Re: partition routing layering in nodeModifyTable.c

On Sun, Oct 18, 2020 at 12:54 AM Alvaro Herrera <alvherre(at)alvh(dot)no-ip(dot)org> wrote:
> On 2020-Oct-17, Amit Langote wrote:
> > Hmm, I don't see ri_PartitionCheckExpr as being a piece of routing
> > information, because it's primarily meant to be used when inserting
> > *directly* into a partition, although it's true we do initialize it in
> > routing target partitions too in some cases.
> >
> > Also, ChildToRootMap was introduced by the trigger transition table
> > project, not tuple routing. I think we misjudged this when we added
> > PartitionToRootMap to PartitionRoutingInfo, because it doesn't really
> > belong there. This patch fixes that by removing PartitionToRootMap.
> >
> > RootToPartitionMap and the associated partition slot is the only piece
> > of extra information that is needed by tuple routing target relations.
>
> Well, I was thinking on making the ri_PartitionInfo be about
> partitioning in general, not just specifically for partition tuple
> routing. Maybe Heikki is right that it may end up being simpler to
> remove ri_PartitionInfo altogether. It'd just be a couple of additional
> pointers in ResultRelInfo after all.

So that's 2 votes for removing PartitionRoutingInfo from the tree.
Okay, I have tried that in the attached 0002 patch. Also, I fixed
some comments in 0001 that still referenced PartitionToRootMap.

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