On Fri, Aug 6, 2021 at 2:02 PM vignesh C <vignes...@gmail.com> wrote:
>
> On Wed, Aug 4, 2021 at 4:10 PM Amit Kapila <amit.kapil...@gmail.com> wrote:
> >
> > On Tue, Aug 3, 2021 at 8:38 PM vignesh C <vignes...@gmail.com> wrote:
>
> > 6.
> > + {PublicationSchemaRelationId, /* PUBLICATIONSCHEMAMAP */
> > + PublicationSchemaPsnspcidPspubidIndexId,
> > + 2,
> > + {
> > + Anum_pg_publication_schema_psnspcid,
> > + Anum_pg_publication_schema_pspubid,
> > + 0,
> > + 0
> > + },
> >
> > Why don't we keep pubid as the first column in this index?
>
> I wanted to keep it similar to PUBLICATIONRELMAP, should we keep it as
> it is, thoughts?
>Okay, I see your point. I think for PUBLICATIONRELMAP, we need it because it is searched using the only relid in GetRelationPublications, so, similarly, in the patch, you are using schema_oid in GetSchemaPublications, so probably that will help. I was wondering why you haven't directly used the cache in GetSchemaPublications similar to GetRelationPublications? It seems there is a danger for concurrent object drop. Can you please check how the safety is ensured say when either one wants to drop the corresponding relation/schema or publication? Another point is why can't we use the other index (where the index is on relid or schema_oid (PublicationSchemaObjectIndexId))? -- With Regards, Amit Kapila.
