Re: SQL/MED - file_fdw

On Tue, Jan 18, 2011 at 09:33, Kevin Grittner
<Kevin(dot)Grittner(at)wicourts(dot)gov> wrote:
> Review for CF:

Thank your for the review!

> Since it doesn't appear to be intended to change any user-visible
> behavior, I don't see any need for docs or changes to the regression
> tests.

There might be some user-visible behaviors in error messages
because I rearranged some codes to check errors, But we can see
the difference only if we have two or more errors in COPY commands.
They should be not so serious issues.

> So far everything I've done has been with asserts enabled, so I
> haven't tried to get serious benchmarks, but it seems fast.  I will
> rebuild without asserts and do performance tests before I change the
> status on the CF page.
>
> I'm wondering if it would make more sense to do the benchmarking with
> just this patch or the full fdw patch set?  Both?

I tested the performance on my desktop PC, but I cannot see any
differences. But welcome if any of you could test on high-performance
servers.

Comparison with file_fdw would be more interesting
If they have similar performance, we could replace "COPY FROM" to
"CREATE TABLE AS SELECT FROM foreign_table", that is more flexible.

--
Itagaki Takahiro