Question

Lets suppose I have a scenario with the following model: An Animal table which represents any animal, a Dog table and a Bird table, each one with a 1:1 relationship with the Animal table.

Animal

    INTEGER id (PK)
    STRING name

Bird

    INTEGER id (PK FK referencing `Animal.id`)

Dog

    INTEGER id (PK FK referencing `Animal.id`)                

(I’m giving only the keys just to be clear)

How can I guarantee that a given row in the Animal table will have JUST one referenced row in either the Dog or the Bird table? The model itself allows it…

An animal can’t be a Dog and a Bird at the same time (not in mythology but that’s not the case 😛)

It would be better if this could be done just by playing with the model, without triggers…

Any tip would be appreciated 🙂

Answer 1

This may be what @Pranay meant, but the answer was incomplete. Add a column TYPE to all tables and then constrain it like this:

create table Animal (id integer, 
                     type string,
                     name string,
                     primary key (id),
                     unique (id, type)
                    );

create table Bird (id integer,
                   type string default 'BIRD' check (type='BIRD'),
                   primary key (id),
                   foreign key (id, type) references Animal (id, type)
                  );

create table Dog (id integer,
                  type string default 'DOG' check (type='DOG'),
                  primary key (id),
                  foreign key (id, type) references Animal (id, type)
                 );

See David Portas’s blog fora good explanation of this.