Re: SQL:2011 application time

On Thu, Jul 6, 2023 at 1:13 AM Peter Eisentraut
<peter(dot)eisentraut(at)enterprisedb(dot)com> wrote:
>
> I had talked to Paul about this offline a while ago. btree_gist to core
> is no longer considered a prerequisite. But Paul was planning to
> produce a new patch set that is arranged and sequenced a bit
> differently. Apparently, that new version is not done yet, so it would
> make sense to either close this entry as returned with feedback, or move
> it to the next commit fest as waiting on author.

Here are some new patch files based on discussions from PGCon. The
patches are reorganized a bit to hopefully make them easier to review:

Initially I implement all functionality on just range columns, without
supporting PERIODs yet. There are patches for temporal PRIMARY
KEY/UNIQUE constraints, for simple foreign keys (without CASCADE/SET
NULL/SET DEFAULT), for UPDATE/DELETE FOR PORTION OF, and then for the
rest of the FK support (which depends on FOR PORTION OF). If you
compare these patches to the v11 ones, you'll see that a ton of
clutter disappears by not supporting PERIODs as a separate "thing".

Finally there is a patch adding PERIOD syntax, but with a new
implementation where a PERIOD causes us to just define a GENERATED
range column. That means we can support all the same things as before
but without adding the clutter. This patch isn't quite working yet
(especially ALTER TABLE), but I thought I'd send where I'm at so far,
since it sounds like folks are interested in doing a review. Also it
was a little tricky dealing with the dependency between the PERIOD and
the GENERATED column. (See the comments in the patch.) If anyone has a
suggestion there I'd be happy to hear it.

My goal is to include another patch soon to support hidden columns, so
that the period's GENERATED column can be hidden. I read the
conversation about a recent patch attempt for something similar, and I
think I can use most of that (but cut some of the things the community
was worried about).

All these patches need some polishing, but I think there is enough new
here for them to be worth reading for anyone interested in temporal
progress.

I'll set this commitfest entry back to Needs Review. Thanks for taking a look!

Paul