mysql query optimization - join and orderby - sql

Question

I need a little help by one mysql query optimization. It is a simple query but anything is not right and I can't found it :-(
I have 2 tables: products (> 40000 Rows) and product_tags (> 5 mil)
There is a relation betweet the tables 1 -> N . Every prdoduct can have many tags in the table product tags.
I have this simple Query:
EXPLAIN SELECT t.product_id, kwt.tag_id
FROM products AS t, product_tags AS kwt
WHERE 1
AND t.product_id = kwt.product_id
AND kwt.tag_id =11
ORDER BY t.order_date
wchich returns 55 results.
First Situation: if I have this table structure of the tables:
CREATE TABLE IF NOT EXISTS `products` (
`product_id` int(10) unsigned NOT NULL auto_increment,
`product_source_id` smallint(5) unsigned NOT NULL,
`order_date` int(10) unsigned NOT NULL,
PRIMARY KEY (`product_id`),
KEY `order_date` (`order_date`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 ;
CREATE TABLE IF NOT EXISTS `product_tags` (
`product_tag_id` int(10) unsigned NOT NULL auto_increment,
`tag_id` int(10) unsigned NOT NULL,
`product_id` int(11) NOT NULL,
PRIMARY KEY (`product_tag_id`),
KEY `product_id` (`product_id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 ;
Then the Explain of the query is this:
+----+-------------+-------+-------+---------------+------------+---------+---------------------------+-------+-------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+-------+-------+---------------+------------+---------+---------------------------+-------+-------------+
| 1 | SIMPLE | t | index | PRIMARY | order_date | 4 | NULL | 45392 | Using index |
| 1 | SIMPLE | kwt | ref | product_id | product_id | 4 | t.product_id | 3 | Using where |
+----+-------------+-------+-------+---------------+------------+---------+---------------------------+-------+-------------+
It is getting all the rows from table products, but there is nothing with temporary table.
Second Situation: If I add an index for the field "tag_id" in product_tags, then the picture is different:
+----+-------------+-------+--------+-------------------+---------+---------+-----------------------------+------+---------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+-------+--------+-------------------+---------+---------+-----------------------------+------+---------------------------------+
| 1 | SIMPLE | kwt | ref | product_id,tag_id | tag_id | 4 | const | 55 | Using temporary; Using filesort |
| 1 | SIMPLE | t | eq_ref | PRIMARY | PRIMARY | 4 | kwt.product_id | 1 | Using where |
+----+-------------+-------+--------+-------------------+---------+---------+-----------------------------+------+---------------------------------+
Now it selects only 55 rows, what is right, but the query is havy :(
Where is my mistake here ?
Thanks
Nik

Answer 1

this is what i would do:
Read these resources
http://dev.mysql.com/doc/refman/5.0/en/innodb-index-types.html
http://www.xaprb.com/blog/2006/07/04/how-to-exploit-mysql-index-optimizations/
Redesign your system to take advantage of a clustered primary key
Simplified schema:
drop table if exists products;
create table products
(
prod_id int unsigned not null auto_increment primary key,
name varchar(255) not null unique
)
engine = innodb;
drop table if exists tags;
create table tags
(
tag_id mediumint unsigned not null auto_increment primary key,
name varchar(255) not null unique
)
engine = innodb;
drop table if exists product_tags;
create table product_tags
(
tag_id mediumint unsigned not null,
prod_id int unsigned not null,
created_date date not null,
primary key (tag_id, prod_id), -- note the clustered composite index and the order !!
key (prod_id)
)
engine = innodb;
Test the schema
select
pt.tag_id,
pt.prod_id
from
product_tags pt
inner join products p on pt.prod_id = p.prod_id
where
pt.tag_id = 11
order by
pt.created_date
limit 10;
I may even change the product_tags PK to primary key (tag_id, prod_id, created_date) but it all depends on the typical queries you run. You could ofc, just create a non clustered secondary index on created date if you think that's gonna boost performance.
Hope this helps :)