I have been catching up on my feed reader‘s backlog and I found some interesting posts on primary keys in sql that I thought I’d share:
I don't have any real strong opinions on the subject, just some thoughts:
I do think composite keys are a pain in most situations. And not because of any ORM tools as implied in some comments in the above links. Most of the database code I write is still straight sql: sometimes stored procedures, sometimes parameterized queries.
In the databases I work on, nearly every table has an id field that is an auto-incrementing integer. This works well I think. It's consistent, which is always good. Joins are easy. The sql is easy (sometimes)
When designing a table I have a concept in my head of what makes a record unique. I don't think of it as a “key” usually. A key, at least the way most tools and books talk about it, is a single field; usually meaningless outside the database. The tools never seem to push the term “surrogate key.” Just primary keys and foreign keys.
Any information that is determined to be unique to a record is enforced with a constraint, but it's not referred to as a primary key. At least going by how I see the tools used.
Is it just a difference in terminology? Or, is there more to it? I don‘t think I’d do anything fundamentally differently if I thought in terms of primary-composite keys and surrogate keys. It‘s something I’ll have to keep in mind next time I'm creating a table.