Question

I am creating a bulletin board application. Each bulletin can be liked or disliked by users of the site.To keep track of the likes and dislikes I have created the following database table

id  user_id   bulletin_id   like_dislike
1   1         1             1
2   1         2             0
3   3         1             1
4   2         1             0

In the like_dislike column 1 means ‘Like It’, 0 means ‘Don’t like it’
I know how to ask.
– How many times was bulletin 1 liked (2)
– How many times was bulletin 1 disliked (1)

But How do I do a query to ask those two questions at the same time? That is, how many times was bulletin 1 liked and disliked

liked  disliked
2      1

I have tried the query

  SELECT count(like_dislike) AS likes, count(like_dislike) AS dislikes FROM bulletins_ld
where bulletins_id = 1 AND likes = 1 AND dislikes = 0

but all I get is two twice which is not surprising.
The only solution I can think of is having a separate like and dislike column

Answer 1

You can do this with an aggregate query, using opposing conditions on the single like_dislike column (I am assuming below that a ‘1’ in that column means ‘liked’).

SELECT bulletin_id, 
       SUM(CASE WHEN like_dislike = 1 THEN 1 ELSE 0 END) AS likes, 
       SUM(CASE WHEN like_dislike = 0 THEN 1 ELSE 0 END) AS dislikes
FROM bulletins_ld
GROUP BY bulletin_id

Update: As per the discussion in the comments below, the like/dislike column could be normalized into its own table, like so (example deliberately silly…):

CREATE TABLE how_user_feels(
    feeling_id INT,
    feeling_desc VARCHAR(20)
)

INSERT INTO how_user_feels(feeling_id, feeling_desc) VALUES
(0, 'Undecided'),
(1, 'Likes It'),
(2, 'Could Do Without It')

The Likes_Dislikes column in the Bulletin table is then replaced by the foreign key feeling_id, with a default to 0. Let’s say that you then enter a record in this table when a user first views a bulletin, making them “Undecided” by default, and update that record when they vote on the bulletin. You could query the results like so:

SELECT bulletin_id, 
       SUM(CASE WHEN feelings_id = 1 THEN 1 ELSE 0 END) AS likes, 
       SUM(CASE WHEN feelings_id = 2 THEN 1 ELSE 0 END) AS dislikes,
       SUM(CASE WHEN feelings_id = 0 THEN 1 ELSE 0 END) AS doesnt_seem_to_care
FROM bulletins_ld b
INNER JOIN how_user_feels h ON b.feeling_id = h.feeling_id
GROUP BY bulletin_id

Keep in mind, this is just one approach, and may not be useful in your case. But if you ever decided to change or expand the model by which a user expresses their feelings for a bulletin, say to a five-star rating system, you could do that without changing the database schema – just alter the records in the how_user_feels table, and the associated queries.