Hacker News Reader

Comment by OJFord | Hacker News Reader

Sounds like a 'star schema', just without foreign key constraints on the relationship ('fact') table.

I'm not really sure I understand why not having that constraint is helpful - what sort of 'changes' are significantly cheaper? Schema changes on the thing ('dimension') tables? Or do you mean cheaper in the 'reasoning about it' sense, you can delete a thing without having to worry about whether it should be nullified, deleted, or what in the relationship table? (And sure something starts to 404 or whatever, but in this application that's fine.)

View on HN

> you can delete a thing without having to worry about whether it should be nullified, deleted, or what in the relationship table?

Pretty much this. You could just delete all the relationships, or just some of them. For example, a user could be related to many things (appointments, multiple vendors, customers, etc). Thus a customer can just delete the relationship between them and a user, but the user still keeps their other relationships with the customer (such as appointments, allowing them to see their appointment history, even though the customer no longer has access to the user's information).

wouldn't you be able to do the same even with FK in place? what FK would prevent you from doing would be deleting a user before all his relationships are removed. That should prevent you to ending up with inconsistent data inside the tables.

In my experience with regular Postgres databases shared across many teams, inconsistent/broken references is seen as a much less risky thing to do than, say, cascading a delete. Usually a broken reference is of no consequence. And you don’t use FKs, people tend to bake that into how they code things without much trouble.

I've oft wished for a system where FKs could be used to automatically mark rows as invalid instead of deleting them, and you could inspect the nature of the broken reference as a property of the row/object. Then broken references can be cleaned up when appropriate, possibly never depending on the application.

This gives the neat property of being able to do data entry in parallel, too - different people may insert records into two tables linked by a FK relationship, it's nice for this to happen in either order. It also allows the user to query for exactly what rows will be affected by a cascaded delete, or to perform the operation in steps.

Wouldn't you effectively get those features by just not using FKs? You'd get your broken-reference marker by `where foreign_id not in (select id from foreign_table)`. You could create that as a view or materialized view.

It would be nice to have the schema metadata, at the very least, to be able to analyze a given tables reverse dependencies without resorting to naming conventions.

Reply on news.ycombinator.com