Date: Wed, 3 Mar 2010 17:17:59 +0100
From: Louis-David Mitterrand <vindex+lists-pgsql-sql@apartia.org>
To: pgsql-sql@postgresql.org
Subject: Re: check constraint on multiple tables?
Message-ID: <20100303161759.GB22889@apartia.fr>
Mail-Followup-To: pgsql-sql@postgresql.org
References: <20100303150227.GA18661@apartia.fr>
	<396486431003030714s48b239e2ub8b58a49b9aa29e4@mail.gmail.com>
	<20100303151900.GA19735@apartia.fr>
	<396486431003030729x17f921bcr7ff439113a5644ab@mail.gmail.com>
	<20100303153831.GA20503@apartia.fr>
	<7B7DB40901C947B688FD3BAC874E762F@marktestcr.marktest.pt>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <7B7DB40901C947B688FD3BAC874E762F@marktestcr.marktest.pt>
User-Agent: Mutt/1.5.20 (2009-06-14)

On Wed, Mar 03, 2010 at 04:05:43PM -0000, Oliveiros wrote:
> 
> As your table names seem to imply, type and category are cabin's
> characteristics , not ship characteristics.
> Am I right?

Yes and no :) 

- I'm scanning cruise line web sites with a spider to collect prices so
  I'm building cabin_type's and cabin_category's on the fly,
- each ship class (a cruise line has several) has its own particular
  types (Celebrity Xpedition is the only one with "Xpedition suite"
  type, etc.)

> As Richard pointed out, maybe you could add a relationship between
> cabin and ship and drop the relationship between ship and
> cabin_category you now have
> Then you could add that uniqueness restriction.

That's one option.

> Also, the relationship between type and category is one to many ? Or
> can it be many to many? Put other way, is this overlap between the
> categories that belong to different "types" ?

One cabin_type to many cabin_category's, for example:

- "Sunset Veranda Stateroom" (type) can be on "Vista", "Panorama", etc.
  decks (category) with a different price,

But it's true that there is some overlap in categories between different
ships.

> If the later applies, maybe
> you could have cabin refer to both type and category tables and drop
> the relation between type and category.
> 
> The cabin table would then work as an associative table between
> category and type.
> 
> Ain't saying your schema is wrong, maybe you have strong reasons to
> do that that way, that I am not realizin by now...

You got me thinking about it. Thank you for your interesting comments.