sam wrote:
> FUNCTION UPDATE ()
> BEGIN
> UPDATE statement
>
> EXCEPTION
> END
>
>
> when one record fails data only for that is rolled back the rest of
> the data is saved. EXCEPTION has to be caught.
>
>
As I recently found out, too many savepoints really kill PostgreSQL's
performance in a transaction. A function inserting/updating, say,
100,000 records will perform OK, but statements run in the same
transaction after the function completes will be very slow.
So ... if you find that after your big update statement performance is
terrible, you might need to explicitly check the conditions that might
result in an exception and skip those records, thus avoiding the
EXCEPTION block.
--
Craig Ringer