Date: Sat, 11 Jan 2020 14:31:59 +0100
From: Tomas Vondra <tomas.vondra@2ndquadrant.com>
To: David Fetter <david@fetter.org>
Cc: Andrew Gierth <andrew@tao11.riddles.org.uk>,
	Kyotaro Horiguchi <horikyota.ntt@gmail.com>,
	pgsql-hackers@postgresql.org
Subject: Re: Efficient output for integer types
Message-ID: <20200111133159.ppxnocexkdduo4m6@development>
References: <20190920210916.GF31596@fetter.org>
 <20190920211813.GG31596@fetter.org>
 <87h856pekw.fsf@news-spur.riddles.org.uk>
 <20190921060835.GH31596@fetter.org>
 <87a7ayp4bm.fsf@news-spur.riddles.org.uk>
 <20190922215804.GL31596@fetter.org>
 <87pnjrns08.fsf@news-spur.riddles.org.uk>
 <20190923213506.GN31596@fetter.org>
 <20190924043018.GO31596@fetter.org>
 <20190924052620.GP31596@fetter.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Disposition: inline
In-Reply-To: <20190924052620.GP31596@fetter.org>
Precedence: bulk

Hi,

Any plans regarding committing this patch? I see the thread is silent
since September 24, when the last patch version was posted. The patch is
already marked as RFC since December, when David changed the status. I
don't have any opinion whether the patch is RFC or not (it might well
be), but IMHO it should have been mentioned in this thread.

I did a quick test to see how much more efficient this is, and for a
table with 10 bigint columns and 5M random rows the COPY to /dev/null
went from 3000 ms to ~2700 ms. That's not the 30% speedup mentioned by
David in the first message, but 10% is still pretty nice.

Of course, for real-world use cases the speedup will be lower because of
using other data types too, I/O etc. But it's still nice.

So, is anyone opposed to pushing this? If not, who'll to do that?  I see
Andrew Gierth was involved in the discussions on IRC and it's related to
the Ryu patch, so maybe he want's to take care of this. Andrew?


regards

-- 
Tomas Vondra                  http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services