[development] auto_increment

Karoly Negyesi karoly at negyesi.net
Mon Dec 12 21:26:34 UTC 2005


On Mon, 12 Dec 2005 22:35:11 +0100, Gerhard Killesreiter  
<gerhard at killesreiter.de> wrote:

> Nicolas Tostin wrote:
>
>>>> Btw, due to the chaotic use of sequences table (sometime it's used,
>>>> sometimes not)
>>>>
>>> When not?
>>>
>>>
>>
>> I don't have an exact list, but I remember having found some examples,  
>> so I
>> understood that it was better handling them by DB trigger (for Oracle  
>> schema
>> I mean) than doing nothing .
>>
>
> The general scheme is that we use locking if we want to know the ID in  
> advance or if we need it afterwards for further processing (example:  
> node iD).

There are facilities that return the last ID on a per connection basis for  
pgsql & mysql. As long as we do not use pconnect , we are fine. Those that  
use pconnect are advanced enough to introduce locking.


More information about the development mailing list