In the following schema:


I'm tempted to have a single Upvotes table with a single "entityId" column that stores either the CollectionId, ReportId, or ReviewId. There will also be an enum for Type - storing either collection, report, or review.

The entityId will always be required, and the logic will always make sure that inserts enforce uniqueness across each type.

The benefits of this is that adding another type is just a matter of expanding the enum. Everything will live on a single table with no redundancy.

The cost seems to be the added complexity on the logic side, which is contained to the one place in my application logic that inserts new entities into the table.

Practically speaking, is there anything wrong with this approach? What would be some other reasons to avoid this?

  • @yannishristofakis - I'm sorry, I meant polymorphic association. I have three entities whose FKs I'm treating as a single FK in the upvote table, with no referential integrity. I'll throw a diagram together shortly :) – RobVious Dec 8 '13 at 18:16

I can't recommend polymorphic associations.

In my experience, this will lead to data consistency issues since you can't use a foreign key constraint in the schema. That means you're reliant on your application's writes to this part of the database being bug-free, and no one manually altering the data.

It can also make the queries which join Upvotes to >1 of the other tables more complicated/harder to understand.

  • 1
    You would you mind to giving us a better implementation. I was looking for something better a couple of days ago. – giannis christofakis Dec 8 '13 at 19:56
  • 1
    A separate upvotes table for each of collections/reports/reviews. – Matt Ball Dec 8 '13 at 20:03
  • Thnk you Matt for your explanation. – giannis christofakis Dec 8 '13 at 20:15

Not the answer you're looking for? Browse other questions tagged or ask your own question.