MySQL Query does not use index in table join

Question

I am trying to list all the book_sales information for a particular book author. So I have a query and it's not using Index to lookup records.

The following is my tables structure:

-- Table structure for table `books`

CREATE TABLE IF NOT EXISTS `books` (
  `book_id` int(11) NOT NULL auto_increment,
  `author_id` int(11) unsigned NOT NULL,
  `book_type_id` int(11) NOT NULL,
  `book_title` varchar(50) NOT NULL,
  `book_price` smallint(4) NOT NULL,
  `in_stock` char(1) NOT NULL,
  PRIMARY KEY  (`book_id`),
  KEY `book_type_id` (`book_type_id`),
  KEY `author_id` (`author_id`)
) ENGINE=InnoDB  DEFAULT CHARSET=latin1 AUTO_INCREMENT=1 ;

-- Dumping data for table `books`

INSERT INTO `books` (`book_id`, `author_id`, `book_type_id`, `book_title`, `book_price`, `in_stock`) VALUES
(1, 1, 1, 'My Book 1', 10, 'y'),
(2, 2, 1, 'My Book 2', 20, 'n'),
(3, 1, 2, 'My Book 3', 30, 'y'),
(4, 3, 3, 'My Book 4', 40, 'y'),
(5, 4, 2, 'My Book 5', 50, 'n'),
(6, 1, 1, 'My Book 6', 60, 'y'),
(7, 5, 3, 'My Book 7', 70, 'n'),
(8, 6, 2, 'My Book 8', 80, 'n'),
(9, 7, 1, 'My Book 9', 90, 'y'),
(10, 8, 3, 'My Book 10', 100, 'n');

-- Table structure for table `book_sales`

CREATE TABLE IF NOT EXISTS `book_sales` (
  `sale_id` int(11) NOT NULL auto_increment,
  `book_id` int(11) NOT NULL,
  `sale_amount` decimal(8,2) NOT NULL default '0.00',
  `time` datetime NOT NULL default '0000-00-00 00:00:00',
  `price` smallint(8) NOT NULL,
  PRIMARY KEY  (`sale_id`),
  KEY `book_id` (`book_id`),
  KEY `price` (`price`)
) ENGINE=InnoDB  DEFAULT CHARSET=latin1 AUTO_INCREMENT=1 ;

-- Dumping data for table `book_sales`

INSERT INTO `book_sales` (`sale_id`, `book_id`, `sale_amount`, `time`, `price`) VALUES
(1, 1, '10.00', '2010-02-23 10:00:00', 20),
(2, 1, '20.00', '2010-02-24 11:00:00', 20);

My Query:

SELECT sale_amount, price
FROM book_sales
INNER JOIN books ON book_sales.book_id = books.book_id
WHERE books.author_id =1

An EXPLAIN on the above, shows me:

id   select_type    table        type        possible_keys      key      key_len    ref     rows      Extra
1   SIMPLE          books       ref     PRIMARY,author_id   author_id   4         const    3        Using index
1   SIMPLE          book_sales  ALL     book_id             NULL        NULL      NULL     2        Using where

Clearly, book_sales is not using the key 'book_id', although I have it. What can I do to make the book_sales table use the Index?

Thank you.

Edits done based on suggestions (but the result is they still do not use index):

//Does not use the index in book_sales table
EXPLAIN SELECT sale_amount, price
FROM books, book_sales
FORCE INDEX ( book_id ) 
WHERE book_sales.book_id = books.book_id
AND books.author_id =1

//Does not use the index in book_sales table
EXPLAIN SELECT sale_amount, price
FROM book_sales, books
WHERE books.author_id = 1
AND book_sales.book_id = books.book_id

How to force the book_sale table with just 2 rows, to use the index ? Thank you.

Answer 1

As you can see in the EXPLAIN, "book_id" is listed as a possible key. If MySQL doesn't use it, it's just that the optimizer doesn't think it would speed up the query. Which is true if "book_sales" only has 2 rows, and 100% of those rows share the same "book_id". It's called cardinality btw. How to Avoid Table Scans (MySQL Manual)

Try filling it with more rows and you should see that MySQL will use an index for the join.

Edit: the query

SELECT sale_amount, price
FROM books, book_sales
FORCE INDEX ( book_id ) 
WHERE book_sales.book_id = books.book_id
AND books.author_id =1

...will not work either in that case because the optimizer still recognizes that reading the index is suboptimal and switches the table order to avoid doing so. You can force the table order by using STRAIGHT_JOIN. This is, however, a bit of a hack because it forces MySQL to execute the query in a way that is not the best.

      EXPLAIN
       SELECT sale_amount, price
         FROM books
STRAIGHT_JOIN book_sales FORCE INDEX (book_id) ON book_sales.book_id = books.book_id
        WHERE books.author_id = 1