Martijn Tonies wrote:

>News for you :-)
>
>If you add a required column (not NULL) to a table with existing data,
>and include a default for that column, you'll get this behaviour.
>
>
>

You're kidding? When did that happen? Did anyone actually think about
it first?

When a constraint is added, one can make an argument that the table
should be validated after adding the constraint, then removing the
constraint and throwing an error. My preference is to add the
constraint in the assumption that the user has or will fix the data.

To make sure that all backups are restorable, provision should be made
to relax the constraint so the data can be loaded and fixed.