Question

Is there a reason to use a single incrementing field for a primary key instead of multiple fields that actually represent the unique record?

I’m working on an existing php application, and the tables all seem to have a single ‘id’ key instead of using the 2 or more fields that are actually unique to the record (like user, auction, bid).

I’m not a database expert, but that just seems lazy (or inexperienced) to me. Is there any benefit (performance or otherwise)?

Updated: I’m not referring to psudo-unique data (ssn, e-mail address, etc), where you may want to ensure the data is really unique. I’m talking about tables with obvious foreign key references, but instead of using those references along with the unique field(s) in the table itself, every table just has an incrementing ID.

Not trying to start up a subjective debate, it just didn’t make sense to me.

Answer 1

Using synthetic primary keys has several advantages:

• You can change values in key fields without having to take an index update hit
• The indexes are smaller
• It makes foreign key relationships simpler
• Since you’re not dealing with strings, there are never encoding issues

Databases often have specific optimizations around building indexes with monotonically incrementing keys.

That being said, there is nothing wrong with a little denormalization now and then. If the use-case is clear and the tables are relatively small, do what’s convenient.