How to group value within a certain date range or every certain days in SQL server - sql

Question

For example table would be:
Customer OrderDate OrderAmt
-------- ---------- ---------
A1 20140101 920.00
A2 20140111 301.00
A2 20140123 530.00
A1 20140109 800.00
A3 20140110 500.00
A1 20140115 783.00
A3 20140217 500.00
A1 20140219 1650.00
A1 20140225 780.00
...
A3 20140901 637.00
I want to group them and calculate the sum(OrderAmt) within every 20 days and group by customer start from 20140101.

Answer 1

For what it's worth, you can accomplish what you describe with a pretty simple DATEDIFF() / GROUP BY operation, as below: whether or not that is actually what you want might be another question. I suspect that the DateBucket calculation might actually be something else ...
CREATE TABLE #tmpCustomer (Customer VARCHAR(2), OrderDate VARCHAR(10), OrderAmt DECIMAL(6,2))
INSERT INTO #tmpCustomer (Customer, OrderDate, OrderAmt)
SELECT 'A1',20140101, 920.00 UNION
SELECT 'A2',20140111, 301.00 UNION
SELECT 'A2',20140123, 530.00 UNION
SELECT 'A1',20140109, 800.00 UNION
SELECT 'A3',20140110, 500.00 UNION
SELECT 'A1',20140115, 783.00 UNION
SELECT 'A3',20140217, 500.00 UNION
SELECT 'A1',20140219, 1650.00 UNION
SELECT 'A1',20140225, 780.00 UNION
SELECT 'A3',20140901, 637.00
SELECT
Customer,
(DATEDIFF(day, '1/1/2014', CAST(OrderDate AS DATE)) / 20) + 1 AS DateBucket,
SUM(OrderAmt) SumOrderAmt
FROM #tmpCustomer
GROUP BY Customer, (DATEDIFF(day, '1/1/2014', CAST(OrderDate AS DATE)) / 20) + 1
ORDER BY Customer, DateBucket

Answer 2

You need to do two things:
(1) Create some sort of guide hold the '20 day groups' information. A Recursive CTE does this pretty well, and
(2) Recast that varchar date as an actual date for comparison purposes.
Then it's just joining the order data into that daterange grouping and summing the order amounts.
-------------------------
-- Here i'm just recreating your example
-------------------------
DECLARE #customerOrder TABLE (Customer varchar(2), OrderDate varchar(8), OrderAmt decimal(8,2))
INSERT INTO #customerOrder (Customer, OrderDate, OrderAmt)
VALUES
('A1', '20140101', 920.00),
('A2', '20140111', 301.00),
('A2', '20140123', 530.00),
('A1', '20140109', 800.00),
('A3', '20140110', 500.00),
('A1', '20140115', 783.00),
('A3', '20140217', 500.00),
('A1', '20140219', 1650.00),
('A1', '20140225', 780.00),
('A3', '20140901', 637.00)
-------------------------
-- Set up a table that lists off 20 day periods starting from 1/1/2014
-------------------------
DECLARE #startDate datetime, #endDate datetime;
SET #startDate = {d N'2014-01-01'};
SET #endDate = {d N'2014-12-31'};
WITH [dates] ([Sequence], [startDate], [maxExcludedDate]) AS
(SELECT 1 AS [Sequence]
,#startDate AS [startDate]
,DATEADD(d, 20, #startDate) AS [maxExcludedDate]
UNION ALL
SELECT Sequence + 1 AS Sequence
,DATEADD(d, 20, [startDate]) AS [startDate]
,DATEADD(d, 40, [startDate]) AS [maxExcludedDate]
FROM [dates]
WHERE [startDate] < #endDate
)
, dateFrame AS
(
SELECT
[startDate]
,[maxExcludedDate]
FROM [dates]
)
-------------------------
-- Set up a table that holds the orderDates as actual dates
-------------------------
, castData AS
(
SELECT
cast(orderdate as datetime) castDate
,OrderAmt
FROM #customerOrder
)
-------------------------
-- JOIN and sum.
-------------------------
SELECT
[startDate]
, Sum(OrderAmt) perodAmt
FROM
dateFrame df
left join castData cd
on cd.castDate >= df.startDate
and cd.castDate < df.maxExcludedDate
GROUP BY
[startDate]
ORDER by
[startDate]

Answer 3

Assuming that the OrderDate is a numeric field (not varchar). I'm also assuming that you don't need to go much more than a year in the future. If you want the gaps shown, keep the left join, if you don't want the gaps, then make it an inner join. (You can also make the hardcoded date a variable of where to start, I just kept it as the 20140101 that you mentioned.
with Numbers as
(Select 0 as Num
UNION ALL
Select Num+1
FROM Numbers
WHERE Num+1<= 20
)
, DateList AS
(Select Year(DateAdd(dd,20*(Num), Cast('2014-01-01' as date))) * 10000+Month(DateAdd(dd,20*(Num), Cast('2014-01-01' as date)))*100+Day(DateAdd(dd,20*(Num), Cast('2014-01-01' as date))) Groupingdatemin
, Year(DateAdd(dd,20*(Num+1)-1, Cast('2014-01-01' as date)))*10000+ MONTH(DateAdd(dd,20*(Num+1)-1, Cast('2014-01-01' as date)))*100+DAY(DateAdd(dd,20*(Num+1)-1, Cast('2014-01-01' as date))) Groupingdatemax
from Numbers
)
select Customer, sum(orderamt), Groupingdatemin, Groupingdatemax from DateLIst d LEFT JOIN
<yourtable> t on t.orderdate between d.Groupingdatemin and d.Groupingdatemax
group by customer, Groupingdatemin, Groupingdatemax