I have a question and answer website like stackoverflow. Here is the structure of some tables:
-- {superfluous} means some other columns which are not related to this question
// q&a
+----+-----------------+--------------------------+------+-----------+-----------+
| id | title | body | type | related | author_id |
+----+-----------------+--------------------------+------+-----------+-----------+
| 1 | How can I ... | I'm trying to make ... | q | NULL | 3 |
| 2 | | You can do that by ... | a | 1 | 1 |
| 3 | Why should I .. | I'm wonder, why ... | q | NULL | 1 |
| 4 | | First of all you ... | a | 1 | 2 |
| 5 | | Because that thing ... | a | 3 | 2 |
+----+-----------------+--------------------------+------+-----------+-----------+
// users
+----+--------+-----------------+
| id | name | {superfluous} |
+----+--------+-----------------+
| 1 | Jack | |
| 2 | Peter | |
| 3 | John | |
+----+--------+-----------------+
// votes
+----+----------+-----------+-------+-----------------+
| id | user_id | post_id | value | {superfluous} |
+----+----------+-----------+-------+-----------------+
| 1 | 3 | 4 | 1 | |
| 2 | 1 | 1 | -1 | |
| 3 | 2 | 1 | 1 | |
| 4 | 3 | 2 | -1 | |
| 5 | 1 | 4 | 1 | |
| 6 | 3 | 5 | -1 | |
+----+--------+-------------+-------+-----------------+
// tags
+----+------------+-----------------+
| id | name | {superfluous} |
+----+------------+-----------------+
| 1 | PHP | |
| 2 | SQL | |
| 3 | MySQL | |
| 4 | HTML | |
| 5 | CSS | |
| 6 | C# | |
+----+------------+-----------------+
// q&aTag
+-------+--------+
| q&aid | tag_id |
+-------+--------+
| 1 | 1 |
| 1 | 4 |
| 3 | 5 |
| 3 | 4 |
| 4 | 6 |
+-------+--------+
Now I need to find top users in a specific tag. For example, I need to find Peter
as top user in PHP tag. Because his answer for question1 (which has PHP
tag) has earned 2 upvotes. Is doing that possible?
Try this:
select q1.title, u.id, u.name, sum(v.value) total from `q&a` q1
left join `q&atag` qt ON q1.id = qt.`q&aid`
inner join tags t ON qt.tag_id = t.id
left join `q&a` q2 ON q2.related = q1.id
left join users u ON q2.author_id = u.id
left join votes v ON v.post_id = q2.id
where t.name = 'PHP'
group by q1.id, u.id
and here is a simple divided solution:
Let us divide it into sub queries:
id
of the tag you will search for: select id from tags where name = 'PHP'
select 'q&aid' from 'q&aTag' where tag_id = 1.
id
s of answers for that question: select id, author_id from
q&awhere related in (2.)
select user_id, sum(value) from votes where post_id in (3.) group by user_id
Now combining them all give the result:
select user_id, sum(`value`) total from votes
where post_id in (
select id from `q&a` where related in (
select `q&aid` from `q&aTag` where tag_id IN (
select id from tags where name = 'PHP'
)
)
)
group by user_id
you can add this at the end if you want only one record:
order by total desc limit 1