Max 3 rows in table for each user - php

Question

Is it possible to make a script without else/if that deletes the oldest row for a user if the new row is the 4 row for that specific user?
I have a table called points_history. Fields are:
date(datetime),
fk_player_id(int),
points(int)
Here is my insert:
mysqli_query($mysqli,"INSERT INTO points_history (date,fk_player_id,points) VALUES (NOW(),$player,$points)");
The reason for this taht I want to be able to go back in the players history and check points, but only the last 3 points and don't want a table with a million of rows.
Can it be done in one sql query?
Hoping for help and thanks in advance :-)

Answer 1

This is quite easy to do if you add a primary key to your table points_history.
Part 1:
Use the following script to add a primary key called points_history_id to your table:
ALTER TABLE points_history RENAME TO points_history_old;
CREATE TABLE points_history
(
`points_history_id` int(11) UNSIGNED NOT NULL AUTO_INCREMENT,
`date` datetime NOT NULL,
`fk_player_id` int(11) NOT NULL,
`points` int(11) NOT NULL,
PRIMARY KEY (`points_history_id`)
);
INSERT INTO points_history (date, fk_player_id, points)
SELECT date, fk_player_id, points
FROM points_history_old;
-- Drop table if migration succeeded (up to you)
-- DROP TABLE points_history_old;
This needs to be run only once!
Part 2:
Now you can use the following SQL script to add a new record and delete the outdated:
-- First insert the new record
INSERT INTO points_history (date,fk_player_id,points)
VALUES (NOW(),:player,:points);
-- Create temporary table with records to keep
CREATE TEMPORARY TABLE to_keep AS
(
SELECT points_history_id
FROM points_history
WHERE fk_player_id = :player
ORDER BY date DESC
LIMIT 3
);
SET SQL_SAFE_UPDATES = 0;
-- Delete all records not in table to_keep
DELETE FROM points_history
WHERE points_history_id NOT IN (SELECT points_history_id FROM to_keep);
SET SQL_SAFE_UPDATES = 1;
-- Drop temporary table
DROP TEMPORARY TABLE to_keep;
If you use a database supporting transactions, I strongly recommend to wrap this script in a transaction. I tested it on MySQL 5.5.29 and it runs fine.