Horton

Re: [PATCH 4/4] Add tests to dblink covering use of COPY TO FUNCTION

Jeff Davis <pgsql@j-davis.com>

View thread View on postgr.es
From: Jeff Davis <pgsql@j-davis.com>
To: Simon Riggs <simon@2ndQuadrant.com>
Cc: Daniel Farina <drfarina@gmail.com>, Pavel Stehule <pavel.stehule@gmail.com>, Hannu Krosing <hannu@krosing.net>, Greg Smith <greg@2ndquadrant.com>, Robert Haas <robertmhaas@gmail.com>, Daniel Farina <dfarina@truviso.com>, pgsql-hackers@postgresql.org
Date: 2009-11-27T20:55:49Z
Lists: pgsql-hackers 
On Fri, 2009-11-27 at 13:39 +0000, Simon Riggs wrote:
> On Tue, 2009-11-24 at 22:13 -0800, Jeff Davis wrote:
> 
> > My disagreement with the row-by-row approach is more semantics than
> > performance. COPY translates records to bytes and vice-versa, and your
> > original patch maintains those semantics.
> 
> The bytes <-> records conversion is a costly one. Anything we can do to
> avoid that in either direction will be worth it. I would regard
> performance as being part/most of the reason to support this.
> 

Right. I was responding to an idea that copy support sending records
from a table to a function, or from a function to a table, which is
something that INSERT/SELECT can already do.

Our use case is a table to a remote table, so it would go something
like:
 1. COPY TO WITH BINARY on local node
 2. stream output bytes from #1 to remote node
 3. COPY FROM WITH BINARY on remote node

The only faster mechanism that I could imagine is sending the records
themselves, which would be machine-dependent.

Regards,
	Jeff Davis