MySQL Indexes on InnoDB table - not working right? - mysql

Question

I am reading the book High Performance MySQL and messing around with a new database testing somethings.
I am not sure if I am doing something wrong though..
I have a table called table_users
Structure:
ID(Integer)
FullName(Char)
UserName(Char)
Password(Char)
SecurityID(TinyINT)
LocationID(TinyINT)
Active(TinyINT)
My indexes are as follows:
PRIMARY : ID
FullName : UNIQUE : FullName
FK_table_users_LocationID (foreign key reference) : INDEX : LocationID
FK_table_users_SecurityID (foreign key reference) : INDEX : SecurityID
Active : INDEX : Active
All are BTREE
While reading the book, I am trying to use the following mysql statement to view the extras involved with a SELECT statement
EXPLAIN
SELECT * FROM table_users WHERE
FullName = 'Jeff';
No matter what the WHERE statement points to with this call, the extra result is either nothing or Using where. If I SELECT ID ... WHERE FullName = 'Jeff' it returns Using where, Using Index. But not whenever I do SELECT FullName .... WHERE FullName = 'Jeff'..
I am not familiar at all with indexes and trying to wrap my head around them bit having a bit of confusion with this. Shouldn't they return Using Index if I am referencing an indexed column?
Thanks.

Answer 1

Using index doesn't mean what it seems to mean. Have a look at covering indexes. If it says "using index" it means that mysql could return the data for your query without reading the actual rows. SELECT * - is only going to be able to use a covering index if even column of the table is in the index. Usually this is not the case.
I seem to remember a Chapters in High Performance Mysql that talks about covering indexes and how to read EXPLAIN results.

Answer 2

What version of MySQL are you using? Here's a test I ran on Percona Server 5.5.16:
mysql> create table table_users (
id int auto_increment primary key,
fullname char(20),
username char(20),
unique key (fullname)
);
Query OK, 0 rows affected (0.03 sec)
mysql> insert into table_users values (default, 'billk', 'billk');
Query OK, 1 row affected (0.00 sec)
mysql> explain select * from table_users where fullname='billk'\G
*************************** 1. row ***************************
id: 1
select_type: SIMPLE
table: table_users
type: const
possible_keys: fullname
key: fullname
key_len: 21
ref: const
rows: 1
Extra:
1 row in set (0.00 sec)
This shows that it's using the fullname index, looking up by a constant value, but it's not an index-only query.
mysql> explain select fullname from table_users where fullname='billk'\G
*************************** 1. row ***************************
id: 1
select_type: SIMPLE
table: table_users
type: const
possible_keys: fullname
key: fullname
key_len: 21
ref: const
rows: 1
Extra: Using index
1 row in set (0.00 sec)
This is as expected, it's able to get the fullname column from the fullname index, so this is an index-only query.
mysql> explain select id from table_users where fullname='billk'\G
*************************** 1. row ***************************
id: 1
select_type: SIMPLE
table: table_users
type: const
possible_keys: fullname
key: fullname
key_len: 21
ref: const
rows: 1
Extra: Using index
1 row in set (0.00 sec)
Searching on fullname but fetching the primary key is also an index-only query, because the leaf nodes of InnoDB secondary indexes (e.g. the unique key) implicitly contain the primary key value. So this query is able to traverse the BTREE for fullname, and as a bonus it gets the id too.
mysql> explain select fullname, username from table_users where fullname='billk'\G
*************************** 1. row ***************************
id: 1
select_type: SIMPLE
table: table_users
type: const
possible_keys: fullname
key: fullname
key_len: 21
ref: const
rows: 1
Extra:
1 row in set (0.00 sec)
As soon as the select-list includes any column that's not part of the index, it can no longer be an index-only query. First it searches the BTREE for fullname, to find the primary key value. Then it uses that id value to traverse the BTREE for the clustered index, which is how InnoDB stores the whole table. There it finds the other columns for the given row, including username.