Question

I have three tables, bookmarks, tag and tagging. tagging is the association table for linking tags to bookmarks. I would like to select related tags. The current query takes ages to run, but I could not wrap my head around the JOINs that could replace the subquery that selects all the related bookmarks that are tagged with specific tag IDs.

SELECT ta.name, count(*) 
FROM tagging t2 
JOIN tag ta ON t2.tag_id=ta.id  
WHERE t2.bookmark_id IN (
   SELECT bookmark_id 
   FROM tagging t1 
   WHERE t1.tag_id IN (1, 7) 
   GROUP BY t1.bookmark_id 
   HAVING COUNT(t1.id) = 2
) 
GROUP BY ta.id

Here is the result of the EXPLAIN of that query:

1   PRIMARY            ta  index PRIMARY    PRIMARY         8   NULL         3   
1   PRIMARY            t2  ref   tag_id_idx tag_id_idx      8   blinkz.ta.id 89   Using where
2   DEPENDENT SUBQUERY t1  index tag_id_idx bookmark_id_idx 8   NULL         71   Using where 

I have an index for the bookmark_id and tag_id fields in the tagging table. The tag table has a primary key named id

How could this query be optimised, preferably with a JOIN instead of a subquery?

Answer 1

The poor performance is due to a bug in MySQL that will hopefully be fixed soon. Until then try this instead:

SELECT tag.name, COUNT(*) AS cnt
FROM tagging T1
JOIN tag
ON T1.tag_id = tag.id  
JOIN (
   SELECT bookmark_id 
   FROM tagging
   WHERE tag_id IN (1, 7) 
   GROUP BY bookmark_id 
   HAVING COUNT(id) = 2
) T2
ON T1.bookmark_id = T2.bookmark_id
GROUP BY tag.id

I’m assuming that there is a unique contraint on (bookmark_id, tag_id).