Question

I know this is pretty elementary but here it goes.

I would like to know how you know what columns are a primary key in a table that does not have a primary key? Is there a technique or something that I should read?

Thank you in advance

Answer 1

You need to take a look at your data structures.

A primary key must:

• never be NULL (no exceptions)
• reliably and uniquely identify each single row

and it helps if it’s

• small and easy to use
• stable (doesn’t change at all, or at least not often)
• a single column (or at most two)

Check your data – which columns or set of columns can fulfill these requirements??

Once you have those potential primary keys (the “candidate keys”) – think about how you will access the data, and what other data might need to be associated with this one entity in question – what would make sense as a foreign key? Do you want to reference your department by its name? Probably not a good idea, since the name could be misspelled, it might change over time etc. By the department’s office location? Bad choice, too. But something like a unique “department ID” might be a good idea.

If you don’t find any appropriate column(s) in your actual data that could serve as primary key and would make sense, it’s a common practice to introduce a “surrogate key” – an extra column, often an INT (and often something like an “auto-increment” INT) that will serve as an artificial identifier for each row. If you do this, one common best practice is to never show that artificial key on any data screen – it has no meaning whatsoever to the users of your system – so don’t even show it to them.

Checking these requirements, and a lot of experience, will help you find the right primary key.