November 16, 2017 at 3:57 pm
Comments posted to this topic are about the item Reporting by date - DateTrunc
December 19, 2017 at 2:22 pm
See the example below in which it appears the function does perform as expected:
December 20, 2017 at 3:39 am
I'm not sure what you'd expect the function to return when the number is not a divisor of the number of units in the implied containing period.
For example, if we have an interval of months, we are typically looking at how they repeat within a year, and it makes sense to have n as 1,2,3,4 or 6. So, for example, 3 months = 1 quarter and we get 4 per year. But if we asked for a period of 5 months then the generated dates will not be aligned to a year and the results will not have an obvious meaning. Likewise we might be looking at minutes within an hour, so 1,2,3,4,5,6,10,12,15,20,30 make sense; for example to report an activity for every 5 minutes within the hour makes sense, but asking for the value in groups of 13 minutes would not as it wouldn't repeat each hour.
In your example, 3 quarters (9 months) is not a divisor of a year, aka 4 quarters aka 12 months, so the result has no intuitive meaning.
I should have added this as a conceptual limitation of the function.
December 20, 2017 at 5:26 am
I think this a serious limitation for a great function but why not do something like this:
--where zero is the current value of the unit
CREATE FUNCTION [dbo].[DateTrunc]
(
@Unit varchar(10),
@DateTime datetime,
@Num int
)
RETURNS DateTime
AS
BEGIN
RETURN
case left(@Unit, 1)
when 'y' then dateadd(yy, datediff(yy, 0, @DateTime) -@num, 0)
when 'q' then dateadd(qq, datediff(qq, 0, @DateTime) -@num, 0)
when 'w' then dateadd(wk, datediff(wk, 0, @DateTime) -@num, 0)
when 'd' then dateadd(dd, datediff(dd, 0, @DateTime) -@num, 0)
when 'h' then dateadd(hh, datediff(hh, 0, @DateTime) -@num, 0)
when 'n' then dateadd(mi, datediff(mi, 0, @DateTime) -@num, 0)
when 's' then dateadd(ss, datediff(ss, dateadd(dd, datediff(dd, 0, @DateTime), 0), @DateTime) -@num, dateadd(dd, datediff(dd, 0, @DateTime), 0))
when 'm' then
Case when @Unit in ('month', 'mm', 'm') then dateadd(mm, datediff(mm, 0, @DateTime) -@num, 0)
when @Unit in ('minute', 'mi') then dateadd(mi, datediff(mi, 0, @DateTime) -@num, 0)
else null
end
else null
end;
END
GO
December 20, 2017 at 6:57 am
Perhaps I'm misunderstanding how I should use the variant of the function you suggest.
Here is an example of how I would expect my version to be used, in this case for multiples of minutes.
Taking the 5th line as an example, if I wanted to have times truncated to 5 minute intervals I would use n = 5.declare @dt as datetime; set @dt = '2017-12-20 13:17:44'
select '1 min slot', dbo.datetrunc('mi', @dt, 1)
select '2 min slot', dbo.datetrunc('mi', @dt, 2)
select '3 min slot', dbo.datetrunc('mi', @dt, 3)
select '4 min slot', dbo.datetrunc('mi', @dt, 4)
select '5 min slot', dbo.datetrunc('mi', @dt, 5)
select '6 min slot', dbo.datetrunc('mi', @dt, 6)
select '10 min slot', dbo.datetrunc('mi', @dt, 10)
select '15 min slot', dbo.datetrunc('mi', @dt, 15)
select '20 min slot', dbo.datetrunc('mi', @dt, 20)
Which gives me (as the minutes portion)
1 -> 17, 2 -> 16, 3 -> 15, 4 -> 16, 5 -> 15, 6 -> 12, 10 ->10, 15 -> 15, 20 -> 0 - that is the portion truncated to multiples of n minutes.
I can't see how I can get your flavour to do that.
Here's another example. I want to see how many jobs have been created in each 5 minute interval for each day in some particular weekselect dbo.datetrunc('d', CreatedOn, 5), dbo.datetrunc('mi', CreatedOn, 5), count(*)
from tblJob
where CreatedOn >='2017-02-01' and CreatedOn < '2017-02-08'
group by dbo.datetrunc('d', CreatedOn, 5), dbo.datetrunc('mi', CreatedOn, 5)
order by 1,2
This lists the day, the time truncated to 5 minutes, and the count.
In my mind there is no meaning to a 7 or 23 minute truncation - if I wanted that then I'd use datepart and do a relevant division on that within the context of the date truncated to an hour or day.
Viewing 5 posts - 1 through 4 (of 4 total)
You must be logged in to reply to this topic. Login to reply