Re: Sequence Access Methods, round two

On Mon, Aug 18, 2025 at 06:15:05PM +0500, Kirill Reshke wrote:
> Yes, this resonates with me better.

After sleeping on it, yes, perhaps that's the right move.
test_ddl_deparse reports the attributes as related to a sequence, but
it's really how this is handled internally anyway with the internal
call of DefineRelation(). This entirely decouples the relation
creation and the creation of its attributes.

> @@ -3365,7 +3365,7 @@ match_previous_words(int pattern_id,
> COMPLETE_WITH("TYPE");
> /* Complete "CREATE ACCESS METHOD <name> TYPE" */
> else if (Matches("CREATE", "ACCESS", "METHOD", MatchAny, "TYPE"))
> - COMPLETE_WITH("INDEX", "TABLE");
> + COMPLETE_WITH("INDEX", "TABLE", "SEQUENCE");
> /* Complete "CREATE ACCESS METHOD <name> TYPE <type>" */
> else if (Matches("CREATE", "ACCESS", "METHOD", MatchAny,
> "TYPE", MatchAny))
> COMPLETE_WITH("HANDLER");

Right, thanks.
--
Michael