SQL Many-to-Many Query Problem

Question

I have three tables: videos, videos_categories, and categories.

The tables look like this:

videos: video_id, title, etc...
videos_categories: video_id, category_id
categories: category_id, name, etc...

In my app, I allow a user to multiselect categories. When they do so, I need to return all videos that are in every selected category .

I ended up with this:

SELECT * FROM videos WHERE video_id IN (
    SELECT c1.video_id FROM videos_categories AS c1
    JOIN c2.videos_categories AS c2
    ON c1.video_id = c2.video_id
    WHERE c1.category_id = 1 AND c2.category_id = 2
)

But for every category I add to the multiselect, I have to add a join to my inner select:

SELECT * FROM videos WHERE video_id IN (
    SELECT c1.video_id FROM videos_categories AS c1
    JOIN videos_categories AS c2
    ON c1.video_id = c2.video_id
    JOIN videos_categories AS c3
    ON c2.video_id = c3.video_id
    WHERE c1.category_id = 1 AND c2.category_id = 2 AND c3.category_id = 3
)

I can't help but feel this is the really wrong way to do this, but I'm blocked trying to see the proper way to go about it.

Answer 1

if this is a primary key:

 videos_categories: video_id, category_id

then a GROUP BY and HAVING should work, try this:

SELECT
    * 
    FROM videos 
    WHERE video_id IN (SELECT 
                           video_id
                           FROM videos_categories
                           WHERE category_id IN (1,2,3)
                           GROUP BY video_id
                           HAVING COUNT(video_id)=3
                      )

Answer 2

Here's a FOR XML PATH solution:

--Sample data
CREATE TABLE Video
    (
    VideoID int,
    VideoName varchar(50)
    )

CREATE TABLE Videos_Categories
    (
    VideoID int,
    CategoryID int
    )

INSERT  Video(VideoID, VideoName)
SELECT  1, 'Indiana Jones'
UNION ALL
SELECT  2, 'Star Trek'

INSERT Videos_Categories(VideoID, CategoryID)
SELECT 1, 1
UNION ALL
SELECT 1, 2
UNION ALL
SELECT 1, 3
UNION ALL
SELECT 2, 1
GO

--The query
;WITH   GroupedVideos
AS
(
SELECT  v.*,
        SUBSTRING(
                    (SELECT  (', ') + CAST(vc.CategoryID AS varchar(20))
                    FROM     Videos_Categories AS vc
                    WHERE    vc.VideoID = v.VideoID
                    AND      vc.CategoryID IN (1,2)
                    ORDER BY vc.CategoryID
                    FOR XML PATH('')), 3, 2000) AS CatList
FROM    Video AS v
)

SELECT  *
FROM    GroupedVideos
WHERE   CatList = '1, 2'

(Ignore everything below - I misread the question)

Try

WHERE c1.category_id IN (1,2,3)

or

...
FROM videos v
JOIN Vedeos_categories vc ON v.video_id = vc.video_id
WHERE vc.category_id IN (1,2,3)

Multiple joins aren't at all necessary.

Edit: to put the solutions in context (I realize it's not obvious):

SELECT * 
FROM videos 
WHERE video_id IN 
(    SELECT c1.video_id 
FROM videos_categories AS c1
WHERE c1.category_id = IN (1,2,3))

or

SELECT *
FROM videos v
JOIN Vedeos_categories vc ON v.video_id = vc.video_id
WHERE vc.category_id IN (1,2,3)

Answer 3

Sounds similar to SQL searching for rows that contain multiple criteria

To avoid having to another join for each category (and hence changing the structure of the query), you can put the categories into a temp table and then join against that.

CREATE TEMPORARY TABLE query_categories(category_id int);
INSERT INTO query_categories(category_id) VALUES(1);
INSERT INTO query_categories(category_id) VALUES(2);
INSERT INTO query_categories(category_id) VALUES(3);

SELECT * FROM videos v WHERE video_id IN (
  SELECT video_id FROM video_categories vc JOIN query_categories q ON vc.category_id = qc.category_id
  GROUP BY video_id
  HAVING COUNT(*) = 3
)

Although this is ugly in its own way, of course. You may want to skip the temp table and just say 'category_id IN (...)' in the subquery.