From: Luke O. <lu...@me...> - 2003-06-04 05:49:31
|
> Yes, it seems wierd the way I do it, but not really. What I'm trying to > do (using my Cat/Dog/Person example) is capture the relationship between > a person and all his/her animals within the same mapping table. There > are a variety of legitimate reasons for doing so. > > When you do this, a null value in one row need not necessarily indicate > invalid data. I do have a check constraint to ensure that each row has > at least pair of non-null values. Hmm, I was just going to reply to your other post that this merged intersection table seems odd. I have a few three-way intersections in my latest project, but not where either-this-or-that-or-both is the meaning (no NULLs here). Hmm. I've proposed that joins silently pass over NULL values before, but always changed my mind because it always seemed to point to bad data for us, and we didn't want to hide that. I suppose it could become a constructor argument for the join, allowNulls(?). But I'm only convinced by the legacy possibility of your table. :) Care to elaborate on the design reasons for sticking these both in one table instead of two? - Luke |