MySQL group-by very slow

问题

I have the folowwing SQL query

SELECT CustomerID FROM sales WHERE `Date` <= '2012-01-01' GROUP BY CustomerID

The query is executed over 11400000 rows and runs very slow. It takes over 3 minutes to execute. If I remove the group-by part, this runs below 1 second. Why is that?

MySQL Server version is '5.0.21-community-nt'

Here is the table schema:
CREATE TABLE `sales` (
  `ID` int(11) NOT NULL auto_increment,
  `DocNo` int(11) default '0',
  `CustomerID` int(11) default '0',
  `OperatorID` int(11) default '0',
  PRIMARY KEY  (`ID`),
  KEY `ID` (`ID`),
  KEY `DocNo` (`DocNo`),
  KEY `CustomerID` (`CustomerID`),
  KEY `Date` (`Date`)
) ENGINE=MyISAM AUTO_INCREMENT=14946509 DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci

回答1:

Try putting an index on (Date,CustomerID).

Have a look at the mysql manual for optimizing group by queries:- Group by optimization

You can find out how mysql is generating the result if you use EXPLAIN as follows:-

EXPLAIN SELECT CustomerID FROM sales WHERE `Date` <= '2012-01-01' GROUP BY CustomerID

This will tell you which indexes (if any) mysql is using to optimize the query. This is very handy when learning which indexes work for which queries as you can try creating an index and see if mysql uses it. So even if you don't fully understand how mysql calculates aggregate queries you can create a useful index by trial and error.

回答2:

Without knowing what your table schema looks like, it's difficult to be certain, but it would probably help if you added a multiple-column index on Date and CustomerID. That'd save MySQL the hassle of doing a full table scan for the GROUP BY statement. So try ALTER TABLE sales ADD INDEX (Date,CustomerID).

回答3:

try this one :

SELECT distinct CustomerID FROM sales WHERE `Date` <= '2012-01-01'

回答4:

I had the same problem, I changed the key fields to the same Collation and that fix the problem. Fields to join the tables had different Collate value.

回答5:

Wouldn't this one be a lot faster and achieve the same?

SELECT DISTINCT CustomerID FROM sales WHERE `Date` <= '2012-01-01'

Make sure to place an index on Date, of course. I'm not entirely sure but indexing CustomerID might also help.

来源：https://stackoverflow.com/questions/10278878/mysql-group-by-very-slow

标签

mysql

performance

group-by