"Alex Bolenok" <[EMAIL PROTECTED]> writes:
> peroon=# INSERT INTO t_foo (foo_value) VALUES (2000);
> NOTICE:  fn_foo_ains: Start
> NOTICE:  fn_foo_ains: End
> NOTICE:  fn_bar_ains: Start
> NOTICE:  fn_bar_ains: End

Looking at the code, it seems that all AFTER triggers are implicitly
handled as DEFERRED triggers, ie, they're queued up and executed at
end of statement.  This seems wrong to me --- DEFERRED mode is useful,
certainly, but it shouldn't be the only form of AFTER trigger.

Also, it'd seem to me that DEFERRED mode ought to mean defer till end
of transaction, not just end of statement...

Jan, any comments here?

                        regards, tom lane

Reply via email to