Re: Support logical replication of DDLs

On Tue, Jun 13, 2023 at 1:21 PM Michael Paquier <michael(at)paquier(dot)xyz> wrote:
>
> On Mon, Jun 12, 2023 at 01:47:02AM +0000, Wei Wang (Fujitsu) wrote:
> > # load test cases from the regression tests
> > -my @regress_tests = split /\s+/, $ENV{REGRESS};
> > +#my @regress_tests = split /\s+/, $ENV{REGRESS};
> > +my @regress_tests = ("create_type", "create_schema", "create_rule", "create_index");
> > ```
> > I think @regress_tests should include all SQL files, instead of just four. BTW,
> > the old way (using `split /\s+/, $ENV{REGRESS}`) doesn't work in meson.
>
> I have been studying what is happening on this thread, and got a look
> at the full patch set posted on 2023-06-08.
>
> First, the structure of the patch set does not really help much in
> understanding what would be the final structure aimed for. I mean, I
> understand that the goal is to transform the DDL Nodes at a given
> point in time, fetched from the event query code paths, into a
> structure on top of which we want to be apply to apply easily
> filtering rules because there could be schema changes or even more..
> But, take patch 0001. It introduces ObjTree to handle the
> transformation state from the DDL nodes, and gets replaced by jsonb
> later on in 0008. This needs to be reworked and presented in a shape
> that's suited for review, split into more parts so as this is
> manageable.
>

We have to choose one of the approaches between 0001 and 0008. I feel
we don't need an intermediate ObjTree representation as that adds
overhead and an additional layer which is not required. As mentioned
in my previous email I think as a first step we should merge 0001 and
0008 and avoid having an additional ObjTree layer unless you or others
feel we need it. I think that will reduce the overall patch size and
help us to focus on one of the approaches. Surely, as suggested by
you, we should also evaluate if we can generate this code for the
various command types.

--
With Regards,
Amit Kapila.