Re: Tuple-routing for certain partitioned tables not working as expected

On 2017/08/30 17:20, Etsuro Fujita wrote:
> Here is a patch to skip the CheckValidResultRel checks for a
> target table if it's a foreign partition to perform tuple-routing for,
> as proposed by Robert.

In the patch, to skip the checks, I passed to CheckValidResultRel a new
flag indicating whether the target relation is a partition to do
tuple-routing for, but while updating another patch for tuple-routing
for foreign partitions in PG11, I noticed that it would be better to
pass ResultRelInfo than that flag. The reason is: (1) we can see
whether the relation is such a partition by looking at the
ResultRelInfo's ri_PartitionRoot, instead of that flag, and (2) this is
for tuple-routing for foreign partitions, but we could extend that
function with that argument to do the checks without throwing an error
and save the result in a new member of ResultRelInfo (eg,
ri_PartitionIsValid) so that we can use that info to determine in
ExecInsert whether a foreign partition chosen by ExecFindPartition is
valid for tuple-routing. So, I updated the patch that way. Attached is
an updated version of the patch.

(I discussed about lazy initialization of partition ResultRelInfos
before, but I changed my mind because I noticed that the change to
EXPLAIN for INSERT into a partitioned table, which I'd like to propose
in the tuple-routing-for-foereign-partitions patch, needs partition
ResultRelInfos.)

Best regards,
Etsuro Fujita