Subject: Re: Unicode normalization SQL functions
To: Daniel Verite <daniel@manitou-mail.org>
References: <2c5e8df9-43b8-41fa-88e6-286e8634f00a@manitou-mail.org>
Cc: Andreas Karlsson <andreas@proxel.se>,
 pgsql-hackers <pgsql-hackers@postgresql.org>
From: Peter Eisentraut <peter.eisentraut@2ndquadrant.com>
Organization: 2ndQuadrant
Message-ID: <b0c0dbbd-9142-6bf8-4483-3a46360c55d3@2ndquadrant.com>
Date: Wed, 19 Feb 2020 21:23:12 +0100
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.14; rv:52.0)
 Gecko/20100101 Thunderbird/52.9.1
MIME-Version: 1.0
In-Reply-To: <2c5e8df9-43b8-41fa-88e6-286e8634f00a@manitou-mail.org>
Content-Type: text/plain; charset=iso-8859-15; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 2020-02-17 20:08, Daniel Verite wrote:
> Concerning execution speed, there's an excessive CPU usage when
> normalizing into NFC or NFKC. Looking at pre-existing code, it looks
> like recompose_code() in unicode_norm.c looping over the
> UnicodeDecompMain array might be very costly.

Yes, this is a known issue and I think room for future optimization work.

> Another point is that the ICU-based implementation appears
> to be significantly faster in all cases, which makes me wonder
> why ICU builds should not just use ICU instead of the PG-core
> implementation.

That would require linking libpq to ICU (for SASLprep), and in general 
would either make ICU required or require maintaining multiple 
implementations.  I don't think we're there yet.

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