Except does not make correct intersection between 2 sets - sql

Question

I want to get best sellers from march 2019, while excluding the top 3 sellers of january. I tried using except where first SELECT gives best sellers of march (all of them) and the second SELECT gives top 3 of january.
SELECT * FROM (SELECT fullname, SUM(sale) sales
FROM mytable
WHERE oredrdate BETWEEN '2019-03-01' AND '2019-03-31'
GROUP BY fullname
ORDER BY sales DESC) X
EXCEPT
SELECT * FROM (SELECT fullname, SUM(sale) sales
FROM mytable
WHERE oredrdate BETWEEN '2019-01-01' AND '2019-01-31'
GROUP BY fullname
ORDER BY sales DESC
LIMIT 3) Y;
The problem is that EXCEPT does not intersect as I wished it would. What each SELECT returns and my desired output with data:
First SELECT returns:
fullname sales
Tommy Williams 8320
Ryan Atkinson 7310
Petey Cruiser 6200
Anna Mull 5840
Gail Forcewind 4120
Paige Turner 3300
Bob Frapples 2100
... ...
Seconds SELECT returns:
fullname sales
Tommy Williams 9220
Anna Mull 8100
Greta Life 7891
Desired OUTPUT:
fullname sales
Ryan Atkinson 7310
Petey Cruiser 6200
Gail Forcewind 4120
Paige Turner 3300
Bob Frapples 2100
... ...
How should I change my code to achieve this?

Answer 1

This can be done with a LEFT JOIN where you exclude the matching rows:
SELECT X.*
FROM (
SELECT fullname, SUM(sale) sales
FROM mytable
WHERE oredrdate BETWEEN '2019-03-01' AND '2019-03-31'
GROUP BY fullname
) X LEFT JOIN (
SELECT fullname, SUM(sale) sales
FROM mytable
WHERE oredrdate BETWEEN '2019-01-01' AND '2019-01-31'
GROUP BY fullname
ORDER BY sales DESC
LIMIT 3
) Y ON Y.fullname = X.fullname
WHERE Y.fullname IS NULL
ORDER BY X.sales DESC

Answer 2

You could use:
SELECT fullname, SUM(sales) AS total
FROM mytable
WHERE oredrdate BETWEEN '2019-03-01' AND '2019-03-31'
AND fullname NOT IN (SELECT fullname, SUM(sales) AS total
FROM mytable
WHERE oredrdate BETWEEN '2019-01-01' AND '2019-01-31'
AND fullname IS NOT NULL
GROUP BY fullname
ORDER BY total DESC LIMIT 3)
GROUP BY fullname
ORDER BY total DESC;
I would group by some kind of unique column like employee_id, there is possibility that two persons could have the same name.

Answer 3

The problem is that EXCEPT is considering both the name and the amount columns. It is unlikely that the second would match.
One way to write this is:
WITH jan3 as (
SELECT TOP (3) fullname, SUM(sale) as sales
FROM mytable
WHERE orderdate >= '2019-01-01' AND
orderdate < '2019-02-01'
GROUP BY fullname
ORDER BY sales DESC
)
SELECT m.fullname, SUM(m.sale) as sales
FROM mytable m
WHERE m.orderdate >= '2019-03-01' AND
m.orderdate < '2019-04-01' AND
NOT EXISTS (SELECT 1
FROM jan3
WHERE jan3.fullname = m.fullname
)
GROUP BY fullname
ORDER BY sales DESC;
Note that this changes the date comparisons to use >= and <. This is considered a best practice, because it works for dates and datetime (timestamp) values.
There are other ways of writing this using only a single aggregation. For instance:
WITH s as (
SELECT m.fullname,
SUM(CASE WHEN m.orderdate < '2019-02-01' THEN m.sale END) as sales_jan,
SUM(CASE WHEN m.orderdate >= '2019-03-01' THEN m.sale END) as sales_mar
FROM mytable m
WHERE m.orderdate >= '2019-01-01' AND
m.orderdate < '2019-04-01'
)
SELECT s.*
FROM (SELECT s.*,
ROW_NUMBER() OVER (ORDER BY sales_jan DESC) as seqnum_jan
FROM s
) s
WHERE seqnum_jan > 3
ORDER BY s.sales_mar;