views:

120

answers:

1

I want to fetch a set of Posts w/ vote count listed, sorted by vote count (e.g.)

Post 1 - Post Body blah blah - Votes: 500
Post 2 - Post Body blah blah - Votes: 400
Post 3 - Post Body blah blah - Votes: 300
Post 4 - Post Body blah blah - Votes: 200

I have 2 tables:

Posts - columns - id, body, is_hidden
Votes - columns - id, post_id, vote_type_id

Here is the query I've tried:

SELECT p.*, v.yes_count
FROM posts p
LEFT JOIN
    (SELECT post_id, vote_type_id, COUNT(1) AS yes_count
    FROM votes 
    WHERE (vote_type_id = 1) 
    GROUP BY post_id
    ORDER BY yes_count DESC
    LIMIT 0, 10) v 
ON v.post_id = p.id
WHERE (p.is_hidden = 0)
ORDER BY yes_count DESC
LIMIT 0, 10

Correctness: The above query almost works. The subselect is including votes for posts that have is_hidden = 1, so when I left join it to posts, if a hidden post is in the top 10 (ranked by votes), I can end up with records with NULL on the yes_count field.

Performance: I have ~50k posts and ~500k votes. On my dev machine, the above query is running in .4sec. I'd like to stay at or below this execution time.

Indexes: I have an index on the Votes table that covers the fields: vote_type_id and post_id

EXPLAIN

id  select_type  table  type  possible_keys  key  key_len  ref  rows  Extra
1  PRIMARY  p  ALL  NULL  NULL  NULL  NULL  45985  Using where; Using temporary; Using filesort
1  PRIMARY  <derived2>  ALL  NULL  NULL  NULL  NULL  10   
2  DERIVED  votes  ref  VotingPost  VotingPost  4     319881  Using where; Using index; Using temporary; Using filesort
A: 

Try

SELECT p.*, count(*) yes_count
FROM posts p
LEFT OUTER JOIN votes v ON (v.post_id = p.id and v.vote_type_id = 1)
WHERE p.is_hidden = 0
GROUP BY p.id
ORDER BY yes_count DESC
LIMIT 0,10

Since this is mysql, group by only p.id s will work (but this won't be portable to other dbs)

DrewM
this is correct, but it is taking 1 second to execute. i am thinking about storing the vote count in each post record..
try adding an index on vote_type_id.post_id. That should speed it up.
DrewM
already have that index, still 1-1.1 sec execute