Search code examples
sqlpostgresqldatabase-designforeign-keys

Refactor foreign key to fields

In PostgreSQL I need to refactor a table (Purchases); it has a foreign key to another table (Shop). Instead I want two fields that keep the relation in a textual way. I must NOT lose any information, the tables already contain data.

Purchases.shop_id: (long)          -- is the field I need to drop
Purchases.shop: (characters)       -- will hold the Shop's name
Purchases.shop_user: (characters)  -- will hold the Shop's user name.

Shop.id: (long, pk)      -- still referenced from Purchases
Shop.name: (characters)  -- Shop's name
Shop.user: (characters)  -- Shop's user name

Two fields are necessary because a Shop is unique on (name,user) (or by id of course).

ALTER TABLE Purchases ADD COLUMN shop CHARACTER VARYING(255);
ALTER TABLE Purchases ADD COLUMN shop_user CHARACTER VARYING(255);

-- ???

ALTER TABLE Purchases DROP CONSTRAINT shop_id_fk;
ALTER TABLE Purchases DROP COLUMN shop_id;

So the start and the ending is easy, can somebody help with middle-part? :)

I know that foreign keys were made for this but I have to do it this way.

Solution

  • Your seem to go the wrong way. Your original, normalized schema is typically superior. If you need to display shop / user, create a VIEW.

    But you may have your reasons, so here goes:

    UPDATE purchases p
SET   (shop, shop_user) = (s.name, s."user")
FROM   shop s
WHERE  s.id = p.shop_id;

    Don't use the reserved word "user" as identifier.
    And "name" is hardly ever a good name, either.
    And varchar(255) in Postgres typically indicates a misunderstanding.

    About varchar(255):