MySQL Subqueries

November 15, 2004

MYSQL 4.1 has been released as a production version, and with all the new features it is likely a new generation of developers will soon be using MySQL. This month I revisit an old area of contention - subqueries. Until now, MySQL has not supported subqueries, and this lack caused many to write off MySQL as not being a serious DBMS. While the lack was certainly a problem, many developers do not know that subqueries can often be rewritten as a join, sometimes giving a performance benefit in the process. This month I look at subqueries, and how they can be rewritten in a more optimal way.

Rewriting subqueries as joins

First, let's create some sample tables and data:


Click for full CREATE TABLE code

Rewriting subqueries as Inner Joins

Let's look at a simple subquery to return a list of all animals that have been assigned food:

mysql> SELECT name 
  FROM animal 
  WHERE id 
  IN (SELECT animal_id FROM animal_food);
+---------------------+
| name                |
+---------------------+
| Chacma Baboon       |
| Small Grey Mongoose |
+---------------------+
2 rows in set (0.00 sec)

Now, the same query as an inner join.

mysql> SELECT DISTINCT name 
  FROM animal,animal_food 
  WHERE animal.id=animal_id;
+---------------------+
| name                |
+---------------------+
| Chacma Baboon       |
| Small Grey Mongoose |
+---------------------+
2 rows in set (0.00 sec)

Note the DISTINCT keyword. Without it, we will get duplicate records, one animal returned for each record in the animal_food table:

mysql> SELECT name FROM animal,animal_food WHERE animal.id=animal_id;
+---------------------+
| name                |
+---------------------+
| Chacma Baboon       |
| Chacma Baboon       |
| Small Grey Mongoose |
| Small Grey Mongoose |
+---------------------+
4 rows in set (0.01 sec)

Which query is better? Some may argue that the former query is more readable, but I think that to anyone vaguely competent in SQL they should both be clear. However, one is definitely more efficient than the other is. Using EXPLAIN, let's see which. First, the join:


Click for full EXPLAIN SELECT DISTINCT code

MySQL would have to examine 12 rows to return the result set (4x3, as you multiply the result - for more on EXPLAIN see Optimizing MySQL: Queries and Indexes. Now the second query:


Click for full EXPLAIN SELECT code

This time MySQL has to examine 16 rows, a marginal difference in such a small example, but this difference can add up in larger datasets.

Rewriting subqueries as Outer Joins

A common kind of query is finding all records that do not have an associated record in another table--in this case, finding all animals that do not have an associated food. Here is how you would do it with a subquery:

 
mysql> SELECT name FROM animal WHERE id NOT IN (SELECT animal_id FROM animal_food);
+-------------+
| name        |
+-------------+
| Cape Gerbil |
| Porcupine   |
+-------------+
2 rows in set (0.00 sec)

Writing this sort of query without a subquery often taxes novice MySQL developers, but it is actually quite easy, simply requiring a LEFT JOIN and an IS NULL clause:

mysql> SELECT name FROM animal LEFT JOIN  animal_food ON animal.id=animal_id WHERE animal_id IS NULL;
+-------------+
| name        |
+-------------+
| Cape Gerbil |
| Porcupine   |
+-------------+
2 rows in set (0.00 sec)

This time the former query is certainly easier to understand, but does the extra complexity of the second query deliver a performance gain?


Click for full EXPLAIN SELECT (2) code

I can hear some sighs of relief - the subquery in this case is as efficient as the outer join. Outer joins can often deliver performance gains though, so benchmark your queries. Note that an alternative way to write the same subquery is as follows (this is as efficient in this case)

mysql> SELECT name FROM animal 
  WHERE NOT EXISTS (SELECT animal_id FROM animal_food WHERE animal.id=animal_id);
+-------------+
| name        |
+-------------+
| Cape Gerbil |
| Porcupine   |
+-------------+







The Network for Technology Professionals

Search:

About Internet.com

Legal Notices, Licensing, Permissions, Privacy Policy.
Advertise | Newsletters | E-mail Offers