Subject: Re: Indexes on partitioned tables and foreign partitions
References: <20180514172910.wjuvrtcdhcc5y3u7@alvherre.pgsql>
From: Amit Langote <Langote_Amit_f8@lab.ntt.co.jp>
Message-ID: <ed06bf4e-9bd2-1e87-da53-3571b2aa853c@lab.ntt.co.jp>
Date: Tue, 15 May 2018 09:47:24 +0900
User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:52.0) Gecko/20100101
 Thunderbird/52.7.0
MIME-Version: 1.0
In-Reply-To: <20180514172910.wjuvrtcdhcc5y3u7@alvherre.pgsql>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
To: Alvaro Herrera <alvherre@2ndquadrant.com>
Cc: Michael Paquier <michael@paquier.xyz>, Tom Lane <tgl@sss.pgh.pa.us>,
        Robert Haas <robertmhaas@gmail.com>,
        Simon Riggs <simon@2ndquadrant.com>,
        Arseny Sher <a.sher@postgrespro.ru>,
        "pgsql-hackers@lists.postgresql.org"
 <pgsql-hackers@lists.postgresql.org>
Precedence: bulk

On 2018/05/15 2:29, Alvaro Herrera wrote:
> On 2018-May-10, Amit Langote wrote:
> 
>> How about we error out even *before* calling DefineIndex for the 1st time?
>>  I see that ProcessUtilitySlow() gets a list of all partitions when
>> locking them for index creation before calling DefineIndex.  Maybe, just
>> go through the list and error out if one of them is a partition that we
>> don't support creating an index on?
> 
> The overwhelming consensus seems to be for this option, so I pushed your
> patch after some small tweaks -- mostly to simplify unnecessarily
> baroque code.  (I must have been thinking that recursion would happen
> right in ProcessUtilitySlow.  That doesn't match my memories, but I
> can't explain this code otherwise.)  I added a very small test too.

Thank you.

> I think it'd be better to take this out of ProcessUtility also and into
> DefineInde, for cleanliness sake; maybe add a 'recursing' flag to
> DefineIndex.  Not for pg11, though.

Agreed.  Most of the stuff in utility.c is for command dispatch and there
is no reason for this partitioning-related bit to be sitting here.  Moving
it will require perhaps non-trivial adjustment of indexcmds.c code, so
let's leave it for later as you say.

Thanks,
Amit