Question

I have a SQL that looks like this:

SELECT * FROM foo WHERE foo.bar IN (SELECT foobar.bar FROM foobar)

This is not good, right? The nested SELECT will cause things to slow down? How should I query something like this?

Answer 1

The IN clause is perfectly valid SQL, but it’s not always the preferred way. I’ve often found that MySQL’s performance on them is pretty poor, even if foo.bar is indexed.

If you change it to a join, you have to be careful if the second table has multiple rows that match the join condition for each row of the first table, because the join will produce multiple result rows. If this is possible, the join should be:

SELECT f.*
FROM foo f
INNER JOIN (SELECT DISTINCT bar FROM foobar) fb USING (bar);

The ultimate answer is that you should use EXPLAIN to see how different forms of the query will be executed. But if you’re not experiencing a performance problem in the first place, don’t sweat it.

If you want all the rows where f.bar does not exist in foobar, NOT IN is usually the most succinct way to write it. But it can also be written using a LEFT OUTER JOIN:

SELECT f.*
FROM foo f
LEFT OUTER JOIN foobar fb USING (bar)
WHERE fb.bar IS NULL

You don’t need the subquery in this case because you’re only reporting the non-matching rows, so there obviously can’t be multiple matches in the result.