Re: multi-column range partition constraint

On Sat, May 13, 2017 at 12:42 AM, Amit Langote <amitlangote09(at)gmail(dot)com> wrote:
> Attached is the correct version.

Thank you! I committed 0001 with a couple of cosmetic tweaks and with
the change I previously suggested around partexprs_item. You argued
that wouldn't work because the contents of partexprs_item was
modified, but that's not so: partexprs_item in
get_range_key_properties is a pointer the partexprs_item in the
caller. When it modifies *partexprs_item, it's not changing the
contents of the ListCell itself, just the caller's ListCell *. I also
ran pgindent over the patch.

Also committed 0002. In that case, I removed CHECK (...) from the
output; the caller can always add that if it's desired, but since a
partitioning constraint is NOT a CHECK constraint I don't actually
think it's useful in most cases. I also tweaked the regression tests
slightly.

Thanks again.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company