COUNT and GROUP BY on text fields seems slow

问题

I'm building a MySQL database which contains entries about special substrings of DNA in species of yeast. My table looks like this:

+--------------+---------+------+-----+---------+-------+
| Field        | Type    | Null | Key | Default | Extra |
+--------------+---------+------+-----+---------+-------+
| species      | text    | YES  | MUL | NULL    |       |
| region       | text    | YES  | MUL | NULL    |       |
| gene         | text    | YES  | MUL | NULL    |       |
| startPos     | int(11) | YES  |     | NULL    |       |
| repeatLength | int(11) | YES  |     | NULL    |       |
| coreLength   | int(11) | YES  |     | NULL    |       |
| sequence     | text    | YES  | MUL | NULL    |       |
+--------------+---------+------+-----+---------+-------+

There are approximately 1.8 million records. In one type of query I want to see how many DNA substrings are associated with each type of species and region, so I issue this query:

select species, region, count(*) group by species, region;

The species and region columns have only two possible entries (conserved/scer for species, and promoter/coding for region) yet this query takes about 30 seconds.

Is this a normal amount of time to expect for this type of query given the size of the table? Is it slow because I'm using text fields instead of simple integer or boolean values (I prefer text fields as several non-CS researchers will be using the DB). Any other ideas and suggestions would be welcome.

Please excuse if this is a boneheaded question, I am an SQL neophyte.

P.S. I've also seen this question but the proposed solution doesn't seem relevant for what I'm doing.

EDIT: Converting those fields to VARCHARs reduced the runtime to ~2.5 seconds. Note I also timed it against ENUMs which had a similar timing.

回答1:

Why're all your string based columns defined as TEXT? If you read the performance comparison, you'll see that TEXT was ~3x slower than a VARCHAR column using identical indexing: http://forums.mysql.com/read.php?24,105964,105964

回答2:

If your fields are only ever going to have 2 values, you're much better off making them booleans. You should also make everything NOT NULL unless there's a real reason you'll need it to be NULL.

Also take a look at the ENUM type for a better way to use a finite number of human-readable values for a column.

As for slowness, the first thing to try is to create indices on your columns. For the particular query you're showing here, an index on species, region should make a huge difference:

create index on mytablename (species, region);

should do it.

来源：https://stackoverflow.com/questions/3305437/count-and-group-by-on-text-fields-seems-slow