SELECT DISTINCT from one column based on MAX of another - mysql

Question

I have a query that returns the relative activity of users in each region. I want to be returned that list but with each user only in 1 region, so I want to filter out on everyone's MAX applications.
The current query:
SELECT
r.region_id,
ha.user_id,
count(ha.user_id) AS applications
FROM
sit_applications ha
LEFT JOIN
listings_regions r
ON
r.listingID = ha.listingID
AND deleted = 0
WHERE
ha.datetime_applied >= (NOW() - INTERVAL 1 MONTH)
GROUP BY
ha.user_id, r.region_id
HAVING
applications > 0
ORDER BY
r.region_id DESC
I need to filter this query so I only grab each user_id once, and with it's biggest applications for a region. This is so I have a list of all the top performers for each region, with no duplicate users.

Answer 1

In MySQL, you have three basic ways to do this:
Using variables
Using a complex join
Using a hack with substring_index() and group_concat().
The complex join is really a mess when you have aggregation queries. The hack is fun, but does have its limitation. So, let's consider the variables method:
SELECT ur.*
FROM (SELECT ur.*,
(#rn := if(#u = user_id, #rn + 1,
if(#u := user_id, 1, 1)
)
) as rn
FROM (SELECT r.region_id, ha.user_id, count(ha.user_id) AS applications
FROM sit_applications ha LEFT JOIN
listings_regions r
ON r.listingID = ha.listingID AND deleted = 0
WHERE ha.datetime_applied >= (NOW() - INTERVAL 1 MONTH)
GROUP BY ha.user_id, r.region_id
HAVING applications > 0
) ur CROSS JOIN
(SELECT #u := -1, #rn := 0) params
ORDER BY user_id, applications DESC
) ur
WHERE rn = 1;
Note: Aspects of your query do not really make sense, even though I left them in. You are using LEFT JOIN, so r.region_id could be NULL -- and that is usually not desirable. You have a HAVING clause that is totally unnecessary, because the COUNT() is always 1 -- assuming that ha.user_id is never NULL. I suspect that the logic could be replaced with an INNER JOIN, no HAVING clause, and COUNT(*).

Answer 2

You could try wrapping the query and extracting out what you want:
SELECT t2.user_id, t2.region_id, t2.applications
FROM
(
SELECT t.user_id, MAX(t.applications) AS applications
FROM
(
SELECT r.region_id, ha.user_id, COUNT(ha.user_id) AS applications
FROM sit_applications ha LEFT JOIN listings_regions r
ON r.listingID = ha.listingID AND deleted = 0
WHERE ha.datetime_applied >= (NOW() - INTERVAL 1 MONTH)
GROUP BY ha.user_id, r.region_id
HAVING applications > 0
) t
GROUP BY t.user_id
) t1
INNER JOIN
(
SELECT r.region_id, ha.user_id, COUNT(ha.user_id) AS applications
FROM sit_applications ha LEFT JOIN listings_regions r
ON r.listingID = ha.listingID AND deleted = 0
WHERE ha.datetime_applied >= (NOW() - INTERVAL 1 MONTH)
GROUP BY ha.user_id, r.region_id
HAVING applications > 0
) t2
ON t1.user_id = t2.user_id AND t1.applications = t2.applications