Re: Can we solve this -- NFNF and non-1NF at Loggerheads
From: Paul <paul_at_test.com>
Date: Mon, 07 Feb 2005 17:08:33 +0000
Message-ID: <4207a092$0$73319$ed2619ec_at_ptn-nntp-reader03.plus.net>
>
> imagine a catalog table with 2 columns :
> tablename and keycols (RVA ), both atrributes compound the key...
Date: Mon, 07 Feb 2005 17:08:33 +0000
Message-ID: <4207a092$0$73319$ed2619ec_at_ptn-nntp-reader03.plus.net>
strider5 wrote:
>>Maybe it's because I'm used to thinking relationally, but I just can't >>think of any examples where a RVA would be necessary or even just >> useful.
>
> imagine a catalog table with 2 columns :
> tablename and keycols (RVA ), both atrributes compound the key...
Why can't this be done in standard relations?
Surely you just have:
- one table (Tables?) with one row per tablename.
- another table (Keys?) with a primary key of {tablename, key_name} and maybe attributes to denote uniqueness etc.
- a third table (Key_Columns?) with a primary key of {tablename, key_name, col_name}
This is assuming you just want key names to be unique per table rather than database-wide, if not, remove the tablename column. Maybe I have some other details wrong, but you get the idea that it should be possible in principle.
Although this seems at first sight more complex because it uses more tables, I think it's better in the long run because it's forcing you to break down your knowledge into simple first-order predicates.
Paul. Received on Mon Feb 07 2005 - 18:08:33 CET