SQL Query to Roll Up Data from the Last 7 Days - sql

Question

Let's assume I have a table structure like this.
CheckIn
- int checkInId pk
- int companyPositionId
- Date checkInDate
Let's say I want to get a count of all check ins for the last 7 days from a given date. What would be the best way to do this? Right now I'm making a query for each of the 7 dates AND the company positions. Currently this is too slow because there could be many companyPositions * 7 days. How can I roll this up into one query?
Would it be easiest to generate the last 7 days dates and construct a long query? Could I then group count by a date range for each of the 7 days?
An ideal result back could look like:
companyPositionId, date1Count, date2Count, date3Count, date4Count, date5Count, date6Count.
Example Data:
checkInId | companyPositionId | checkInDate
1 | 1 | 1970-01-01
2 | 1 | 1970-01-02
3 | 1 | 1970-01-03
4 | 1 | 1970-01-04
5 | 1 | 1970-01-05
6 | 1 | 1970-01-06
7 | 1 | 1970-01-07
8 | 2 | 1970-01-01
9 | 2 | 1970-01-02
10 | 2 | 1970-01-03
11 | 2 | 1970-01-04
12 | 2 | 1970-01-05
13 | 2 | 1970-01-06
14 | 2 | 1970-01-07*
15 | 2 | 1970-01-07*
My current query is this:
SELECT * FROM CheckIn
WHERE (startDate) <= (inputDate)
AND (inputDate) <= (endDate)
AND companyPositionId = (companyPositionId);
I then loop through each startDate/endDate that is generated from the beginning of the day and end of that day. And then each of the companyPositionId's.
Ideal result:
companyPositionId | date1Count | date2Count | date3Count | date4Count | date5Count | date6Count | date7Count
1 | 1 | 1 | 1 | 1 | 1 | 1 | 1
2 | 1 | 1 | 1 | 1 | 1 | 1 | 2

Answer 1

You can do this with the PIVOT command, or with conditional SUMs:
DECLARE #my_date DATE = CAST(GETDATE() AS DATE)
SELECT
companyPositionId,
SUM(CASE WHEN CAST(checkInDate AS DATE) = DATEADD(DAY, -7, #my_date) THEN 1 ELSE 0 END) AS date1Count,
SUM(CASE WHEN CAST(checkInDate AS DATE) = DATEADD(DAY, -6, #my_date) THEN 1 ELSE 0 END) AS date2Count,
SUM(CASE WHEN CAST(checkInDate AS DATE) = DATEADD(DAY, -5, #my_date) THEN 1 ELSE 0 END) AS date3Count,
SUM(CASE WHEN CAST(checkInDate AS DATE) = DATEADD(DAY, -4, #my_date) THEN 1 ELSE 0 END) AS date4Count,
SUM(CASE WHEN CAST(checkInDate AS DATE) = DATEADD(DAY, -3, #my_date) THEN 1 ELSE 0 END) AS date5Count,
SUM(CASE WHEN CAST(checkInDate AS DATE) = DATEADD(DAY, -2, #my_date) THEN 1 ELSE 0 END) AS date6Count,
SUM(CASE WHEN CAST(checkInDate AS DATE) = DATEADD(DAY, -1, #my_date) THEN 1 ELSE 0 END) AS date7Count
FROM
CheckIn
WHERE
checkInDate BETWEEN DATEADD(DAY, -7, GETDATE()) AND DATEADD(DAY, -1, GETDATE())
GROUP BY
companyPositionId
If your checkInDate has a time component then you'll need to account for that.
I declared the #my_date variable just to avoid having to repeat that expression in the query a bunch of times, but you could replace the variable with that expression and it would work as well. You could also use BETWEEN which might have better performance since the optimizer could then potentially use an index on your checkInDate. Just calculate midnight/11:59:59 of each of the days instead of looking for equality.

Answer 2

Look at
Declare #d date= '2016-05-05'; -- parameter, we need this day and six previous
Select
companyPositionId
,date1Count = count(case checkInDate when #d then checkInId end)
,date2Count = count(case checkInDate when dateadd(d,-1,#d) then checkInId end)
--...
from checkIn
where checkInDate between dateadd(d,-6,#d) and #d
group by companyPositionId
order by companyPositionId;

Answer 3

Here's a quick pivot example assuming you're looking for the previous 7 days of the #inputDate. This should work even with the time component since it goes off a DATEDIFF with days only.
DECLARE #InputDate DATE = '1/8/1970'
;WITH CTE AS (
SELECT *, DATEDIFF(DD, checkInDate, #InputDate) AS DaysAgo FROM #CheckIns
)
SELECT
companyPositionId,
SUM([7]) AS date1Count,
SUM([6]) AS date2Count,
SUM([5]) AS date3Count,
SUM([4]) AS date4Count,
SUM([3]) AS date5Count,
SUM([2]) AS date6Count,
SUM([1]) AS date7Count -- Reversed columns since you wanted 7 to be most recent day
FROM CTE
PIVOT (COUNT(checkInId) FOR DaysAgo IN ([1], [2], [3], [4], [5], [6], [7])) PVT
GROUP BY
companyPositionId
This gives the following results matching your desired output:
companyPositionId date1Count date2Count date3Count date4Count date5Count date6Count date7Count
----------------- ----------- ----------- ----------- ----------- ----------- ----------- -----------
1 1 1 1 1 1 1 1
2 1 1 1 1 1 1 2

Answer 4

Create Data:
CREATE TABLE #checkin (
checkInId INT IDENTITY(1, 1),
companyPositionId int,
checkInDate DATE
)
DECLARE #counter INT = 100
WHILE #counter > 0
BEGIN
INSERT INTO #checkin
( companyPositionId, checkInDate )
VALUES ( RAND() * 10 + 1, -- 10 possible companies?
DATEADD(day, RAND()*-7, GETDATE()) -- random days in the last 2 weeks
)
SET #counter = #counter - 1
END
Logic starts here:
DECLARE
#now DATETIME = GETDATE(),
#days INT = -7 -- Our interval of interest
This logic is partially borrowed from Bogdan Sahlean, his results were getting all '1' for me, so I changed the dayNum calculation
SELECT *
FROM (
SELECT companyPositionId, checkInId, 1 + CAST(DATEDIFF(DAY,checkInDate, #now) AS INT) AS DayNum
FROM #CheckIn
WHERE checkInDate BETWEEN CAST(DATEADD(DAY, #days, #now) AS DATE) AND #now
) AS ps -- Pivot source
PIVOT ( COUNT(checkInId) FOR DayNum IN ([1], [2], [3], [4], [5], [6], [7]) ) AS p
DROP TABLE #checkin