Re: [RFC] Removing "magic" oids

On 09/30/2018 05:48 AM, Andres Freund wrote:> I think we should drop
WITH OIDs support.

+1

> 2) We need to be able to manually insert into catalog tables. Both
> initdb and emergency surgery. My current hack is doing so directly
> in nodeModifyTable.c but that's beyond ugly. I think we should add
> an explicit DEFAULT clause to those columns with something like
> nextoid('tablename', 'name_of_index') or such.

Adding a function to get the next OID sounds like a good solution for
both our catalog and legacy applications. The only potential
disadvantage that I see is that this function becomes something we need
to maintain if we ever decide to move from OIDs to sequences.

> 3) The quickest way to deal with the bootstrap code was to just assign
> all oids for oid carrying tables that don't yet have any assigned.
> That doesn't generally seem terrible, although it's certainly badly
> implemented right now. That'd mean we'd have three ranges of oids
> probably, unless we somehow have the bootstrap code advance the
> in-database oid counter to a right state before we start with
> post-bootstrap work. I like the idea of all bootstrap time oids
> being determined by genbki.pl (I think that'll allow to remove some
> code too).

Agreed, having genbki.pl determine all oids in the bootstrap data sounds
ideal.

Andreas