Negative Primary Keys

Question

Are there any repercussions using Negative Primary Keys for tables (Identity Increment -1, Identity Seed -1 in SQL Server 2005)?

The reason for this is we\'re creati

Answer 1

You'll want to review legacy code and look for where developers have sorted on primary key as a lazy/sloppy way of sorting by date (because identity pk's usually are strongly or perfectly correlated to time).

Answer 2

This is perfectly fine from SQL Server's perspective. The real question is going to be your application.

Answer 3

Another option is to prefix the legacy keys with a string like "OLD_". The ony problem is your key field will be non-numeric.

If you have to have numeric keys, you could introduce a "legacy" indicator column, and the primary key would be a combination of the numeric ID and the legacy indicator (hopefully, that combination should be unique).

Answer 4

Not a problem.

It is slightly un-orthodox, but apart from that, fine.

The default offered by SQL Server is just that - a default, and can be changed to suit needs. Looks like you got a good compromise.

Answer 5

Only problem is you won't be able to add a third data source this way!

Answer 6

Like others have said, the database is fine with this.

But it would be a problems for a .NET application that uses DataSet+DataAdapter as they use negative keys as temporaries for new records.

Other data-access layers may use similar tricks.