Re: pg_sequence catalog

On 11/11/16 10:06 PM, Andreas Karlsson wrote:
> As pointed out by Peter this patch also requires the changes to
> pg_upgrade. I have not looked at those patches.

The required changes to pg_upgrade have been committed, so that will
work now.

> - The pg_dump tests fails due to the pg_dump code not being updated. I
> have attached a patch which fixes this.

fixed

> I was a bit worried that the extra syscache lookups might slow down
> nextval(), but I got a measurable speed up on a very simple workload
> which consisted of only calls to nextval() to one sequence. The speedup
> was about 10% on my machine.

I have done a fair amount of performance testing and the results were
usually in the neighborhood of 1%-2% slower or faster, nothing really
clear.  But running nextval by itself in a tight loop isn't really a
normal use.  The important thing is the concurrency behavior.

>  > @@ -1155,6 +1156,8 @@ doDeletion(const ObjectAddress *object, int flags)
>  >                     else
>  >                         heap_drop_with_catalog(object->objectId);
>  >                 }
>  > +               if (relKind == RELKIND_SEQUENCE)
>  > +                   DeleteSequenceTuple(object->objectId);
>  >                 break;
>  >             }
>
> I think it might be cleaner here to put this as a "else if" just like
> "relKind == RELKIND_INDEX".

The sequence tuple has to be deleted in addition to the heap drop.  I
have added a comment to make the clearer.

> The patch does not update catalogs.sgml which it should do.

added

--
Peter Eisentraut              http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services