Mysql query the most visited url per user - mysql

Question

I need to count the urls with the most visits per user.
Table name:visit_actions: mysql version:5.7
+----+--------+---------+---------------------+
| id | url_id | user_id | server_time |
+----+--------+---------+---------------------+
| 1 | 265338 | 4 | 2019-11-07 08:54:47 |
| 2 | 265405 | 1 | 2019-11-07 08:55:21 |
| 3 | 265391 | 4 | 2019-11-07 08:56:03 |
| 4 | 265338 | 1 | 2019-11-07 08:57:36 |
| 5 | 265338 | 1 | 2019-11-07 10:02:46 |
| 21 | 265207 | 5 | 2019-11-08 02:17:30 |
| 22 | 265207 | 5 | 2019-11-08 02:17:30 |
+----+--------+---------+---------------------+
I have tried this sql:
SELECT
url_id,
user_id,
count( * ) AS visit_times
FROM
visit_actions
GROUP BY
user_id,
url_id
ORDER BY
visit_times DESC
I expect the output :
+--------+---------+-------------+
| url_id | user_id | visit_times |
+--------+---------+-------------+
| 265338 | 4 | 1 |
| 265207 | 5 | 2 |
| 265338 | 1 | 2 |
+--------+---------+-------------+
Each user only finds the one with the most url_id.
Click Here Online Demo . Thanks folks!

Answer 1

On MySQL 8+ a fairly clean solution uses ROW_NUMBER with aggregation:
WITH cte AS (
SELECT url_id, user_id, COUNT(*) AS cnt,
ROW_NUMBER() OVER (PARTITION BY user_id ORDER BY COUNT(*) DESC) rn
FROM visit_actions
GROUP BY url_id, user_id
)
SELECT
url_id,
user_id,
cnt AS visit_times
FROM cte
WHERE rn = 1;
If you had to do this on MySQL 5.7, here is one way:
SELECT
t1.url_id,
t1.user_id,
t1.cnt AS visit_times
FROM
(
SELECT url_id, user_id, COUNT(*) AS cnt
FROM visit_actions
GROUP BY url_id, user_id
) t1
INNER JOIN
(
SELECT user_id, MAX(cnt) AS max_cnt
FROM
(
SELECT url_id, user_id, COUNT(*) AS cnt
FROM visit_actions
GROUP BY url_id, user_id
) t
GROUP BY user_id
) t2
ON t1.user_id = t2.user_id AND
t1.cnt = t2.max_cnt;

Answer 2

Here's a MySQL 5.7 solution. Basically you have to find the maximum number of visits per user and then join the count of visits per user to that table to give the user and urls they have visited most. Note in your sample that yields 6 rows for user 1 as they have visited 6 sites twice.
SELECT c.url_id, c.user_id, c.visit_times
FROM (SELECT url_id, user_id, count( * ) AS visit_times
FROM visit_actions
GROUP BY user_id, url_id) c
JOIN (SELECT user_id, MAX(visit_times) AS max_visits
FROM (SELECT url_id, user_id, count( * ) AS visit_times
FROM visit_actions
GROUP BY user_id, url_id ) c
GROUP BY user_id) m ON m.user_id = c.user_id AND m.max_visits = c.visit_times
Output:
url_id user_id visit_times
265207 0 2
265338 1 2
265391 1 2
265394 1 2
265396 1 2
265410 1 2
265431 1 2
265338 4 1
Demo on SQLFiddle