From: | "David Clarke" <pigwin32(at)gmail(dot)com> |
---|---|
To: | "operationsengineer1(at)yahoo(dot)com" <operationsengineer1(at)yahoo(dot)com> |
Cc: | "Aaron Bono" <postgresql(at)aranya(dot)com>, pgsql-sql(at)postgresql(dot)org |
Subject: | Re: Alternative to serial primary key |
Date: | 2006-07-06 20:30:57 |
Message-ID: | 12b7ac1e0607061330o34adbd6ds9bd624d3889c7fe7@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-sql |
On 7/7/06, operationsengineer1(at)yahoo(dot)com <operationsengineer1(at)yahoo(dot)com> wrote:
> i agree. all my primary keys are abstract - even
> though some don't have to be. iow, i'm comfortable
> using serials as my primary key even when i don't
> absolutely need to.
Yes I had in fact already created my table using a serial as the
primary key but I've been reading Celko's SQL Programming Style and
the use of a hash on the address column as the primary key (and for
use in FK's) meets a number of the requirements for a good key. The
address column itself is the natural primary key but it doesn't make
for a good FK. Plus I feel I would be remiss in not exploring an
alternative to the serial key.
To recap, yes there is only a single column, yes it is varchar. I need
to do a lookup on the address column which is unique and use it as a
foreign key in other tables. Using a serial id would obviously work
and has been recommended. But having a hash function over the address
column as the primary key means I can always regenerate my primary key
from the data which is impossible with a serial key. I believe the
risk of collision using md5 is effectively zero on this data and I can
put a unique index over it.
I'm kind of new to sql so apologies if this is a naive approach.
Thanks to all for responses.
Dave
From | Date | Subject | |
---|---|---|---|
Next Message | D'Arcy J.M. Cain | 2006-07-06 21:18:16 | Re: Alternative to serial primary key |
Previous Message | Aaron Bono | 2006-07-06 19:32:45 | Re: Foreign Key: what value? |