Question

I have a simple schema in which soft deletes are used (that’s how it is designed and that can’t be changed). There are two tables that participate in the schema: Company (id, is_deleted) and Employee (id, company_id, is_deleted) where company_id ofcourse is a FK to the Company table. The rules are:

• If a Company has is_deleted = true, then all Employee referring to that company should have is_deleted = true.
• But an Employee may have is_deleted = true even if the parent Company has is_deleted = false.

My two problems are a) how to enforce these constraints? b) how to easiest ensure that is_deleted = true is cascaded when a Company is soft-deleted.

I added the tags postgresql and sql server because those are the databases I’m mostly interested in. If there are other solutions in other rdbms:es I’d like to hear about them too.

Answer 1

Strictly speaking, the only way to cascade values like that is by using ON UPDATE CASCADE. To do that, the column “is_deleted” has to be part of a unique constraint.

That alone isn’t too hard. If company.id is your primary key, then the pair of columns {id, is_deleted} will also be unique. A unique constraint on that pair of columns would allow you to cascade updates through a foreign key reference.

But that won’t work in your case, because you need to allow referencing values to be different from the referenced values.

So in your case, I think you have three options.

• Triggers
• Stored procedures
• Application code

In all those cases, you need to pay attention to permissions (probably revoking delete permissions) and to cases that can avoid your code. For example, the dbms command-line interface and GUI interface can be used to get around constraints in application code and, depending on permissions, in stored procedures.