Re: partition routing layering in nodeModifyTable.c

On Thu, Oct 8, 2020 at 9:35 PM Amit Langote <amitlangote09(at)gmail(dot)com> wrote:
> On Wed, Oct 7, 2020 at 9:07 PM Heikki Linnakangas <hlinnaka(at)iki(dot)fi> wrote:
> > On 07/10/2020 12:50, Amit Langote wrote:
> > > I have thought about something like this before. An idea I had is to
> > > make es_result_relations array indexable by plain RT indexes, then we
> > > don't need to maintain separate indexes that we do today for result
> > > relations.
> >
> > That sounds like a good idea. es_result_relations is currently an array
> > of ResultRelInfos, so that would leave a lot of unfilled structs in the
> > array. But in on of your other threads, you proposed turning
> > es_result_relations into an array of pointers anyway
> > (https://www.postgresql.org/message-id/CA+HiwqE4k1Q2TLmCAvekw+8_NXepbnfUOamOeX=KpHRDTfSKxA@mail.gmail.com).
>
> Okay, I am reorganizing the patches around that idea and will post an
> update soon.

Attached updated patches.

0001 makes es_result_relations an RTI-indexable array, which allows to
get rid of all "result relation index" fields across the code.

--
Amit Langote
EDB: http://www.enterprisedb.com