Re: row filtering for logical replication

On Fri, Nov 26, 2021 at 4:05 PM houzj(dot)fnst(at)fujitsu(dot)com
<houzj(dot)fnst(at)fujitsu(dot)com> wrote:
>
> On Fri, Nov 26, 2021 11:32 AM Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
> > On Thu, Nov 25, 2021 at 7:39 PM Euler Taveira <euler(at)eulerto(dot)com> wrote:
> > >
> > > On Thu, Nov 25, 2021, at 10:39 AM, houzj(dot)fnst(at)fujitsu(dot)com wrote:
> > >
> > > When researching and writing a top-up patch about this.
> > > I found a possible issue which I'd like to confirm first.
> > >
> > > It's possible the table is published in two publications A and B,
> > > publication A only publish "insert" , publication B publish "update".
> > > When UPDATE, both row filter in A and B will be executed. Is this behavior
> > expected?
> > >
> > > Good question. No. The code should check the action before combining
> > > the multiple row filters.
> > >
> >
> > Do you mean to say that we should give an error on Update/Delete if any of the
> > publications contain table rowfilter that has columns that are not part of the
> > primary key or replica identity? I think this is what Hou-san has implemented in
> > his top-up patch and I also think this is the right behavior.
>
> Yes, the top-up patch will give an error if the columns in row filter are not part of
> replica identity when UPDATE and DELETE.
>
> But the point I want to confirm is that:
>
> ---
> create publication A for table tbl1 where (b<2) with(publish='insert');
> create publication B for table tbl1 where (a>1) with(publish='update');
> ---
>
> When UPDATE on the table 'tbl1', is it correct to combine and execute both of
> the row filter in A(b<2) and B(a>1) ?(it's the current behavior)
>
> Because the filter in A has an unlogged column(b) and the publication A only
> publish "insert", so for UPDATE, should we skip the row filter in A and only
> execute the row filter in B ?
>

But since the filters are OR'ed together does it even matter?

Now that your top-up patch now prevents invalid updates/deletes, this
other point is only really a question about the cache performance,
isn't it?

------
Kind Regards,
Peter Smith.
Fujitsu Australia.