MySQL multiple columns as unordered sets - mysql

Question

I'm somewhat new to MySQL and SQL in general, so hopefully this isn't a simple question.
I have a table that represents items in a customer's basket at checkout. This table represents a situation in which a customer is limited to 3 items, so I currently have a column for each item in the basket. It looks like this:
+------------------------------------------------------+
+ id | item1 | item2 | item3 | val |
+------------------------------------------------------+
where val is just some value associated with the basket. The ordering of the items means nothing in terms of my processing, so in theory I would like to have them represented as an unordered set. This means that a row of (i1,i2,i3,val) is functionally equivalent to (i2,i1,i3,val).
My question is, how do I implement this in my table and/or in SQL such that selecting (i3,i2,i1,val) will return the row for (i1,i2,i3,val)?
I also need to have something that catches uniqueness when I'm inserting. For example, if I insert (i2,i3,i1,newval), I would want the table to update (i1,i2,i3,val) to be (i1,i2,i3,newval).

Answer 1

You could standardise your model by using a 0 to many relation between customer and item:
-- assuming the existing table to be named `yourtable`
-- assuming your customer's table to be named `customer`
-- assuming your customer's id in the customer's table to be named `id`
-- assuming innodb (remove fk constraint if not)
CREATE TABLE `customer_item` (
`id` INT(10) NOT NULL AUTO_INCREMENT,
`id_customer` INT(10) NOT NULL,
`item` VARCHAR(255) NOT NULL,
PRIMARY KEY (`id`)
)
ENGINE=innodb
SELECT NULL AS `id`, t.id AS `id_customer`, t.`item`
FROM (
SELECT id, item1 AS `item`
FROM
yourtable
UNION
SELECT id, item2 AS `item`
FROM
yourtable
UNION
SELECT id, item3 AS `item`
FROM
yourtable
) t
ORDER BY t.id ASC
;
CREATE INDEX UNIQUE `idx_customer_item_cust` ON `customer_item` (`id_customer`, `item`);
ALTER TABLE `customer_item` ADD CONSTRAINT `fk_customer_item_cust` FOREIGN KEY (`id_customer`) REFERENCES `customer` (`id`) ON DELETE CASCADE ON UPDATE CASCADE;
-- once you check the data is consistent:
DROP TABLE `yourtable`;
Once that done, no item could be possibly inserted twice for the same customer.
Please note:
the union select skips duplicates already at table creation, in case some items were repeated for some customers
your data is normalised, from the customer to item point of view
your data is still not normalised, from the item point of view. You should have an item table, and the customer_item table should reference the id of items in the item table instead of using item names or description for varchars.