Re: Feature: triggers on materialized views

Hi!

On Fri, Mar 15, 2019 at 2:46 AM David Steele <david(at)pgmasters(dot)net> wrote:
> The reason is that you have not gotten any committer support for this
> patch or attracted significant review, that I can see. On the contrary,
> three committers have expressed doubts about all or some of the patch
> and it doesn't seem to me that their issues have been addressed.

To my understanding many comments were to early version of the patch
which were or addressed or explained why proposed changes not work
(use TRUNCATE/INSERT instead of swapping heads is slower). If I missed
anything and have not addressed it, please point this out.

The only pending/unaddressed comment is about the philosophical
question of what it means to be a trigger. There it seems we simply
disagree with the reviewer and I do not know how to address that. I
just see this as a very pragmatical feature which provides features
you would have if you would not use PostgreSQL abstraction. If you can
use features then, why not also with the abstraction?

So in my view this looks more like a lack of review feedback on the
latest version of the patch. I really do not know how to ask for more
feedback or to move the philosophical discussion further. I thought
that commit fest is in fact exactly a place to motivate and collect
such feedback instead of waiting for it in the limbo.

> What you need to be doing for PG13 is very specifically addressing
> committer concerns and gathering a consensus that the behavior of this
> patch is the way to go.

To my understanding the current patch addresses all concerns made by
reviewers on older versions of the patch, or explains why proposals
cannot work out, modulo the question of "does this change what trigger
is".

Moreover, it improves performance of CONCURRENT REFRESH for about 5%
based on my tests because of the split to INSERT/UPDATE/DELETE instead
of TRUNCATE/INSERT, when measuring across a mixed set of queries which
include just UPDATEs to source tables.

Thank you everyone who is involved in this process for your time, I do
appreciate. I am just trying to explain that I am a bit at loss on
concrete next steps I could take here.

Mitar

--
http://mitar.tnode.com/
https://twitter.com/mitar_m