Message-ID: <53F5BC47.3080000@lab.ntt.co.jp>
Date: Thu, 21 Aug 2014 18:30:47 +0900
From: Etsuro Fujita <fujita.etsuro@lab.ntt.co.jp>
User-Agent: Mozilla/5.0 (Windows NT 6.0;
 rv:24.0) Gecko/20100101 Thunderbird/24.6.0
MIME-Version: 1.0
To: Noah Misch <noah@leadboat.com>, shigeru.hanada@gmail.com,
 Kyotaro HORIGUCHI <horiguchi.kyotaro@lab.ntt.co.jp>
CC: pgsql-hackers@postgresql.org
Subject: Re: inherit support for foreign tables
References: <53393E18.3080901@lab.ntt.co.jp> <533C01C7.6030809@lab.ntt.co.jp>
 <53706FB6.6000403@lab.ntt.co.jp>
 <20140620.170406.123403564.horiguchi.kyotaro@lab.ntt.co.jp>
 <20140702022327.GC1586927@tornado.leadboat.com>
In-Reply-To: <20140702022327.GC1586927@tornado.leadboat.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk
Sender: pgsql-hackers-owner@postgresql.org

(2014/07/02 11:23), Noah Misch wrote:
> On Fri, Jun 20, 2014 at 05:04:06PM +0900, Kyotaro HORIGUCHI wrote:
>> Attached is the rebased patch of v11 up to the current master.

> The rest of these review comments are strictly observations; they're not
> requests for you to change the patch, but they may deserve more discussion.
>
> We use the term "child table" in many messages.  Should that change to
> something more inclusive, now that the child may be a foreign table?  Perhaps
> one of "child relation", plain "child", or "child foreign table"/"child table"
> depending on the actual object?  "child relation" is robust technically, but
> it might add more confusion than it removes.  Varying the message depending on
> the actual object feels like a waste.  Opinions?

IMHO, I think that "child table" would not confusing users terribly.

> LOCK TABLE on the inheritance parent locks child foreign tables, but LOCK
> TABLE fails when given a foreign table directly.  That's odd, but I see no
> cause to change it.

I agree wth that.

> The longstanding behavior of CREATE TABLE INHERITS is to reorder local columns
> to match the order found in parents.  That is, both of these tables actually
> have columns in the order (a,b):
>
> create table parent (a int, b int);
> create table child (b int, a int) inherits (parent);
>
> Ordinary dump/reload always uses CREATE TABLE INHERITS, thereby changing
> column order in this way.  (pg_dump --binary-upgrade uses ALTER TABLE INHERIT
> and some catalog hacks to avoid doing so.)  I've never liked that dump/reload
> can change column order, but it's tolerable if you follow the best practice of
> always writing out column lists.  The stakes rise for foreign tables, because
> column order is inherently significant to file_fdw and probably to certain
> other non-RDBMS FDWs.  If pg_dump changes column order in your file_fdw
> foreign table, the table breaks.  I would heartily support making pg_dump
> preserve column order for all inheritance children.  That doesn't rise to the
> level of being a blocker for this patch, though.

I agree with that, too.  I think it would be better to add docs for now.

Thanks,

Best regards,
Etsuro Fujita


-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers