Given a user, select the user with most common ratings - mysql

Question

Say I have a table of ratings:
create table ratings (
user_id int unsigned not null,
post_id int unsigned not null,
rating set('like', 'dislike') not null,
primary key (user_id, post_id)
);
And a given user with id 1, how can I select the user with more likes in common? And the user with more dislikes in common? And the user with more ratings (likes or dislikes) in common? I guess that the queries would be very similar, buy I can't figure any of them out yet. I'll update with any progress I make.
Any help is appreciated, thanks!

Answer 1

select
r1.user_id as user1
,r2.user_id as user2
,r1.rating as rating
,count(*) as num_matching_ratings
from
ratings r1
inner join ratings r2
on r1.post_id = r2.post_id
and r1.rating = r2.rating
and r1.user_id <> r2.user_id --don't want to count
--matches with self
where
r1.user_id = 1 -- change this to any user, or use a
-- variable to increase reusebility
and r1.rating = 'like' -- set this to dislike to common dislikes
group by
r1.user_id
,r2.user_id
,r1.rating
having
count(*) > 1 --show only those with more than 1 in common
order by
count(*) desc
/* limit 1 -- uncomment to show just the top match */
By joining the tables together, we can count the number of occurances where the second user has rated an article similarly. This query will return the evalution from the most in common to the least. If you uncomment the "limit 1" statement, it will only return the match with the most in common.

Answer 2

Give this a try:
select r2.user_id from (
select post_id, rating from ratings,
(select #userId := 2) init
where user_id = #userId
) as r1
join ratings r2
on r1.post_id = r2.post_id and r1.rating = r2.rating
where r2.user_id != #userId and r2.rating = 'like'
group by r2.user_id
order by count(*) desc
limit 1
It should work for likes and dislikes by changing the string. And to change the user just modify the variable assignation.
The following should work for both dislikes and likes in common (just by removing the filtering condition):
select r2.user_id from (
select post_id, rating from ratings,
(select #userId := 2) init
where user_id = #userId
) as r1
join ratings r2
on r1.post_id = r2.post_id and r1.rating = r2.rating
where r2.user_id != #userId
group by r2.user_id
order by count(*) desc
limit 1

Answer 3

pardon my syntax, i don't write raw sql very often. you can consider this psudocode.
first, i'd get the table where id is 1
view1 = SELECT * FROM ratings, WHERE ( user_id = 1)
then i'd join it with ratings
view2 = select * from view1, ratings, where(view1.rating = ratings.rating AND view1.post_id = records.post_id)
then i'd aggregate by count
view3 = select count from view2 group by (user_id)
and then i'd get the max of that.
now, that's only an algorithmic overview of what my first thoughts would be. I don't think it would be particularly efficient, and you probably wouldn't use that syntax.

Answer 4

Building on Chris's and Mostacho's answers, I made the following query. I'm not 100% sure that it works every time, but I havent found a flaw yet.
select r2.user_id
from ratings r1
join ratings r2
on r1.user_id <> r2.user_id
and r1.post_id = r2.post_id
and r1.rating = r2.rating
where r1.user_id = 1
and r1.rating = 'like'
group by r2.user_id
order by count(r2.user_id) desc
limit 1
This query returns the id of the user with more common likes with the user 1. To fetch the user with more common ratings, just remove and r1.rating = 'like' from the where clause.