DATEADD(MINUTE, 30, DATEADD(MINUTE, -1 * DATEPART(MINUTE, date), date))
END
FROM TestDates
)
SELECT rounded_date = GeneratedPeriod.date
, ocurrences = COUNT(RoundedTestDates.date)
FROM (SELECT DATEADD(MINUTE, 30 * seq, (SELECT MIN(date) FROM RoundedTestDates))
FROM GeneratedRows
) AS GeneratedPeriod (date)
LEFT JOIN RoundedTestDates
ON GeneratedPeriod.date = RoundedTestDates.date
WHERE GeneratedPeriod.date <= (SELECT MAX(date) FROM RoundedTestDates)
GROUP BY GeneratedPeriod.date
ORDER BY 1
Here is the code you need: (tested in sql2008 and works fine!)
-- Table with the 48 30mins periods of the day
CREATE TABLE #Periods
(
Num INT
)
DECLARE #idt INT
SET #idt = 1
WHILE (#idt <= 48)
BEGIN
INSERT INTO #Periods VALUES (#idt)
SET #idt = #idt + 1
END
--Average of the count for each period on all days.
SELECT DayTable.Num, AVG(CAST(DayTable.DayCount AS DECIMAL))
FROM
( --Total incidents for each interval on each day.
SELECT CAST(FLOOR(CAST(#MyLog.LogDate AS FLOAT)) AS DATETIME) AS DayWithOutTime,
#Periods.Num AS Num,
COUNT(#MyLog.ID) AS DayCount
FROM #Periods LEFT JOIN #MyLog
ON #Periods.Num = (DATEPART(hh, #MyLog.LogDate)*60 + DATEPART(mi,#MyLog.LogDate))/30
GROUP BY CAST(FLOOR(CAST(#MyLog.LogDate AS FLOAT)) AS DATETIME),
#Periods.Num
) AS DayTable
GROUP BY DayTable.Num
DROP TABLE #Periods
Where #NyLog is the table where your datetime is. It shows the count of incidences for each 30min period. The Period 1 is 00:00 -> 00:30 and Period 48 is 23:30 -> 24:00.
In sybase sql is something like this, in sql-server you might need to do some changes but not much :)
create procedure Test #startDay varchar(8), #endDay varchar(8)
as
declare #ocurrence int
declare #numberOfDays int
select #numberOfDays = 0
create table #intervals (
interval_hour int,
interval_min_minute int,
interval_max_minute int,
ocurrences int
)
create table #insertions (
hour int,
minute int
)
declare #hour int, #minute int
select #hour = 0
-- create the intervals
while (#hour <> 24)
begin
insert into #intervals values(#hour,0,29,0)
insert into #intervals values(#hour,30,59,0)
select #hour = #hour + 1
end
while(#startDay <> #endDay)
begin
insert into #insertions
select datepart(hh, *yourcolumn*), datepart(mm, *yourcolumn*) from *yourdb..yourtable* where convert(varchar(8), *yourcolumn*, 112) = #startDay
select #startDay = convert(varchar(8), dateadd(dd, 1, convert(datetime, #startDay, 112)), 112)
select #numberOfDays = #numberOfDays + 1
end
declare cursor1 cursor for
select hour, minute from #insertions
open cursor1
fetch cursor1 into #hour, #minute
while (##sqlstatus=0)
begin
update #intervals
set i.ocurrences = i.ocurrences + 1
from #intervals i
where interval_hour = #hour and #minute between interval_min_minute and interval_max_minute
fetch cursor1 into #hour, #minute
end
close cursor1
select interval_hour 'hour', interval_min_minute 'min minute', interval_max_minute 'max minute', ocurrences,
case when ocurrences > 0 then convert(float, ocurrences) / convert(float, #numberOfDays) else 0 end 'ocurrences average' from #intervals
drop table #intervals
drop table #insertions
go
What I've done is use an auxiliary table of numbers (a 1 column table with number 1 to 1 million) and join to it, adding the value of the number with the dateadd function to the midnight of the date.
since you want 30 minute intervals, then you want to use the dateadd(minute, number*30, yourdate) where number <= 48 (since there are 1440 minutes in a day)/30 = 48 intervals. This will create your time intervals.
Then simply count your occurrences that happen in between the time intervals.