Question

It is common to use SELECT within SELECT to reduce the number of queries; but as I examined this leads to slow query (which is obviously harmful for mysql performance). I had a simple query as

SELECT something
 FROM posts
 WHERE id IN (
  SELECT tag_map.id
  FROM tag_map
  INNER JOIN tags
  ON tags.tag_id=tag_map.tag_id
  WHERE tag IN ('tag1', 'tag2', 'tag3', 'tag4', 'tag5', 'tag6')
  )

This leads to slow queries of “query time 3-4s; lock time about 0.000090s; with about 200 rows examined”.

If I split the SELECT queries, each of them will be quite fast; but this will increase the number of queries which is not good at high concurrency.

Is it the usual situation, or something is wrong with my coding?

Answer 1

In MySQL, doing a subquery like this is a “correlated query”. This means that the results of the outer SELECT depend on the result of the inner SELECT. The outcome is that your inner query is executed once per row, which is very slow.

You should refactor this query; whether you join twice or use two queries is mostly irrelevant. Joining twice would give you:

SELECT something
FROM posts
INNER JOIN tag_map ON tag_map.id = posts.id
INNER JOIN tags ON tags.tag_id = tag_map.tag_id
WHERE tags.tag IN ('tag1', ...)

For more information, see the MySQL manual on converting subqueries to JOINs.

Tip: EXPLAIN SELECT will show you how the optimizer plans on handling your query. If you see DEPENDENT SUBQUERY you should refactor, these are mega-slow.