[Rt-devel] Postgres performance issues and some "odd" ways to get IDs...

[Joby Walker]

Andrew Sullivan wrote:
> On Mon, Aug 14, 2006 at 10:03:36AM -0700, Joby Walker wrote:
>> Jesse,
>>
>> The new behavior seems very broken:
>>
>> 1) Insert row (sequence nextval returns x)
>> 2) get current value of sequence (returns x)
>>
>> As far as I can see SearchBuilder doesn't wrap a transaction around the 
>> Insert sub so you could end up with unpredictable results:
> 
> No, you can't, even without explicit BEGIN;  . . . COMMIT;  Please
> see the Postgres docs on nextval() an currval().  But the basic thing
> is that currval() is private to _your connection_, not to your
> transaction.
> 
> A
> 


Your absolutely right.  Thanks for correcting me!  Looks like currval is 
safer than I thought...