SQL Server select with multiple groupings - sql

Question

I have two tables describing users and their payments:
CREATE TABLE test_users
(id int IDENTITY NOT NULL,
name varchar(25),
PRIMARY KEY (id));
CREATE TABLE test_payments
(id int IDENTITY NOT NULL,
user_id int NOT NULL,
money money NOT NULL,
date datetime NOT NULL,
PRIMARY KEY (id));
INSERT INTO test_users (name)
VALUES ('john');
INSERT INTO test_users (name)
VALUES ('peter');
INSERT INTO test_payments (user_id, money, date)
VALUES (1, $1, CONVERT(datetime, '15.12.2012'));
INSERT INTO test_payments (user_id, money, date)
VALUES (1, $2, CONVERT(datetime, '16.12.2012'));
INSERT INTO test_payments (user_id, money, date)
VALUES (2, $1, CONVERT(datetime, '16.12.2012'));
INSERT INTO test_payments (user_id, money, date)
VALUES (2, $3, CONVERT(datetime, '17.12.2012'));
INSERT INTO test_payments (user_id, money, date)
VALUES (1, $1, CONVERT(datetime, '19.12.2012'));
Table test_users:
id name
-------------
1 john
2 peter
Table test_payments:
id user_id money last_activity
---------------------------------------
1 1 1.0000 2012-12-15
2 1 2.0000 2012-12-16
3 2 1.0000 2012-12-16
4 2 3.0000 2012-12-17
5 1 1.0000 2012-12-19
I need to make a users statistic which will show me :
username
total fee for a period of time
the date of the last
user's activity (general, not for a time period).
For example taking the period 15-18.12.12 I expect the following results:
name total last_activity
--------------------------------
peter $4 2012-12-17
john $3 2012-12-19
I've tried the following query:
SELECT u.*, SUM(p.money) total, MAX(p.date) last_activity
FROM test_users u
JOIN test_payments p
ON u.id= p.user_id
WHERE p.date BETWEEN CONVERT(datetime, '15.12.2012') AND CONVERT(datetime, '18.12.2012')
GROUP BY u.id, u.name
ORDER BY total DESC;
but getting wrong result for last_activity as it is also in the date range:
id name total last_activity
--------------------------------
2 peter 4.0000 2012-12-17
1 john 3.0000 2012-12-16
Please suggest a solution.

Answer 1

Looks like a couple of other answers popped up while I worked on mine, but here it is anyhow. There is a working sql fiddle here: http://sqlfiddle.com/#!3/14808/6
Basically, you need a query to pull the max date regardless of the date range. I chose to do this as a correlated subquery.
SELECT
u.id,
u.name,
SUM(IsNull(money,0)) as TotalMoneyInRange,
(SELECT max(date) FROM test_payments where user_id = u.id) AS LastPaymentOverAll
FROM test_users AS u
LEFT JOIN test_payments AS p
ON u.id = p.user_id
WHERE
p.date IS NULL OR
p.date between
CAST('12-11-2012' AS datetime) --range begin
and
CAST('12-16-2012' as datetime) --range end
GROUP BY u.id, u.name

Answer 2

You need to move the condition from the where clause to a case statement:
SELECT u.id, u.name,
SUM(case when p.date BETWEEN CONVERT(datetime, '15.12.2012') AND CONVERT(datetime, '18.12.2012')
then p.money
end) total,
MAX(p.date) last_activity
FROM test_users u JOIN
test_payments p
ON u.id= p.user_id
GROUP BY u.id, u.name
ORDER BY total DESC;
If you only want users who had a payment in that period, then you can include:
having total is not null
If you want the NULL values to appear as 0 instead of NULL, then include else 0 in the case statement.

Answer 3

You can also use subqueries to get the result:
SELECT u.*, total, last_activity
FROM test_users u
JOIN
(
select sum(money) total, user_id
from test_payments
WHERE date BETWEEN CONVERT(datetime, '2012-12-15')
AND CONVERT(datetime, '2012-12-18')
group by user_id
) p
ON u.id= p.user_id
inner join
(
select user_id, max(date) last_activity
from test_payments
group by user_id
) p1
on p.user_id = p1.user_id
ORDER BY total DESC;
See SQL Fiddle with Demo

Answer 4

You could add a sub query for the MAX date that doesn't have the WHERE clause like so:
SELECT
u.*
,SUM(p.money) total
,a.max_date last_activity
FROM test_users u
INNER JOIN test_payments p ON u.id = p.user_id
INNER JOIN (SELECT user_id, MAX(date) AS max_date
FROM test_payments
GROUP BY user_id) a ON u.id = a.user_id
WHERE p.date BETWEEN CONVERT(datetime, '15.12.2012') AND CONVERT(datetime, '18.12.2012')
GROUP BY u.id, u.name, a.max_date
ORDER BY total DESC;