Message-ID: <5B5ABE70.40606@lab.ntt.co.jp>
Date: Fri, 27 Jul 2018 15:40:48 +0900
From: Etsuro Fujita <fujita.etsuro@lab.ntt.co.jp>
User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64;
 rv:11.0) Gecko/20120327 Thunderbird/11.0.1
MIME-Version: 1.0
Subject: Re: print_path is missing GatherMerge and CustomScan support
References: 
 <CAD21AoBQMLoc=ohH-oocuAPsELrmk8_EsRJjOyR8FQLZkbE0wA@mail.gmail.com>
 <20180718062202.GC8565@paquier.xyz>
 <CAFjFpReLXu9qa2a-bt9-yDiT6UZn1+nh_1dEVkJuCXu6y838Aw@mail.gmail.com>
 <20180719000722.GA3411@paquier.xyz>
 <CAFjFpRf3yGODbCyrR4zfcbWbQOv=DqtfvwZ=qUh2Wfa3ODe-TA@mail.gmail.com>
 <5B5958D2.7060805@lab.ntt.co.jp>
 <CA+TgmoaDYiXj6mLZ0Cg3nN3yPgpjxQ+8sjanxRdKo1nTan9dXA@mail.gmail.com>
In-Reply-To: 
 <CA+TgmoaDYiXj6mLZ0Cg3nN3yPgpjxQ+8sjanxRdKo1nTan9dXA@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
To: Robert Haas <robertmhaas@gmail.com>
Cc: Ashutosh Bapat <ashutosh.bapat@enterprisedb.com>,
        Michael Paquier <michael@paquier.xyz>,
        Masahiko Sawada <sawada.mshk@gmail.com>,
        PostgreSQL-development <pgsql-hackers@postgresql.org>
Precedence: bulk

(2018/07/27 4:50), Robert Haas wrote:
> On Thu, Jul 26, 2018 at 1:14 AM, Etsuro Fujita
> <fujita.etsuro@lab.ntt.co.jp>  wrote:
>> because we currently don't consider gathering partial child-scan or
>> child-join paths.  I think we might consider that in future, though.
>
> You generally want to put the Gather node as high up in the plan tree
> as possible.  I think the only case in which this is beneficial is if
> you can't put the Gather or Gather Merge node above the Append because
> only some of the children are parallel-safe.  In that case, a separate
> Gather per child can be better than no parallelism at all.  It's a
> rare case, but it can happen. Actually, I thought we had code for this
> already: see the end of apply_scanjoin_target_to_paths().

Agreed.  Thanks for the explanation!

(I think that at least currently, there is no need for the Gather and 
GatherMerge cases in reparameterize_path_by_child, but I don't object to 
keeping those as-is there.)

Best regards,
Etsuro Fujita