Thread: How to best grab a chunk of Ids from a sequence

I have a process that will be creating a large number of records (about
4Million).  I am thinking of making this process run quicker by writing the
records to a file in 'dump' format and then running that file through psql.

The problem is each record has a sequence number and I need to know the
sequence ids for other
purposes in this process.

My question is is there a way to grab 4 million IDs from a sequence without
calling nextval once for each ID.  Note, this sequence is being actively
drawn on by other processes.

---------
Bryan White

"Bryan White" <bryan@arcamax.com> writes:
> My question is is there a way to grab 4 million IDs from a sequence without
> calling nextval once for each ID.  Note, this sequence is being actively
> drawn on by other processes.

I can't think of any really clean way, but you could do something like
select setval('seq', nextval('seq') + 4000000 + 100);

and then use the 4m IDs just before the returned value.  Because the
nextval and setval aren't an atomic operation, it's possible that
some other processes will get in and do nextval in between.  The extra
100 advance of the sequence should provide an adequate buffer for that
... but of course it could theoretically be insufficient...
        regards, tom lane

What you could do is use the setval() function to set the value of the
sequence to whatever it is now + 4 million.  That would give you a
great big hole in your sequence from which you could draw sequence
values from.  Your processes that still needed to use the sequence
would still be able to, and the information that you are importing
could use the sequence numbers that you skipped.

Jason

"Bryan White" <bryan@arcamax.com> writes:

> I have a process that will be creating a large number of records (about
> 4Million).  I am thinking of making this process run quicker by writing the
> records to a file in 'dump' format and then running that file through psql.
> 
> The problem is each record has a sequence number and I need to know the
> sequence ids for other
> purposes in this process.
> 
> My question is is there a way to grab 4 million IDs from a sequence without
> calling nextval once for each ID.  Note, this sequence is being actively
> drawn on by other processes.
> 
> ---------
> Bryan White
> 
> 
> 
> ---------------------------(end of broadcast)---------------------------
> TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org