Re: Sequence Access Method WIP

On 29 January 2016 at 14:48, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Alvaro Herrera <alvherre(at)2ndquadrant(dot)com> writes:
>> I would guess that the DDL boilterplate should come from Alexander
>> Korotkov's patch, right? I think a first easy step may be to combine
>> parts both patches so that we get the "amkind" column from this patch
>> and the DDL support from Alexander's patch (means that his proposed
>> command needs a new clause to specify the amkind);
>
> Uh, what? Surely we would provide a bespoke command for each possible
> sort of handler. As an example, CREATE INDEX ACCESS METHOD ought to check
> that the provided function has the right signature, and then it would put
> the correct amkind into the pg_am entry automatically.
>
> If we skimp on this infrastructure then we're just relying on the
> user not to make a typo, which doesn't seem like a good idea.
>

Yeah it has to be completely separate command for both that do
completely different sanity checks. It can't even be called CREATE
INDEX/SEQUENCE ACCESS METHOD unless we are willing to make ACCESS a
keyword due to preexisting CREATE INDEX/SEQUENCE <name> commands. The
previous version of the patch which used somewhat different underlying
catalog structure already had DDL and honestly writing the DDL part is
quite easy. I used CREATE ACCESS METHOD FOR SEQUENCE there.

> (Agreed though that a reasonable first step would be to add amkind to
> pg_am and make the appropriate adjustments to existing code, ie check
> that amkind is correct when fetching an index handler. I considered
> putting that into the AM API refactor patch, but desisted.)
>

Well I was wondering how to handle this as well, the submitted patch
currently just adds Asserts, because IMHO the actual ERROR should be
thrown in the DDL not in the code that just uses it.

--
Petr Jelinek http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services