Question

Only for MSSQL 2000-2005 by SQL query and no stored procedures

No cursors and temp tables

The table

create table my_table
(
row_id int identity(1,1),
_key varchar(20),
_total decimal(18,2)
)

Insert the data

insert into my_table (_key,_total) Values('qwe',10)
insert into my_table(_key,_total) Values ('qwe',10)
insert into my_table(_key,_total) Values ('asd',10)
insert into my_table(_key,_total) Values('asd',10)
insert into my_table(_key,_total) Values('zxc',10)
insert into my_table(_key,_total) Values('zxc',10)
insert into my_table(_key,_total) Values('qwe',100)
insert into my_table(_key,_total) Values('qwe',100)
insert into my_table(_key,_total) Values('asd',100)
insert into my_table(_key,_total) Values('asd',100)
insert into my_table(_key,_total) Values('zxc',100)
insert into my_table(_key,_total) Values('zxc',100)
insert into my_table(_key,_total) Values('qwe',50)
insert into my_table(_key,_total) Values('qwe',50)
insert into my_table(_key,_total) Values('asd',50)
insert into my_table(_key,_total) Values('asd',50)
insert into my_table(_key,_total) Values('zxc',50)
insert into my_table(_key,_total) Values('zxc',50)

I need to delete duplicates by _key field and leave rows with maximum _total and among them maximum row_id.

I need to get this result set

8 qwe 100.00
10 asd 100.00
12 zxc 100.00

Answer 1

To delete them, which is what I figured “remove duplicates” means:

DELETE FROM dbo.my_table
 WHERE NOT EXISTS(SELECT NULL                       
                    FROM dbo.my_table x
                    JOIN (SELECT _key,
                                  MAX(_total) AS max_total
                             FROM dbo.my_table
                         GROUP BY _key) y ON y._key = x._key
                                         AND y.max_total = x._total 
                   WHERE x._key = my_table._key
                     AND x._total = my_table._total
                GROUP BY x._key, x._total
                  HAVING MAX(x.row_id) = my_table.row_id)

The query to select the rows you list is:

  SELECT MAX(x.row_id),
         x._key,
         x._total
    FROM dbo.my_table x
    JOIN (SELECT _key,
                 MAX(_total) AS max_total
            FROM dbo.my_table
        GROUP BY _key) y ON y._key = x._key
                      AND y.max_total = x._total 
GROUP BY x._key, x._total