Restrict FOREIGN KEY to a part of the referenced table

From: Matthias Nagel <matthias(dot)h(dot)nagel(at)gmail(dot)com>
To: pgsql-sql(at)postgresql(dot)org
Subject: Restrict FOREIGN KEY to a part of the referenced table
Date: 2013-04-11 07:55:20
Message-ID: 3358559.iSkkklqms2@hek506
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-sql

Hello,

is there any best practice method how to create a foreign key that only allows values from those rows in the referenced table that fulfill an additional condition?

First I present two pseudo solutions to clarify what I would like to do. They are no real solutions, because they are neither SQL standard nor postgresql compliant. The third solution actually works, but I do not like it for reason I will explain later:

CREATE TABLE parent (
id SERIAL,
discriminator INT NOT NULL,
attribute1 VARCHAR,
...
);

Pseudo solution 1 (with a hard-coded value):

CREATE TABLE child (
id SERIAL NOT NULL,
parent_id INT NOT NULL,
attribute2 VARCHAR,
...,
FOREIGN KEY ( parent_id, 42 ) REFERENCES parent ( id, discriminator )
);

Pseudo solution 2 (with a nested SELECT statement):

CREATE TABLE child (
id SERIAL NOT NULL,
parent_id INT NOT NULL,
attribute2 VARCHAR,
...,
FOREIGN KEY ( parent_id ) REFERENCES ( SELECT * FROM parent WHERE discriminator = 42 ) ( id )
);

Working solution:

CREATE TABLE child (
id SERIAL NOT NULL,
parent_id INT NOT NULL,
parent_discriminator INT NOT NULL DEFAULT 42,
attribute2 VARCHAR,
...,
FOREIGN KEY ( parent_id, parent_discriminator ) REFERENCES parent ( id, discriminator ),
CHECK ( parent_discriminator = 42 )
);

The third solution work, but I do not like it, because it adds an extra column to the table that always contains a constant value for the sole purpose to be able to use this column in the FOREIGN KEY clause. On the one hand this is a waste of memory and on the other hand it is not immediately obvious to an outside person what the purpose of this extra column and CHECK clause is. I am convinced that any administrator who follows me might get into problems to understand what this is supposed to be. I would like to have a more self-explanatory solution like 1 or 2.

I wonder if there is something better.

Best regards, Matthias

----------------------------------------------------------------------
Matthias Nagel
Willy-Andreas-Allee 1, Zimmer 506
76131 Karlsruhe

Telefon: +49-721-8695-1506
Mobil: +49-151-15998774
e-Mail: matthias(dot)h(dot)nagel(at)gmail(dot)com
ICQ: 499797758
Skype: nagmat84

Responses

Browse pgsql-sql by date

  From Date Subject
Next Message androclos 2013-04-11 16:38:39 Slow update with ST_Contians()
Previous Message Gavin Flower 2013-04-11 00:00:20 Re: Advice for index design