Finding Duplicate Orders (by time proximity) - sql

Question

I have a table of orders that I know have duplicates
customer order_number order_date
---------- ------------ -------------------
1 1 2012-03-01 01:58:00
1 2 2012-03-01 02:01:00
1 3 2012-03-01 02:03:00
2 4 2012-03-01 02:15:00
3 5 2012-03-01 02:18:00
3 6 2012-03-01 04:30:00
4 7 2012-03-01 04:35:00
5 8 2012-03-01 04:38:00
6 9 2012-03-01 04:58:00
6 10 2012-03-01 04:59:00
I want to find all duplicates (order by same customer within 60 minutes of eachother). Either a resultset consisting of the 'duplicate' rows or a set of all customers with a count of how many duplicates.
Here is what I have tried
SELECT
customer,
count(*)
FROM
orders
GROUP BY
customer,
DATEPART(HOUR, order_date)
HAVING (count(*) > 1)
This doesn't work when duplicates are within 60 minutes of each other but are in different hours i.e 1:58 and 2:02
I've also tried this
SELECT
o1.customer,
o1.order_number,
o2.order_number,
DATEDIFF(MINUTE,o1.order_date, o2.order_date) AS [diff]
FROM
orders o1 LEFT OUTER JOIN
orders o2 ON o1.customer = o2.customer AND o1.order_number <> o2.order_number
WHERE
ABS(DATEDIFF(MINUTE,o1.order_date, o2.order_date)) < 60
Now this gives me all of the duplicates but it also gives me multiple rows per duplicate order. i.e (o1, o2) and (o2, o1) which wouldn't be so bad if there were'nt some orders with multiple duplicates. In those cases I get (o1, o2), (o1,o3), (o2, o1), (o2, o3), (o3, o1), (o3, o2) etc. I get all of the permutations.
Anyone have some insight? I'm not necessarily looking for the best performing answer here, just one that works.

Answer 1

SELECT
*,
CASE WHEN EXISTS (SELECT *
FROM orders AS lookup
WHERE customer = orders.customer
AND order_date < orders.order_date
AND order_date >= DATEADD(hour, -1, order_date)
)
THEN 'Principle Order'
ELSE 'Duplicate Order'
END as Order_Status
FROM
orders
Using EXISTS and a correlated sub-query you can check if there were any preceding orders in the last hour.

Answer 2

Maybe something like this:
Test data:
DECLARE #tbl TABLE(customer INT,order_number INT,order_date DATETIME)
INSERT INTO #tbl
VALUES
(1,1,'2012-03-01 01:58:00'),
(1,2,'2012-03-01 02:01:00'),
(1,3,'2012-03-01 02:03:00'),
(2,4,'2012-03-01 02:15:00'),
(3,5,'2012-03-01 02:18:00'),
(3,6,'2012-03-01 04:30:00'),
(4,7,'2012-03-01 04:35:00'),
(5,8,'2012-03-01 04:38:00'),
(6,9,'2012-03-01 04:58:00'),
(6,10,'2012-03-01 04:59:00')
Query
;WITH CTE
AS
(
SELECT
MIN(datediff(minute,'1990-1-1',order_date)) OVER(PARTITION BY customer) AS minDate,
datediff(minute,'1990-1-1',order_date) AS DateTicks,
tbl.customer
FROM
#tbl AS tbl
)
SELECT
CTE.customer,
SUM(CASE WHEN (CTE.DateTicks-CTE.minDate)<60 THEN 1 ELSE 0 END)
FROM
CTE
GROUP BY
CTE.customer

Answer 3

The following query identifies all possible permutations of orders within the proximity of 60 minutes of one another:
DECLARE #orders TABLE (CustomerId INT, OrderId INT, OrderDate DATETIME)
INSERT INTO #orders
VALUES
(1, 1, '2012-03-01 01:58:00'),
(1, 2, '2012-03-01 02:01:00'),
(1, 3, '2012-03-01 02:03:00'),
(2, 4, '2012-03-01 02:15:00'),
(3, 5, '2012-03-01 02:18:00'),
(3, 6, '2012-03-01 04:30:00'),
(4, 7, '2012-03-01 04:35:00'),
(5, 8, '2012-03-01 04:38:00'),
(6, 9, '2012-03-01 04:58:00'),
(6, 10, '2012-03-01 04:59:00');
with ProximityOrderCascade(CustomerId, OrderId, ProximateOrderId, MinutesDifference, OrderDate, ProximateOrderDate)
as
(
select o.customerid, o.orderid, null, null, o.orderdate, o.orderdate
from #orders o
union all
select o.customerid, o.orderid, p.orderid, datediff(minute, p.OrderDate, o.OrderDate), o.OrderDate, p.OrderDate
from ProximityOrderCascade p
inner join #orders o
on p.customerid = o.customerid
and abs(datediff(minute, p.OrderDate, o.OrderDate)) between 0 and 60
and o.orderid <> p.orderid
where proximateorderid is null
)
select * from ProximityOrderCascade
where
not ProximateOrderId is null
From there, you can transform the results into a query of your choice. The results of this function identify only customers 1 and 6 as having 'duplicate' orders.
CustomerId OrderId ProximateOrderId MinutesDifference OrderDate ProximateOrderDate
----------- ----------- ---------------- ----------------- ----------------------- -----------------------
6 9 10 -1 2012-03-01 04:58:00.000 2012-03-01 04:59:00.000
6 10 9 1 2012-03-01 04:59:00.000 2012-03-01 04:58:00.000
1 1 3 -5 2012-03-01 01:58:00.000 2012-03-01 02:03:00.000
1 2 3 -2 2012-03-01 02:01:00.000 2012-03-01 02:03:00.000
1 1 2 -3 2012-03-01 01:58:00.000 2012-03-01 02:01:00.000
1 3 2 2 2012-03-01 02:03:00.000 2012-03-01 02:01:00.000
1 2 1 3 2012-03-01 02:01:00.000 2012-03-01 01:58:00.000
1 3 1 5 2012-03-01 02:03:00.000 2012-03-01 01:58:00.000
(8 row(s) affected)