Need some help with putting this query together. I'm using Mysql
I have two tables
Video - contains videos uploaded by users
- video_id
- user_id
- category_id
Vote - contains a vote given by any user for a particular video
- vote_id
- video_id
- user_id
I don't want to hardcode the categories in the query - the Categories are stored in the Category table which has category_id and category_name
I basically want a query that pulls the top 3 videos (ones with max votes) for each category.
Sample Data - Video Table
video_id | user_id | category_id 1 100 10 2 101 10 3 102 11 4 103 11 5 104 11 6 105 11 7 105 12
Sample Data - Vote Table
vote_id | video_id | user_id 11 3 105 12 3 102 13 3 111 14 3 121 15 4 200 16 4 201 17 1 222
Sample Data - Category Table
category_id | category_name 10 HipHop 11 Rap 12 Country
This is the type of problem that is trivial to solve with ranking functions. However, since MySQL does not yet support them, it makes it more difficult. In this design, I assumed that video_id
was the primary key of the Video table.
Select video_id, user_id, category_id, vote_count, vote_rank
From (
Select VoteCounts.video_id, VoteCounts.user_id
, VoteCounts.category_id, VoteCounts.vote_count
, (
Select Count(*) + 1
From (
Select V1.video_id, V1.user_id, V1.category_id
, Count(vote_id) As vote_count
From Videos As V1
Left Join Votes As V2
On V2.video_id = V1.video_id
Group By V1.video_id, V1.user_id, V1.category_id
) As VoteCounts1
Where VoteCounts1.category_id = VoteCounts.category_id
And (
VoteCounts1.vote_count > VoteCounts.vote_count
Or (VoteCounts1.vote_count = VoteCounts.vote_count
And VoteCounts1.video_id < VoteCounts.video_id )
)
) As vote_rank
From (
Select V1.video_id, V1.user_id, V1.category_id
, Count(vote_id) As vote_count
From Videos As V1
Left Join Votes As V2
On V2.video_id = V1.video_id
Group By V1.video_id, V1.user_id, V1.category_id
) As VoteCounts
) As VoteRanks
Where VoteRanks.vote_rank <= 3
I don't know about MySQL, but here's how I would try to do it in SQL Server, and bringing perhaps some adjustments depending on the results which make sense.
select Category
, Film
, MAX(Votes) as Votes
from (
select TOP 3 c.category_name as Category
, vd.video_id as Film
, COUNT(vt.votes) as Votes
from Category c
inner join Video vd on vd.category_id = c.category_id
inner join Votes vt on vt.video_id = vd.video_id
where c.category_id = 10 -- HipHop
group by c.category_name
, vd.video_id
union
select TOP 3 c.category_name
, vd.video_id
, COUNT(vt.votes) as Votes
from Category c
inner join Video vd on vd.category_id = c.category_id
inner join Votes vt on vt.video_id = vd.video_id
where c.category_id = 11 -- Rap
group by c.category_name
, vd.video_id
-- You need as much UNION as there are categories.
) Top3VotesPerCategory
group by Category
, Film
Depending whether you accepte downvotes on a film, one could register a negative vote a film, and this wouldn't mean that he recommends this video. If you do support such functional feature, then you might want to consider a conditional SUM()
, instead of a COUNT()
, for the number of votes, so that negative votes would be considered and would decrease its overall points in regards of votes.
精彩评论