MySQL optimizing orderby date query - mysql

Question

I did my best to solve the following two simple queries but for each 10 rows result set it scans the full table or at least 10K rows. Currently there are 20000 rows in books table.
ALTER TABLE books ADD INDEX search_INX (`book_status`, `is_reviewed`,`has_image`,`published_date`)
mysql> EXPLAIN SELECT book_id FROM books ORDER BY published_date DESC LIMIT 10;
+----+-------------+-------+-------+---------------+------------+---------+------+-------+-----------------------------+
| id | se ref |lect_type | table | type | possible_keys | key | key_len | rows | Extra |
+----+-------------+-------+-------+---------------+------------+---------+------+-------+-----------------------------+
| 1 | SIMPLE | books | index | NULL | search_INX | 11 | NULL | 20431 | Using index; Using filesort |
+----+-------------+-------+-------+---------------+------------+---------+------+-------+-----------------------------+
mysql> EXPLAIN SELECT book_id FROM books WHERE book_status='available' AND is_reviewed=true AND has_image=true ORDER BY published_date DESC LIMIT 10;
+----+-------------+-------+------+---------------+------------+---------+-------------------+-------+--------------------------+
| id | select_type | table | type ref || possible_keys | key | key_len | rows | Extra |
+----+-------------+-------+------+---------------+------------+---------+-------------------+-------+--------------------------+
| 1 | SIMPLE | books | ref | search_INX | search_INX | 3 | const,const,const | 10215 | Using where; Using index |
+----+-------------+-------+------+---------------+------------+---------+-------------------+-------+--------------------------+
mysql> EXPLAIN SELECT book_id FROM books WHERE book_status='available' AND is_reviewed=true AND has_image=true ORDER BY published_date DESC LIMIT 10\G
*************************** 1. row ***************************
id: 1
select_type: SIMPLE
table: books
type: ref
possible_keys: search_INX
key: search_INX
key_len: 3
ref: const,const,const
rows: 10215
Extra: Using where; Using index
1 row in set (0.00 sec)
Create Table: CREATE TABLE `books` (
`book_id` int(10) unsigned NOT NULL auto_increment,
`has_image` bit(1) NOT NULL default '',
`is_reviewed` bit(1) NOT NULL default '\0',
`book_status` enum('available','out of stock','printing') NOT NULL default 'available',
`published_date` datetime NOT NULL,
PRIMARY KEY (`book_id`),
KEY `search_INX` (`is_reviewed`,`has_image`,`book_status`,`published_date`)
) ENGINE=InnoDB AUTO_INCREMENT=162605 DEFAULT CHARSET=latin1
1 row in set (0.00 sec)
Does any one have clue to how to solve this problem?

Answer 1

The cardinality of:
KEY `search_INX` (`is_reviewed`,`has_image`,`book_status`,`published_date`)
...is poor. If you were to put published_date at the front, it would speed up your query. Further, why are you indexing is_reviewed & has_image? Boolean columns cannot even be indexed in the likes of SQL Server as there is no point to doing so (again, cardinality). Either rearrange your key, or put a unique key on the column I mentioned.

Answer 2

Also, mysql in rows column show not the number of rows that was affected, but the approximate amount of rows that could be affected, excluding the LIMIT clause.

Answer 3

At a quick glance, the problem seems to be you're missing an index on published_date. Order by is using this column. Add this index and see what happens.

Answer 4

If you use the FORCE INDEX command, does that help?

Answer 5

I'm no expert on indexing, but can you create an index just on published_date as well as the index you have made on all four fields?
ALTER TABLE books DROP INDEX `search_INX`;
ALTER TABLE books ADD INDEX `published_INX` (`published_date`);

Answer 6

#Zerkms #jason I just thought of another way to solve this.
It is unorthodox, but would work. If your primary key was (publish_date, book_id) with DESC sort, you would easily be able to get the last 10 results. The query engine would scan the table, applying the where clause, until it found 10 results, then quit.
It would work great. Just add another index on book_id if you need to specifically query by book_id.
The reason this makes sense is the DB would naturally store books by date (InnoDB uses clustered indexes), which is exactly what you are trying to query.