Horton

Re: Declarative partitioning - another take

Amit Langote <langote_amit_f8@lab.ntt.co.jp>

View thread View on postgr.es
From: Amit Langote <Langote_Amit_f8@lab.ntt.co.jp>
To: Ashutosh Bapat <ashutosh.bapat@enterprisedb.com>
Cc: Pg Hackers <pgsql-hackers@postgresql.org>
Date: 2016-09-02T06:53:48Z
Lists: pgsql-hackers 
On 2016/09/02 15:22, Ashutosh Bapat wrote:
>>
>>
>>> 2. A combination of constraints on the partitions should be applicable to
>>> the parent. We aren't doing that.
>>
>> How about on seeing that a RELOPT_OTHER_MEMBER_REL is partitioned parent
>> table, we can have get_relation_constraints() include a constant false
>> clause in the list of constraints returned for
>> relation_excluded_by_constraints() to process so that it is not included
>> in the append result by way of constraint exclusion.  One more option is
>> to mark such rels dummy in set_rel_size().
>>
>>
> I am not complaining about having parent relation there. For the people who
> are used to seeing the parent relation in the list of append relations, it
> may be awkward. But +1 if we can do that. If we can't do that, we should at
> least mark with an OR of all constraints on the partitions, so that
> constraint exclusion can exclude it if there are conditions incompatible
> with constraints. This is what would happen in inheritance case as well, if
> there are constraints on the parent. In the above example, the parent table
> would have constraints CHECK ((a >= 0 AND a < 250) OR (a >= 250 and a <
> 500) OR (a >= 500 or a < 600)). It will probably get excluded, if
> constraint exclusion is smart enough to understand ORing.

I guess constraint exclusion would be (is) smart enough to handle that
correctly but why make it unnecessarily spend a *significant* amount of
time on doing the proof (when we *know* we can just skip it).  Imagine how
long the OR list could get.  By the way, my suggestion of just returning a
constant false clause also does not work - neither in case of a query with
restrictions (predicate has to be an OpExpr to go ahead with the proof
which constant false clause is not) nor in case of a query without
restrictions (proof is not performed at all).  So, that one is useless.

Getting rid of the parent table in the append list by other means may be a
way to go.  We know that the table is empty and safe to just drop.

Thanks,
Amit

Commits

  1. Fix typo.

  2. Document trigger-firing behavior for inheritance/partitioning.

  3. Fire per-statement triggers on partitioned tables.

  4. Set ecxt_scantuple correctly for tuple routing.

  5. Fix interaction of partitioned tables with BulkInsertState.

  6. Avoid core dump for empty prepared statement in an aborted transaction.

  7. Fix some problems in check_new_partition_bound().

  8. Remove unnecessary arguments from partitioning functions.

  9. Fix reporting of constraint violations for table partitioning.

  10. Fix tuple routing in cases where tuple descriptors don't match.

  11. Invalid parent's relcache after CREATE TABLE .. PARTITION OF.

  12. Doc: improve documentation about inheritance.