Re: Inheriting table AMs for partitioned tables

On Tue, Mar 5, 2019 at 5:17 AM Andres Freund <andres(at)anarazel(dot)de> wrote:
>
> Hi,
>
> In the pluggable storage patch [1], one thing that I'm wondering about
> is how exactly to inherit the storage AM across partitions. I think
> that's potentially worthy of a discussion with a wider audience than I'd
> get in that thread. It seems also related to the recent discussion in [2]
>
> Consider (excerpted from the tests):
>
> CREATE TABLE tableam_parted_heap2 (a text, b int) PARTITION BY list (a) USING heap2;
>
> SET default_table_access_method = 'heap';
> CREATE TABLE tableam_parted_a_heap2 PARTITION OF tableam_parted_heap2 FOR VALUES IN ('a');
>
> SET default_table_access_method = 'heap2';
> CREATE TABLE tableam_parted_b_heap2 PARTITION OF tableam_parted_heap2 FOR VALUES IN ('b');
>
> CREATE TABLE tableam_parted_c_heap2 PARTITION OF tableam_parted_heap2 FOR VALUES IN ('c') USING heap;
> CREATE TABLE tableam_parted_d_heap2 PARTITION OF tableam_parted_heap2 FOR VALUES IN ('d') USING heap2;
>
> It seems pretty clear that tableam_parted_heap2, tableam_parted_d_heap2
> would be stored via heap2, and tableam_parted_c_heap2 via heap.
>
> But for tableam_parted_a_heap2 tableam_parted_b_heap2 the answer isn't
> quite as clear. I think it'd both be sensible for new partitions to
> inherit the AM from the root, but it'd also be sensible to use the
> current default.
>

Yeah, we can go either way.

> Out of laziness (it's how it works rn) I'm inclined to to go with using
> the current default, but I'd be curious if others disagree.
>

I think using the current default should be okay as that will be the
behavior for non-partitioned tables as well. However, if people have
good reasons to go other way, then that is fine too.

--
With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com