How to properly GROUP BY datetime columns in specific intervals? - sql-server

Question

I am making a stored procedure in MSSQL and I want to group my results by the interval day, I am unsure if this is correct though:
(this is a portion)
INSERT #Results
(
[Day],
[Month],
[Year],
[Result]
)
SELECT
DATEPART(DD, DATEADD(MI, #GmtOffset, EventDate)),
DATEPART(MM, DATEADD(MI, #GmtOffset, EventDate)),
DATEPART(YY, DATEADD(MI, #GmtOffset, EventDate)),
Result = CASE WHEN #Metric = 'Take Rate' THEN NULL ELSE COUNT(*) END
FROM BundleEvent
WHERE BundleEventTypeId = CASE WHEN #Metric = 'Take Rate' THEN #TypeTakeId ELSE #BundleEventTypeId END
AND EventDate >= #StartTime AND EventDate <= #EndTime
GROUP BY
DATEPART(YY, DATEADD(MI, #GmtOffset, EventDate)),
DATEPART(MM, DATEADD(MI, #GmtOffset, EventDate)),
DATEPART(DD, DATEADD(MI, #GmtOffset, EventDate))
My fear is it is ultimately going to group the data by the date where the year or month do not matter. Does anyone know if I am doing this correctly? EventDate is the DateTime field that I want to do a GROUP BY on by interval of day.
Thanks
EDIT:
This is the correct way, a simple format exchange... sometimes I need more coffee.
INSERT #Results
(
[Date],
[Result]
)
SELECT
CAST(DATEADD(MI, #GmtOffset, BundleEvent.EventDate) AS DATE),
Result = CASE WHEN #Metric = 'Take Rate' THEN 0 ELSE COUNT(*) END
FROM dbo.BundleEvent WITH (NOLOCK)
JOIN dbo.BundleUser WITH (NOLOCK)
ON BundleEvent.BundleId = BundleUser.BundleId
JOIN dbo.Bundle WITH (NOLOCK)
ON BundleEvent.BundleId = Bundle.BundleId
WHERE BundleEvent.EventDate >= #StartTimeGmt AND BundleEvent.EventDate <= #EndTimeGmt
AND BundleEvent.BundleEventTypeId = CASE WHEN #Metric = 'Take Rate' THEN #TypeTakeId ELSE #BundleEventTypeId END
AND BundleUser.UserId = CASE WHEN #UserId IS NULL THEN BundleUser.UserId ELSE #UserId END
AND Bundle.BundleType = 1
GROUP BY
CAST(DATEADD(MI, #GmtOffset, BundleEvent.EventDate) AS DATE)
Then I do a sub compare with the format exchange:
CAST(
(SELECT COUNT(*)
FROM dbo.BundleEvent WITH (NOLOCK)
JOIN dbo.BundleUser WITH (NOLOCK)
ON BundleEvent.BundleId = BundleUser.BundleId
JOIN dbo.Bundle WITH (NOLOCK)
ON BundleEvent.BundleId = Bundle.BundleId
WHERE CAST(DATEADD(MI, #GmtOffset, BundleEvent.EventDate) AS DATE) = [Date]
AND BundleEvent.BundleEventTypeId = #TypeTakeId
AND BundleUser.UserId = CASE WHEN #UserId IS NULL THEN BundleUser.UserId ELSE #UserId END
AND Bundle.BundleType = 1)
AS DECIMAL(5,2)
So in essence the query is matching up correctly and using a new format.

Answer 1

Since you are not using and aggregates (except Count(*), Are you sure you want a GROUP BY? Using DISTINCT will produce a single insert entry for all matching dates; perhaps something like:
DECLARE #DATE_GMT datetime
SET #DATE_GMT=DATEADD(MI, #GmtOffset, EventDate)
INSERT #Results
(
[Day],
[Month],
[Year],
[Result]
)
SELECT DISTINCT
DATEPART(DD, #DATE_GMT),
DATEPART(MM, #DATE_GMT),
DATEPART(YY, #DATE_GMT),
Result = CASE WHEN #Metric = 'Take Rate' THEN NULL ELSE COUNT(*) END
FROM BundleEvent
WHERE BundleEventTypeId = CASE WHEN #Metric = 'Take Rate' THEN #TypeTakeId ELSE #BundleEventTypeId END
AND EventDate Between #StartTime AND #EndTime