Re: change oid of a pg_type - Mailing list pgsql-hackers

From Andrew Chernow
Subject Re: change oid of a pg_type
Date
Msg-id 49E112A8.8000306@esilo.com
Whole thread Raw
In response to Re: change oid of a pg_type  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
Tom Lane wrote:
> Andrew Chernow <ac@esilo.com> writes:
>> Tom Lane wrote:
>>> Merlin Moncure <mmoncure@gmail.com> writes:
>>>> Is there any way to adjust the oid of an already created pg_type
>>> No.
> 
>> Then why are oids sent in some type's external binary format?
> 
> How is that related?
> 

Well, the format is called "external", yet the oid is internal.  It seem 
like a contradiction.  Maybe I'm being to literal but i think external 
format should present an identifier other servers can understand/map.

>> If you 
>> can't fix oids, external format is useless to the outside world.
> 
> You have to be prepared to find out what type the OID belongs to.
> There's at least one client library (maybe JDBC) that maintains
> a cache of a SELECT from pg_type for exactly this purpose.  Most
> 

Yeah, libpqtypes looks them up by schema.name as one registers them, 
PQregisterTypeHandler.  Although this works, it feel more like a work 
around.  I think it would be cleaner if a dns-like resolution was not 
required.

Andrew







pgsql-hackers by date:

Previous
From: Tom Lane
Date:
Subject: Re: change oid of a pg_type
Next
From: Abhijit Menon-Sen
Date:
Subject: Re: Allow COMMENT ON to accept an expression rather than just a string