Deleting Duplicates in MySQL

Question

I have a table like this

userid  visitorid   time
1       10          2009-12-23
1       18          2009-12-06
1       18          2009-12-14
1       18             


        

Answer 1

You need to work around MySQL bug#6980, with a doubly nested subquery:

DELETE FROM foo_table
WHERE foo_table.time IN (
    SELECT time FROM (
        SELECT time FROM
            foo_table
            LEFT OUTER JOIN (
                SELECT MAX(time) AS time
                FROM foo_table
                GROUP BY userid, visitorid
                ) AS foo_table_keep
                    USING (time)
        WHERE
            foo_table_keep.time IS NULL
        ) AS foo_table_delete
    );

Using GROUP BY collapses duplicates down to a single row, and MAX(time) chooses which value you want. Use another aggregate function than MAX if you want.

Wrapping the subquery twice, providing aliases for each, avoids the error:

ERROR 1093 (HY000): You can't specify target table 'foo_table' for update in FROM clause

and has the extra advantage that it's clearer how the statement is choosing what to keep.