Question

I’m having a problem in an oracle SQL project I’m working on:

I have a (weak) entity ‘Ticket’ that has as primary key: (Customer_id,packet_id,project_id,ticket_id). Customer_id, packet_id and project_id are also foreign keys.

But, a ticket can only be for a packet OR a project, and because they are both in the primary key, they can’t be null, while one of them actually will always be null. But I need both because a ticket always belongs to a one of them.

I thought of a possible solution and I thought I could make 1 ID for all products, but then there is another problem, because if I want to implement this product_id as a foreign key, I don’t know to which entity I should let it reference to.

Is there a way to make ‘optional primary keys’ or work with if-statements when creating tables? Or a way to make an optional reference? I’ve tried if-statements and a case but it didn’t work.

Answer 1

A lot will depend on exactly what these entities represent. Since a ticket can be for either a packet or a project, that implies that there is some higher-level entity that combines both packets and projects. That could either be modeled as a new table that is the parent to both PACKET and PROJECT or it could be modeled by combining the PACKET and PROJECT table and adding a TYPE column that differentiates between packets and projects. The TICKET table would then reference either the new parent table or the combined table.

Taking a step back, though, if the TICKET entity has a TICKET_ID, it seems likely that the TICKET_ID alone should be the primary key. It seems unusual that you would have a composite primary key if you have a TICKET_ID column– do you really want to allow multiple rows in the TICKET table with the same TICKET_ID that are related to different customers, packets, and projects? If TICKET_ID alone is the primary key, then both PACKET_ID and PROJECT_ID can be nullable foreign keys and you can create a CHECK constraint that ensures that exactly one of the two is non-NULL.