Re: Joins with nulls

"David Cressey" <david_at_dcressey.com> wrote in message news:<XT3F9.297$0I3.31567_at_petpeeve.ziplink.net>...
> All of which is quite beside the point about nulls. As I said before, it's
> possible to design a relational schema which accommodates your needs
> including partial information without recourse to NULLS. Whether that
> should be done, or whether one should just use NULLS, is another matter.

I guess the question is whether you want to have facts about "unknown" information denoted by the syntax (using NULL) or by the semantics (using flag values, empty strings etc.).

I suppose the advantage of using the syntax method is that the database itself knows what it means. i.e. you can search for information that is unknown without knowing anything about the semantics.

With the semantics method it's up to the database owner to know the interpretation applied to the data. The advantages being flexibility (define your own different kinds of unknown data) and the fact that all the other "known" data is stored semantically.

Maybe it's a bit like Holme's dog that didn't bark in the night - the absence of a fact is itself a fact. i.e. saying Bill is age unknown is just as valid information as saying Bill is age 20.

You've still got the problem of OUTER JOINs bringing in NULLs though but like another poster said I'm sure you could get round it, in effect by specifying an actual value to use in place of the NULLs.

Paul. Received on Thu Nov 28 2002 - 11:39:23 CET