pg.ddx.io  pgsql-sql@postgresql.org mailing list archive  
help / color / mirror / Atom feed
From: Josh Berkus <josh@agliodbs.com>
To: Cedar Cox <cedarc@visionforisrael.com>
Cc: pgsql-sql@postgresql.org
Subject: Re: REFERENCES constraint
Date: Sun, 12 Aug 2001 19:06:20 -0700
Message-ID: <web-100993@davinci.ethosmedia.com> (raw)
In-Reply-To: <Pine.LNX.4.21.0108122137590.11622-100000@nanu.visionforisrael.com>

Cedar,

> I guess I wasn't clear.  Let me try to explain again:
> 
>   CREATE TABLE obj_weights (
>     object_id int4 REFERENCES ( apple_objects(id) OR
> banana_objects(id) )
>     weight float4,
>   )
> 
> "apple_objects" doesn't necessarily have anything to do with
> "banana_objects".  Ok, don't ask why you would want to store weights
> of
> apples and bananas in the same table.. (and if you know, please tell
> me).  
> This is all actually for someone else's database that I just picked
> up.  
> They did something like this.. single sequence for the whole
> database,
> multiple object tables, and a table(s) referencing objects that could
> come
> from any of those tables.  Maybe this is just bad design.  Thoughts
> anyone?

That actually makes a *lot* more sense.  In fact, I'm doing the same
thing with two tables in my database ... one called "mod_data" and
another one called "notes".  The first stores modifications users and
timestamps for 6 different tables, the second stores scrolling user
notes for foure different tables.  Thus mod_data should look like:

CREATE TABLE mod_data (
	usq INT NOT NULL REFERENCES (clients(usq) OR candidates(usq) or
orders(usq) or ... )
	entry_user INT NOT NULL
	entry_date TIMESTAMP NOT NULL DEFAULT current_timestamp
	mod_user INT NOT NULL
	mod_date TIMESTAMP NOT NULL DEFAULT current_timestamp
)

But, as you've observed, this is quite impossible.  The entire structure
of relationality has been constructed around the heirarchy and/or star
topology; there is no provision for this kind of relationship.  I can
think of a number of good reasons not to attempt to implement REFERENCES
for this kind of structure.

So you can deal with it as I did:

1. Users have no ability to add or modify records in the child-of-many
table.  They must push data through functions that I have defined, which
take care of creating/updating the dependant records.
2. All major tables subscribe to a single "Univeral Sequence" that
supplies unique surrogate keys for the six tables.  This makes the ID
unique not only within but between the 6 tables.  The one disadvantage
is that it is a postgresql-only implementation.
3. Users cannot modify this "usq".  Nor can they delete records.

Thus my DB integrity is protected.  In a database where user access is
less restrained, you may find that the cost of creating all the triggers
necessary to deal with user updates and deletes is a lot more than the
effort to duplicate a few tables.

An alternate approach is for you to define your own updatable views.
While quite labor-intensive, this approach takes care of quite a few
complex relationship structures by forcing the users to push their
inserts and updates through the views.  Properly designed, the user
would not even realize that banana_weight and apple_weight are in the
same table and are in a different table from banana_data and apple_data.

Or, perhaps, as a very advanced user, I'm just making some very advanced
mistakes ...

-Josh Berkus


______AGLIO DATABASE SOLUTIONS___________________________
                                       Josh Berkus
  Complete information technology      josh@agliodbs.com
   and data management solutions       (415) 565-7293
  for law firms, small businesses        fax 621-2533
    and non-profit organizations.      San Francisco







view thread (10+ messages)  latest in thread

Message-ID: <web-100993@davinci.ethosmedia.com>
Permalink:  ../web-100993@davinci.ethosmedia.com/
Also on:    postgresql.org/message-id/web-100993@davinci.ethosmedia.com

 · 

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: pgsql-sql@postgresql.org
  Cc: josh@agliodbs.com, cedarc@visionforisrael.com
  Subject: Re: REFERENCES constraint
  In-Reply-To: <web-100993@davinci.ethosmedia.com>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox