On 12/1/05, Uwe C. Schroeder <uwe(at)oss4u(dot)com> wrote:
> Hi everyone,
> in 8.1 by default tables have no OID's anymore. Since OID's are 4 byte it's
> probably a good idea to discourage the use of them (they produced a lot of
> trouble in the past anyways, particularly with backup/restores etc)
> Now there's the issue with stored procs. A usual construct would be to
> INSERT xxxxxx;
> GET DIAGNOSTICS lastoid=RESULT_OID;
> SELECT .... oid=lastoid;
> Is there anything one could sanely replace this construct with?
> I personally don't think that using the full primary key is really a good
> option. Say you have a 3 column primary key - one being a "serial", the
> others for example being timestamps, one of them generated with "default"
> options. In order to retrieve the record I just inserted (where I don't know
> the "serial" value or the timestamp) I'd have to
> 1) store the "nextval" of the sequence into a variable
> 2) generate the timestamp and store it to a variable
> 3) generate the full insert statement and retain the other values of the
> primary key
> 4) issue a select to get the record.
> Personally I think this adds unneccessary overhead. IMHO this diminishes the
> use of defaults and sequences unless there is some easier way to retrieve the
> last record. I must be missing something here - am I ?
If you are using a SERIAL in your PK, why you need the other two
fields? The serial will undoubtly identify a record?
you just retrieve the current value you inserted with currval
(DBA: DataBase Aniquilator ;)
In response to
pgsql-hackers by date
|Next:||From: Magnus Hagander||Date: 2005-12-01 18:26:54|
|Subject: Re: [HACKERS] Upcoming PG re-releases|
|Previous:||From: Tom Lane||Date: 2005-12-01 18:20:24|
|Subject: Re: generalizing the planner knobs |
pgsql-general by date
|Next:||From: Tom Lane||Date: 2005-12-01 18:31:25|
|Subject: Re: undefined behaviour for sub-transactions? |
|Previous:||From: Michael Fuhr||Date: 2005-12-01 18:21:51|
|Subject: Re: undefined behaviour for sub-transactions?|