Solution 1:

Since Primary Keys are UNIQUE by default, this makes this relation One to One.

No, that makes the relation "one to zero or one". Is that what you actually need?

If yes, then then your "second solution" is better:

  • it's simpler,
  • takes less storage1 (and therefore makes cache "larger")
  • hes less indexes to maintain2, which benefits data manipulation,
  • and (since you are using InnoDB) naturally clusters the data, so users that are close together will have their accounts stored close together as well, which may benefit cache locality and certain kinds of range scans.

BTW, you'll need to make accounts.id an ordinary integer (not auto-increment) for this to work.

If no, see below...

What is the best way to create One to One relation in MySQL?

Well, "best" is an overloaded word, but the "standard" solution would be the same as in any other database: put both entities (user and account in your case) in the same physical table.

Are there any other solutions other than these two?

Theoretically, you could make circular FKs between the two PKs, but that would require deferred constraints to resolve the chicken-and-egg problem, which are unfortunately not supported under MySQL.

And if I import any of these solutions into MySQL Workbench EER diagram, it recognizes relations as One to Many :S Thats also confusing.

I don't have much practical experience with that particular modeling tool, but I'm guessing that's because it is "one to many" where "many" side was capped at 1 by making it unique. Please remember that "many" doesn't mean "1 or many", it means "0 or many", so the "capped" version really means "0 or 1".


1 Not just in the storage expense for the additional field, but for the secondary index as well. And since you are using InnoDB which always clusters tables, beware that secondary indexes are even more expensive in clustered tables than they are in heap-based tables.

2 InnoDB requires indexes on foreign keys.

Solution 2:

Your first approach creates two candidate keys in the accounts table: id and user_id.

I therefore suggest the second approach i.e. using the foreign key as the primary key. This:

  • uses one less column
  • allows you to uniquely identify each row
  • allows you to match account with user