How is MySQL using an index when doing a case sensitive comparison?

Question

I noticed that Ruby on Rails (ActiveRecord) was using the BINARY operator for a case sensitive comparison. Yet, when I look at the EXPLAIN for this, MySQL can still use the appropriate index for a case insensitive column. How is that possible? Are all text columns indexed case sensitivly?

Without BINARY operator:

  select_type: SIMPLE
        table: my_table
         type: const
possible_keys: my_table_unq,my_table_idx
          key: my_table_unq
      key_len: 771
          ref: const,const
         rows: 1
        Extra: Using index

With BINARY operator:

  select_type: SIMPLE
        table: my_table
         type: range
possible_keys: my_table_unq,my_table_idx
          key: my_table_unq
      key_len: 771
          ref: NULL
         rows: 1
        Extra: Using where; Using index

Looks like the later is using a range.

Answer 1

Index is case insensitive in your case, but optimizer is clever enough to use it efectively anyway.

In first explain, mysql uses ref access method - that means that it is able to directly fetch rows with given value.

In second explain, it says "using where" and uses range access - it takes the string, fetches all values from index searching for that string case insensitive, and for each value it checks the (case senstive/BINARY) WHERE condition to decide if it should be really returned - efectiveness of the access is the same, only drawback is that the case sensitive comparison has to be done afterwards, which is only small penalty compared to table scan which would have to compare every row the same way.

Answer 2

Note the link you gave on the BINARY operator where is it says the following

Posted by Yann Neuhaus on July 20 2005 12:34pm

Per default the search operation in not case sensitive, example, looking for 'HYPE BEAU' returns 'Hype Beau' in the table product :

mysql> select prod_name, prod_id from products where prod_name = 'HYPE BEAU';

+-----------+---------+
| prod_name | prod_id |
+-----------+---------+
| Hype Beau |    9600 |
| Hype Beau |   25860 |
| Hype Beau |   42120 |
| Hype Beau |   46185 |
+-----------+---------+

4 rows in set (0.01 sec)

This is because the default collation is case insentitive. With the binary clause the binary collation is used and the search becomes case sensitive :

If the character set's collation is case insensitive, then using the BINARY operator is not necessary. It would only appear to work fine in that instance.

For more clarification on collations being case insensitive, please read the MySQL Documentation section "The _bin and binary Collations"