sql delete all but 2 duplicates - mysql

Question

I want to be able to limit the amount of duplicate records in a mySQL database table to 2.
(Excluding the id field which is auto increment)
My table is set up like
id city item
---------------------
1 Miami 4
2 Detroit 5
3 Miami 4
4 Miami 18
5 Miami 4
So in that table, only row 5 would be deleted.
How can I do this?

Answer 1

MySQL has some foibles when reading and writing to the same table. So I don't actually know if this will work, the syntax is fine in many implementations of SQL, but I don't know if it's MySQL friendly...
DELETE
yourTable
WHERE
1 < (SELECT COUNT(*)
FROM yourTable as Lookup
WHERE city = yourTable.city AND item = yourTable.item AND id < yourTable.id)
EDIT
Amazingly convoluted, but worth a try?
DELETE
yourTable
FROM
yourTable
INNER JOIN
(
SELECT
id
FROM
(
SELECT
id
FROM
yourTable
WHERE
1 < (SELECT COUNT(*)
FROM yourTable as Lookup
WHERE city = yourTable.city AND item = yourTable.item AND id < yourTable.id)
)
AS inner_deletes
)
AS deletes
ON deletes.id = yourTable.id

Answer 2

I think your problem here is that both your code and/or table structure allows inserting duplicates and you are asking this question when you should really fix your db and/or code.

Answer 3

i think a better solution is avoid allow more than 5 registers, you have to implement a validation where if select count(*) > 3 you will not accept the new insert.
because if you want to do this into the data tier, you have to use a stored procedure , because first you need to identify all the register with more than 3 registers and delete only the last .
Saludos

Answer 4

Due to MySQL being notoriously difficult when it comes to updating queried tables (see for example the answers from Dems), the best I can figure out is sadly more than one statement but on the plus side fairly readable;
CREATE TEMPORARY TABLE Dump AS SELECT id FROM table1 WHERE id NOT IN
(SELECT MIN(id) FROM table1 GROUP BY city,item UNION
SELECT MAX(id) FROM table1 GROUP BY city,item);
DELETE FROM table1 where id in (select * from Dump);
DROP TABLE DUMP;
Not sure if it was important which duplicate was removed, this keeps the first and last.

Answer 5

In your reply to Joachim's answer, you ask about saving 3 or 5 rows, this is one way to accomplish it. Depending on how you are using this database, you could either call this in a loop, or you could turn it into a stored procedure. Either way, you would continue to run this entire block of code until Rows Affected = 0:
drop table if exists TempTable;
create table TempTable
select city, item,
count(*) as record_count,
min(id) as ItemToDrop -- this could be changed to max() if you
-- want to delete new stuff instead
from YourTable
group by city, item
having count(*) > 2; -- This value = number of rows you save
delete from YourTable
where id in (select ItemToDrop from TempTable);