Re: Support logical replication of DDLs

On Thu, Jun 1, 2023 at 1:10 PM Yu Shi (Fujitsu) <shiy(dot)fnst(at)fujitsu(dot)com> wrote:
>
> On Wed, May 31, 2023 5:41 PM shveta malik <shveta(dot)malik(at)gmail(dot)com> wrote:
> >
> > On Mon, May 29, 2023 at 11:45 AM Yu Shi (Fujitsu) <shiy(dot)fnst(at)fujitsu(dot)com>
> > wrote:
> > >
> > > 0008 patch
> > > -----
> > > 4.
> > > case AT_AddColumn:
> > > /* XXX need to set the "recurse" bit somewhere? */
> > > Assert(IsA(subcmd->def, ColumnDef));
> > > - tree = deparse_ColumnDef(rel, dpcontext, false,
> > > - (ColumnDef *) subcmd->def, true,
> > &expr);
> > >
> > > mark_function_volatile(context, expr);
> > >
> > > After this change, `expr` is not assigned a value when mark_function_volatile is
> > called.
> > >
> >
> > Corrected.
> >
>
> It looks the call to mark_function_volatile() is removed in this case. I think
> we still need it, otherwise we won't know if the command contains a volatile
> function. (see check_command_publishable().)
>

Modified this. Thanks for pointing this out.

> > > 5.
> > > create table p1(f1 int);
> > > create table p1_c1() inherits(p1);
> > > alter table p1 add constraint inh_check_constraint1 check (f1 > 0);
> > > alter table p1_c1 add constraint inh_check_constraint1 check (f1 > 0);
> > >
> > > The re-formed command of the last command is "ALTER TABLE public.p1_c1",
> > which
> > > seems to be wrong.
> > >
> >
> > Fixed, second alter-table should actually be no-op in terms of
> > deparsing. But when it is run without running the first alter-table
> > command, it should generate the reformed command.
> >
>
> If the second alter-table is no-op in terms of deparsing, the dumped result of
> origin command and that of re-formed command will be different. This seems to be
> because `conislocal` column of pg_constraint has different values. (After the
> second alter-table, its value is changed from false to true.)
>

Okay, I see. I am analyzing it to figure out if we really need the
second call to be replicated on sub. It needs more review around the
impact of 'conislocal'. I will respond once concluded.

Please find the new set of patches attached. The changes are in 0008
only. New changes are majorly for:
--comments given by Shi-san. Thanks Vignesh for working on this.
--addition of more test-cases to cover missing scenarios. Thanks
Shi-san for working on this.

thanks
Shveta