Query response time with WHERE and HAVING

Question

I have a simple SQL request:

SELECT id, title, resum
FROM film
WHERE id = 56 ;

The response time for this request is nearly equal to 0.0009 sec

Whereas, the same query, using Having Clause :

SELECT id, title, resum
FROM film
HAVING id =56 ;

Give such a big response time compered with the 1st request (0.0049 sec).

The 1st request is more optimized, But, Why ? Can you tell me the reason that cause this big difference ?

Best Regards.

Answer 1

At first, let's state the most important thing: Your second query isn't a valid SQL query as it uses the MySQL GROUP BY extensions. When you switch off this MySQL extension with

SET SESSION sql_mode='ONLY_FULL_GROUP_BY';

then you'll get an error like this:

ERROR 1463 (42000): non-grouping field 'id' is used in HAVING clause

So, you should really use the WHERE clause.

But now to your actual question. I assume from your measurements, that there is an index on the "id" field. Because HAVING works (should work) on grouped data, there's no index it could use. I've got a MySQL table here with about 1.2 million rows. A HAVING query on an indexed integer field takes 16 seconds on its first run and still about 0.6 seconds on consecutive calls, whereas the query using WHERE just takes 0.04 seconds.

Using EXPLAIN, MySQL will tell you that it doesn't use an index:

EXPLAIN SELECT id, title, resum FROM film HAVING id =56;

As an example, here are the results from EXPLAIN for the queries on my table:

mysql> EXPLAIN SELECT id, Title FROM `test` HAVING id = 4374354;
+----+-------------+----------+------+---------------+------+---------+------+---------+-------+
| id | select_type | table    | type | possible_keys | key  | key_len | ref  | rows    | Extra |
+----+-------------+----------+------+---------------+------+---------+------+---------+-------+
|  1 | SIMPLE      | test     | ALL  | NULL          | NULL | NULL    | NULL | 1201750 |       |
+----+-------------+----------+------+---------------+------+---------+------+---------+-------+

You see, the "key" field states "NULL", telling you that there isn't any index used. The "rows" field tells you instead, that MySQL traverses 1201750 (all) rows.

mysql> EXPLAIN SELECT id, Title FROM `test` WHERE id = 4374354;
+----+-------------+----------+-------+---------------+---------+---------+-------+------+-------+
| id | select_type | table    | type  | possible_keys | key     | key_len | ref   | rows | Extra |
+----+-------------+----------+-------+---------------+---------+---------+-------+------+-------+
|  1 | SIMPLE      | test     | const | PRIMARY       | PRIMARY | 4       | const |    1 |       |
+----+-------------+----------+-------+---------------+---------+---------+-------+------+-------+

On the other hand, the EXPLAIN for WHERE tells us, that it uses the "PRIMARY" index and thus it just has to read a single row, resulting in a much faster response.