Message-ID: <48C8C6D2.2010809@enterprisedb.com>
Date: Thu, 11 Sep 2008 10:20:50 +0300
Organization: EnterpriseDB
User-Agent: Mozilla-Thunderbird 2.0.0.14 (X11/20080509)
MIME-Version: 1.0
To: David Rowley <dgrowley@gmail.com>
CC: 'Tom Lane' <tgl@sss.pgh.pa.us>, 'Peter Eisentraut' <peter_e@gmx.net>,
	pgsql-hackers@postgresql.org
Subject: Re: [PATCHES] TODO item: Implement Boyer-Moore searching
	(First time hacker)
References: <17EF6EF77BD7405E86761C688A73BAB4@amd64>
In-Reply-To: <17EF6EF77BD7405E86761C688A73BAB4@amd64>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
From: Heikki Linnakangas <heikki.linnakangas@enterprisedb.com>

David Rowley wrote:
> The thing that surprised me was that string_to_array didn't perform as well.
> I expected single character searches to perform a little better. I can't
> think why it would be slower now.

Yes, that's strange. I tried to reproduce that here, with a CVS snapshot 
before the patch, and after. With quick testing with psql and \timing 
and the same query you had in that spreadsheet, I couldn't see that kind 
of performance degradation. Oprofile suggests that, on the contrary, 
slightly less time is spent in text_position_next() after the patch, and 
slightly more in text_position_setup(). Together they account for ~10% 
of CPU time in both tests, so a small difference there would be 
insignificant anyway in that test case.

I think we're fine.

-- 
   Heikki Linnakangas
   EnterpriseDB   http://www.enterprisedb.com