Re: path toward faster partition pruning

From: Amit Langote <Langote_Amit_f8(at)lab(dot)ntt(dot)co(dot)jp>
To: Robert Haas <robertmhaas(at)gmail(dot)com>
Cc: Beena Emerson <memissemerson(at)gmail(dot)com>, Rajkumar Raghuwanshi <rajkumar(dot)raghuwanshi(at)enterprisedb(dot)com>, David Rowley <david(dot)rowley(at)2ndquadrant(dot)com>, Dilip Kumar <dilipbalaut(at)gmail(dot)com>, Pg Hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: path toward faster partition pruning
Date: 2017-10-27 06:12:27
Message-ID: 8499324c-8a33-4be7-9d23-7e6a95e60ddf@lab.ntt.co.jp
Views: Raw Message | Whole Thread | Download mbox
Thread:
Lists: pgsql-hackers

On 2017/10/27 13:57, Robert Haas wrote:
> On Fri, Oct 27, 2017 at 3:17 AM, Amit Langote
> <Langote_Amit_f8(at)lab(dot)ntt(dot)co(dot)jp> wrote:
>>> I don't think we really want to get into theorem-proving here, because
>>> it's slow.
>>
>> Just to be clear, I'm saying we could use theorem-proving (if at all) just
>> for the default partition.
>
> I don't really see why it should be needed there either. We've got
> all the bounds in order, so we should know where there are any gaps
> that are covered by the default partition in the range we care about.

Sorry, I forgot to add: "...just for the default partition, for cases like
the one in Beena's example."

In normal cases, default partition selection doesn't require any
theorem-proving. It proceeds in a straightforward manner more or less
like what you said it should.

After thinking more on it, I think there is a rather straightforward trick
to implement the idea you mentioned to get this working for the case
presented in Beena's example, which works as follows:

For any non-root partitioned tables, we add the list of its partition
constraint clauses to the query-provided list of clauses and use the whole
list to drive the partition-pruning algorithm. So, when partition-pruning
runs for tprt_1, along with (< 10000) which the original query provides,
we also have (>= 1) which comes from the partition constraint of tprt_1
(which is >= 1 and < 50000). Note that there exists a trick in the new
code for the (< 50000) coming from the constraint to be overridden by the
more restrictive (< 10000) coming from the original query.

Thanks,
Amit

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Pavel Stehule 2017-10-27 06:16:56 Re: proposal: schema variables
Previous Message sanyam jain 2017-10-27 05:56:12 Burst in WAL size when UUID is used as PK while full_page_writes are enabled