To: Gregory Stark <stark@enterprisedb.com>
cc: pgsql-patches@postgresql.org, "Jan Wieck" <JanWieck@Yahoo.com>
Subject: Re: Proposed patch to change TOAST compression strategy 
In-reply-to: <87ir0mo8ot.fsf@oxford.xeocode.com> 
References: <26048.1203271939@sss.pgh.pa.us>
	<87ir0mo8ot.fsf@oxford.xeocode.com>
Comments: In-reply-to Gregory Stark <stark@enterprisedb.com>
	message dated "Mon, 18 Feb 2008 10:33:38 +0000"
Date: Fri, 07 Mar 2008 18:23:12 -0500
Message-ID: <8565.1204932192@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Gregory Stark <stark@enterprisedb.com> writes:
> "Tom Lane" <tgl@sss.pgh.pa.us> writes:
>> * Adds an early-failure path to the compressor as suggested by Jan:
>> if it's been unable to find even one compressible substring in the
>> first 1KB (parameterizable), assume we're looking at incompressible
>> input and give up.  (There are various ways this could be done, but
>> this way seems to add the least overhead to the compressor's inner
>> loop.)

> I'm not sure how to test the rest of it, but this bit seems testable. I fear
> this may be too conservative. Even nigh incompressible data will find a few
> backreferences.
> I'll try some tests and see.

On the strength of Teodor's favorable test, I went ahead and applied
this patch as-is.  We can certainly tweak the compressor logic again
if you can show an improvement by changing it further.

			regards, tom lane