We want to group data daily but with start hour let's say 8am .. 7am next day:

Normally we're using for daily group by with start hour = 0:

select
EXTRACT(YEAR FROM V.DateTimeField) as YY,
EXTRACT(MONTH FROM V.DateTimeField) as MM,
EXTRACT(DAY FROM V.DateTimeField) as DD,
EXTRACT(HOUR FROM V.DateTimeField) as HH,
MIN(V.DateTimeField) as CAS_MIN

FROM MYTABLE as V
Where DateTimeField>=:StartTime AND DateTimeField<=:EndTime

GROUP BY DATEPART(YEAR , V.DateTimeField),
DATEPART(MONTH , V.DateTimeField),
DATEPART(DAY , V.DateTimeField)
order by YY,MM,DD asc

How can we make SQL that would group for each day but the start hour as variable ?

The first thing that comes to mind is to use DATEADD in the GROUP BY. Let's say that you had an int parameter named start_hour that would be 7 for 7 AM, 8 for 8 AM, etc. The GROUP BY might be:

Then, for example, if @start_hour was 8, 7 AM on 04/24 would become 11 PM on 04/23 (so it would be grouped with 04/23), while 8 AM on 04/24 would become midnight on 04/24.

I haven't tested that, but it seems like it should work.

If your start "hour" could really be any time, like 7:20 AM or 8:45 AM, then you could convert the parameter to minutes or seconds and use that with DATEADD.

James

0

vojkocendakAuthor Commented:

Thank you,
We'll try.

To make it clear what we need:
one sql with :StartDate (day 1) to :EndDate (day n) and :StartHour

day 1 - 8am (grouped data from >= 8am and < 8am day 2)
day 2 - 8am ...
day 3 - 8am ...
...
day n - 8am ...

this is not working for me. Could you please make yourself a full SQL with results.
Here's mine:
=== Working example with hour=0
select
DATEPART(YEAR ,V.CAS) as YY,
DATEPART(MONTH ,V.CAS) as MM,
DATEPART(DAY ,V.CAS) as DD,
SUM(V.ST0_poraba) as V_SUM
FROM dbo.MERITVE as V
Where DATEPART(YEAR,V.CAS)=2014 and DATEPART(MONTH,V.CAS)=5
GROUP BY DATEPART(YEAR,V.CAS),DATEPART(MONTH,V.CAS),DATEPART(DAY,V.CAS)
order by YY,MM,DD asc

=== Error example with hour=8, I guess
select
DATEPART(YEAR ,V.CAS) as YY,
DATEPART(MONTH ,V.CAS) as MM,
DATEPART(DAY ,V.CAS) as DD,
SUM(V.ST0_poraba) as V_SUM
FROM dbo.MERITVE as V
Where DATEPART(YEAR,V.CAS)=2014 and DATEPART(MONTH,V.CAS)=5
GROUP BY DATEPART(YEAR,DATEADD(hour,-8,V.CAS)),DATEPART(MONTH,DATEADD(hour,-8,V.CAS)),DATEPART(DAY,DATEADD(hour,-8,V.CAS))
order by YY,MM,DD asc

error:
Msg 8120, Level 16, State 1, Line 4
Column 'dbo.MERITVE.Cas' is invalid in the select list because it is not contained in either an aggregate function or the GROUP BY clause.

Sorry. I thought it would be OK as long as the GROUP BY included the datetime field, but it doesn't work because there are two different datetime values, the original value in the SELECT and the adjusted value in the GROUP BY.

Since you're trying to use the adjusted date in the GROUP BY, and the original date in the SELECT, you will (presumably) end up with situations like some rows that are dated after 8 AM on 05/10, so the adjusted date is still 05/10, and some rows that are dated before 8 AM on 05/11, so the adjusted date for those is also 05/10, so they would be grouped together. But they have 2 different original dates (05/10 and 05/11).

If they're grouped together, you can only have one set of values in YY, MM and DD for those rows, so which date would you want to see -- 05/10 or 05/11 ? Maybe you could just use MIN or MAX on the DATEPART results in the SELECT.

Also, just a side question:

In your example code, you're looking for dates in May, 2014. What if there was a row dated at 7 AM on 05/01. When you subtracted 8 hours, that would be 11 PM on 04/30. Should that row be excluded, since the adjusted date is not in May?

James

0

vojkocendakAuthor Commented:

side question:

If for May : >= 1.5 08:00:00 ... < 1.June 08:00:00 sums per day

Yes, of course it's bizarre, but some companies prefer to use start the day from 08am to 08am next day, or whatever.

select
CONVERT(CHAR(10),DATEADD(hh,-8,[CAS]),120) AS MyDate
,COUNT(1) AS cnt
,SUM(V.ST39_poraba) AS poraba
FROM dbo.MERITVE as V
Where V.cas>='2014-05-01 08:00:00' and V.Cas<'2014-06-01 08:00:00' // data for may
GROUP BY CONVERT(CHAR(10),DATEADD(hh,-8,[Cas]),120)
order by 1

Yeah, that will probably run, because it uses the same value (-8 hours) in the SELECT and GROUP BY. If that gives you what you need, great. I was assuming, perhaps incorrectly, that you wanted to use the original values (eg. DATEPART(YEAR ,V.CAS) as YY) in the SELECT, but group by the modified values. Trying to do both at the same time is what was causing the problem.

James

0

vojkocendakAuthor Commented:

thank you James,

DATEPART was our starting point, because we're new to this.

But you gave us starting idea with -8 to begin with. How do we proceed with points (we're also new to that one ...)

If you're grouping by the day and not the hour, did you try David's suggestion for the GROUP BY? That would probably be more efficient than converting the date to a string, although if you're also converting the date to a string in the SELECT, then it may not make much difference if you do it in the GROUP BY too. But I'd give it a try.

As for the points, you could accept your own post with the new code as the solution, and any posts from myself or David that you think helped could be "assists", and split the points however you see fit.

James

0

vojkocendakAuthor Commented:

It started to think with the comments of James and David. The solution I found was from another web source, which helped me to experiment and find right one.

0

Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

GROUP BY DATEPART(YEAR , DATEADD(hour, 0 - @start_hour, V.DateTimeField)),

DATEPART(MONTH , DATEADD(hour, 0 - @start_hour, V.DateTimeField)),

DATEPART(DAY , DATEADD(hour, 0 - @start_hour, V.DateTimeField))

Then, for example, if @start_hour was 8, 7 AM on 04/24 would become 11 PM on 04/23 (so it would be grouped with 04/23), while 8 AM on 04/24 would become midnight on 04/24.

I haven't tested that, but it seems like it should work.

If your start "hour" could really be any time, like 7:20 AM or 8:45 AM, then you could convert the parameter to minutes or seconds and use that with DATEADD.

James