On Tue, Jul 03, 2018 at 11:16:55PM +1200, David Rowley wrote:
> Oh okay. Yeah, you can hit that with a partitionless sub-partitioned
> table.
Thanks for the patch and fixing the typo ;)
+create table list_parted_tbl (a int,b int) partition by list (a);
+create table list_parted_tbl1 partition of list_parted_tbl for values
in(1) partition by list(b);
+select * from list_parted_tbl;
+explain (costs off) select * from list_parted_tbl;
I am not sure if it is much interesting to keep around this table set
for pg_upgrade, so I would drop it. Except for that, the result looks
fine. I'll double-check and wrap it tomorrow on HEAD and REL_11_STABLE.
The optimizations mentioned sound interesting, though I would recommend
to not risk the stability of v11 at this point, so let's keep them for
v12~.
--
Michael