Received-SPF: none (enugu.terra.com.br: 200.176.10.9 is neither permitted nor
	denied by domain of carcass.dhs.org) client-ip=200.176.10.9;
	envelope-from=howe@carcass.dhs.org; helo=[10.0.0.3];
Message-ID: <4395E68E.3000707@carcass.dhs.org>
Date: Tue, 06 Dec 2005 17:29:18 -0200
From: Steve Howe <howe@carcass.dhs.org>
User-Agent: Mozilla Thunderbird 1.0.7 (X11/20051013)
MIME-Version: 1.0
Cc: pgsql-interfaces@postgresql.org, pgsql-odbc@postgresql.org
Subject: Re: [ODBC] BLOB handling compatibility with PostgreSQL
References: <438B792D.6030802@bnl.gov>
	<20051205212806.GA17690@soptik.pzkagis.cz>
	<20051206142811.GA23334@soptik.pzkagis.cz>
	<259.1133881501@sss.pgh.pa.us>
	<20051206184402.GA24475@soptik.pzkagis.cz>
	<15186.1133896033@sss.pgh.pa.us>
In-Reply-To: <15186.1133896033@sss.pgh.pa.us>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Tom Lane wrote:

>Ludek Finstrle <luf@pzkagis.cz> writes:
>  
>
>>I don't want to reinvent the wheel.
>>    
>>
>
>Why do you feel a need to distinguish the domain from its underlying
>type on the client side?  They're the same as regards representation
>and so on.
>
>The reason for the backend change was that client-side drivers (such as
>JDBC and ODBC) want to know the underlying datatype so that they know
>what representation to use etc.  Distinguishing domains made their job
>harder not easier.
>  
>
>If you want an add-on datatype that is really different from OID, then
>make a real datatype (CREATE TYPE).  You can still piggyback on OID as
>the representation type --- steal its I/O functions and so on.
>  
>
A clear example situation in here is how to be able to distinguish a 
large object field. The ODBC driver, for instance, uses the 'lo' type, 
which is the same as an oid (Large Object).

I ran into exactly the same situation as I wrote the pgExpress driver 
for Vita Voom Software: while declaring a domain looks like the clear 
choice, the original type's oid would be returned by the pq_ftype() 
function. So I based my solution more or less like this post by Hiroshi 
Saito:

http://www.mail-archive.com/pgadmin-hackers@postgresql.org/msg01390.html

... and asked the users to create the 'lo' type that way, which would 
create a real type, just like Tom suggests.

For more details on how it was implemented on the pgExpress, please 
check kits documentation:
http://www.vitavoom.com/Products/pgExpress_Driver/docs/advanced_features.html#large_objects_declaration

Best regards,
Steve Howe