Peter Eisentraut wrote:

>Alex Pilosov writes:
>
>>Dmitry's stuff contains both datatype (uniqueidentifier), a function to
>>generate a new object of that type (newid), and a set of functions to
>>implement comparison operators for that type.
>>
>>I don't see anything wrong with that setup, but maybe I'm still missing
>>something?
>>
>
>It would be much simpler if you stored the unique id in varchar or text.
>
Are you sure varchar comparision will be quickly than current 
implementation? Next, varchar will need 36 byte, uniqueidentifier takes 
16. Next, indexing - IMHO current stuff more suitable for indexes. Some 
time ago I saw some stuff which deals with uniqueidentifiers for 
Interbase. It uses your scheme with chars. But it strip "-" from string 
and reverts it to efficiently use indexes (uid sometimes uses 
MAC-address as part of itself, so MAC should go first in string). Weird 
scheme for me!

regards,
Dmitry

In response to

• Re: Re: New data type: uniqueidentifier at 2001-07-03 15:33:09 from Peter Eisentraut

pgsql-hackers by date

Next:	From: Alex Pilosov	Date: 2001-07-03 20:26:38
	Subject: Re: funny (cache (?)) bug in postgres (7.x tested)
Previous:	From: Nathan Myers	Date: 2001-07-03 20:17:40
	Subject: Re: Re: Backup and Recovery