Hi all,
Álvaro has given faced a use case where it would be useful to have a
function which is able to return the top-most parent of a partition
tree:
https://postgr.es/m/20181204184159.eue3wlchqrkh4vsc@alvherre.pgsql
This has been mentioned as well on the thread where was discussed
pg_partition_tree, but it got shaved from the committed patch as many
things happened when discussing the thing.
Attached is a patch to do the work, which includes documentation and
tests. An argument could be made to include the top-most parent as part
of pg_partition_tree, but it feels more natural to me to have a separate
function. This makes sure to handle invalid relations by returning
NULL, and it generates an error for incorrect relkind.
I have included as well a fix for the recent crash on pg_partition_tree
I have reported, but let's discuss the crash on its thread here:
https://www.postgresql.org/message-id/20181207010406.GO2407@paquier.xyz
The bug fix would most likely get committed first, and I'll rebase this
patch as need be.
I am adding this patch to the CF of January. I think that Amit should
also be marked as a co-author of this patch, as that's inspired from
what has been submitted previously, still I have no reused the code.
Thanks,
--
Michael