>>>>> "Hans-J" == Hans-J Krause <[EMAIL PROTECTED]> writes:
Hans-J> A database *never* should accept an insert of an empty string
Hans-J> in a not null column - otherwise what for should a
Hans-J> not-null-check to be set?
If you said "Oracle database", I'd agree. (And, given the rest of
your reply, you're obviously thinking about Oracle databases.)
However, it's entirely reasonable for a database to distinguish
between zero-length strings, and true nulls -- just as Perl does (''
is not the same as undef!)
Yes, this "feature" of Oracle annoys me. (It appears to annoy others
as well; I seem to remember that this is one of the reasons that
Oracle calls it VARCHAR2, and not simply VARCHAR...) I helped with a
migration from Informix to Oracle, and it originally relied on the
fact that you can store zero-length strings in VARCHAR columns
constrained to be NOT NULL.
t.
