How to make MySQL intersect keys of INTEGER field and FLOAT field? - mysql

Question

I have the following MySQL table (table size - around 10K records):
CREATE TABLE `tmp_index_test` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`m_id` int(11) DEFAULT NULL,
`r_id` int(11) DEFAULT NULL,
`price` float DEFAULT NULL,
PRIMARY KEY (`id`),
KEY `m_key` (`m_id`),
KEY `r_key` (`r_id`),
KEY `price_key` (`price`)
) ENGINE=InnoDB AUTO_INCREMENT=16390 DEFAULT CHARSET=utf8;
As you can see, I have two INTEGER fields (r_id and m_id) and one FLOAT field (price).
For each of these fields I have an index.
Now, when I run a query with condition on the first integer AND on the second one, everything is fine:
mysql> explain select * from tmp_index_test where m_id=1 and r_id=2;
+----+-------------+----------------+-------------+---------------+-------------+---------+------+------+-------------------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+----------------+-------------+---------------+-------------+---------+------+------+-------------------------------------------+
| 1 | SIMPLE | tmp_index_test | index_merge | m_key,r_key | r_key,m_key | 5,5 | NULL | 1 | Using intersect(r_key,m_key); Using where |
+----+-------------+----------------+-------------+---------------+-------------+---------+------+------+-------------------------------------------+
Seems like MySQL performs it very well since there is the Using intersect(r_key,m_key) in the Extra field.
I'm not a MySQL expert, but according to what I understand, MySQL is first making the intersection on indexes, and only then collects the result of the intersection from the table itself.
HOWEVER, when I run very similar query, but instead of condition on two integers, I put similar condition on an integer and a float, MySQL refuses to intersect the result on indexes:
mysql> explain select * from tmp_index_test where m_id=3 and price=100;
+----+-------------+----------------+------+-----------------+-----------+---------+-------+------+-------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+----------------+------+-----------------+-----------+---------+-------+------+-------------+
| 1 | SIMPLE | tmp_index_test | ref | m_key,price_key | price_key | 5 | const | 1 | Using where |
+----+-------------+----------------+------+-----------------+-----------+---------+-------+------+-------------+
As you can see, MySQL decides to use the index of price only.
My first question is why, and how to fix it?
In addition to it, I need to run queries with MORE sign (>) instead of the equal sign (=) on price. Currently explain shows that for such queries, MySQL uses the integer key only.
mysql> explain select * from tmp_index_test where m_id=3 and price > 100;
+----+-------------+----------------+------+-----------------+-------+---------+-------+------+-------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+----------------+------+-----------------+-------+---------+-------+------+-------------+
| 1 | SIMPLE | tmp_index_test | ref | m_key,price_key | m_key | 5 | const | 2 | Using where |
+----+-------------+----------------+------+-----------------+-------+---------+-------+------+-------------+
I need to make somehow MySQL first do the intersection on indexes. Anybody has any idea how?
Thanks a lot in advance!

Answer 1

From the MySQL manual:
ref is used if the join uses only a leftmost prefix of the key or if
the key is not a PRIMARY KEY or UNIQUE index (in other words, if the
join cannot select a single row based on the key value). If the key
that is used matches only a few rows, this is a good join type.
price is not unique or primary, so ref is chosen. I don't believe you can force an intersect.