Question

I’ve tested the following seemingly simple query on MySQL 5.0, 5.1, 5.5 and found it to be extremely slow.

select * from entry where session_id in
    (select session_id from entry where created_at > [some timestamp])

Multiple entry’s can have the same session ID, but different created_at timestamps.
The query is meant to grab all entry’s that have at least one entry from the same session_id whose created_at is greater than the specified timestamp.

I’ve seen others speak of MySQL subquery performance issues with similar queries, and that MySQL considers the subquery a dependent query and it is doing a full table scan on the outer query. Suggested workarounds were something like:

select * from entry where session_id in
    (select session_id from
        (select session_id from entry where created_at > [some timestamp])
    as temp)

However, this hack doesn’t work for me and makes it even slower.

Any ideas on how to rewrite this query?

Answer 1

Depending on your data distribution, use this

SELECT  e.*
FROM    (
        SELECT  session_id, MAX(created_at)
        FROM    entry
        GROUP BY
                session_id
        HAVING  MAX(created_at) > $mytimestamp
        ) ed
JOIN    entry e
ON      e.session_id = ed.session_id

(create an index on (session_id, created_at)), or this:

SELECT  DISTINCT e.*
FROM    entry ed
JOIN    entry e
ON      e.session_id = ed.session_id
WHERE   ed.created_at > $mytimestamp

(create two separate indexes on created_at and session_id)