Re: Support logical replication of DDLs

On Mon, Apr 24, 2023 at 8:18 AM shveta malik <shveta.malik@gmail.com> wrote:
>
> On Thu, Apr 20, 2023 at 3:40 PM Amit Kapila <amit.kapila16@gmail.com> wrote:
> >
> > On Thu, Apr 20, 2023 at 9:11 AM shveta malik <shveta.malik@gmail.com> wrote:
> > >
> > >
> > > Few comments for ddl_deparse.c in patch dated April17:
> >
> > >
> > > 6) There are plenty of places where we use 'append_not_present'
> > > without using 'append_null_object'.
> > > Do we need to have 'append_null_object' along with
> > > 'append_not_present' at these places?
> > >
> >
> > What is the difference if we add a null object or not before not_present?
> >
>
> Sorry I missed this question earlier. There is not much difference
> execution wise, The "present": false' attribute is sufficient to
> indicate that the expansion of that element is not needed when we
> convert back json to ddl command. It is only needed for 'verbose'
> mode. The 'append_null_object' call makes the format string complete
> for the user to understand it better.  Example:
>
> --without append_null_object, we get:
> "collation": {"fmt": "COLLATE", "present": false}
>
> --with  append_null_object, we get:
> With append_null_object(tmp_obj, "%{name}D"), it is:
> "collation": {"fmt": "COLLATE %{name}D", "name": null, "present": false}
>
> So fmt:  "COLLATE %{name}D" is more complete (even though not needed
> when the COLLATE clause is absent) and thus aligns to what we expect
> out of verbose mode.
>

Thanks for the explanation. However, I feel we can split this verbose
handling/optimization into a separate patch so that we can focus on
the core part first.

--
With Regards,
Amit Kapila.