On Wed, Sep 19, 2018 at 02:48:58PM -0700, Andres Freund wrote:
> On 2018-09-19 12:06:47 +0900, Michael Paquier wrote:
>> Yeah, my gut is telling me that this would be the best approach for now,
>> still I am not sure that this is the best move in the long term.
>
> ISTM heap_sync() would be the entirely wrong layer to handle
> partitioning. For several reasons: 1) With pluggable storage, we want to
> have multiple different table implementations, doing the syncing on the
> heap_* for partitions would thus be wrong. 2) In just about all cases we
> only want to sync a few partitions, there's not really a use-case for
> doing syncs across all partitions imo.
I haven't considered things from the angle of 1), which is a very good
point. 2) is also a good argument.
>> All the other callers of heap_sync don't care about partitioned
>> tables, so we could add an assertion on RELKIND_PARTITIONED_TABLE.
>
> Or rather, it should assert the expected relkinds?
Yeah, I think that we are coming back to what heap_create assumes in
term of which relkinds should have storage or not, so a global macro
able to do the work would be adapted perhaps?
--
Michael