MySQL indexes when multiple are possible - mysql

Question

Given the following -
drop table if exists learning_indexes;
create table learning_indexes (
id INT NOT NULL,
col1 CHAR(30),
col2 CHAR(30),
col3 CHAR(30),
PRIMARY KEY (id),
index idx_col1 (col1),
index idx_col1_col2 (col1,col2)
);
explain
select
col1,col2
from
learning_indexes
where
col1 = 'FOO'
and col2 = 'BAR'
Why does MySQL pick idx_col1 over idx_col1_col2?
+----+-------------+------------------+------+------------------------+----------+---------+-------+------+-------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+------------------+------+------------------------+----------+---------+-------+------+-------------+
| 1 | SIMPLE | learning_indexes | ref | idx_col1,idx_col1_col2 | idx_col1 | 91 | const | 1 | Using where |
+----+-------------+------------------+------+------------------------+----------+---------+-------+------+-------------+
This is my version information -
+-------------------------+---------------------+
| Variable_name | Value |
+-------------------------+---------------------+
| innodb_version | 1.1.8 |
| protocol_version | 10 |
| slave_type_conversions | |
| version | 5.5.29 |
| version_comment | Source distribution |
| version_compile_machine | i386 |
| version_compile_os | osx10.7 |
+-------------------------+---------------------+

Answer 1

I agree with Floaf that MySQL sometimes chooses the wrong indexes, but I don't think this is the case here. MySQL includes the number of rows and the data structure into its decision which index to choose.
For a rather simple query like this one, MySQL will likely not use any index at all if the table contains less than about 100 rows or is even empty. It seems to be computationally cheaper just to scan all table rows than to use the index. In your explain plan, you can see that the "key" column says idx_col1, but the "Extra" column doesn't say "using index".
If the table contains more than about 100 rows, MySQL will start using idx_col1. The explain plan will show you this. Only when there are more than about 100 rows that actually contain the string 'FOO' in col1, MySQL will notice that using idx_col1 doesn't reduce the tentative result set enough, since it will have to scan the remaining 100 rows for the value 'BAR' in col2. Therefore, it will switch to idx_col1_col2.
I'm not entirely sure how MySQL decides this quickly which index to use, but I think it has something to do with heuristics and the cardinality of the individual rows in the index, i.e. how "selective" an indexed row is.

Answer 2

I can't explain your case here, but sometimes MySQL simply chooses the "wrong" index. Maybe the database is small enough that it understands that it does not make any difference in this case.
This query is so simple that it should understand which index is the most appropriate.
I can say by experience that when the queries are getting more complex and especially when the tables grow very large, MySQL sometimes (random?) decides to pick another index and go with that and then queries can go from 0.01 second to 100+ seconds, so if you know which index is the right one, use FORCE INDEX(). Even if you use USE INDEX() MySQL sometimes chooses another index with various devestating result to the query speed.