DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma;
	h=mime-version:in-reply-to:references:date:message-id:subject:from:to
	:cc:content-type;
	b=GOfePyUpNMmzGuDx25gZR/phrtueyYfuMqD42M0AAGQ3HOuuMUsNZ60CZgGOZxSxjC
	zaRS8P/TyM0qXMwCsuiXN2yZ8P6m9+/ZLcQSMHDv+uQFIaKBao5GC+tk3IJQv2MpI6Ub
	vv6XvzR1WVMeEU0lmc2Nma5ipODqwH50/seBU=
MIME-Version: 1.0
In-Reply-To: <4C5E25FC.9080303@postnewspapers.com.au>
References: <i3c4a7$j6r$1@speranza.aioe.org>
	<4C5E25FC.9080303@postnewspapers.com.au>
Date: Sun, 8 Aug 2010 15:40:07 +0200
Message-ID: <AANLkTinnVM-gqAXzV4jgr6F+CVH3wUa9R7inXmWHpDeP@mail.gmail.com>
Subject: Re: Using SKYLINE command on PostgreSQL
From: Yeb Havinga <yebhavinga@gmail.com>
To: Craig Ringer <craig@postnewspapers.com.au>
Cc: LaboRecherche <nomail@altranx.com>, pgsql-sql@postgresql.org
Content-Type: multipart/alternative; boundary=00163628528cc912d0048d50095e

--00163628528cc912d0048d50095e
Content-Type: text/plain; charset=ISO-8859-1

On Sun, Aug 8, 2010 at 5:35 AM, Craig Ringer <craig@postnewspapers.com.au>wrote:

>
> The masters thesis page does contain a patch against PostgreSQL 8.3:
>  http://skyline.dbai.tuwien.ac.at/
>

That page has a nice web page and examples where queries can be run. 'The
skyline operator' is described in
http://www.dbis.ethz.ch/research/publications/38.pdf. It reminds me a bit of
the staircase join operator, something on my personal wish list for postgres
as well (latest patch seems to have got lost on the internet).

>
> They say they're porting the patch to 8.4devel, but I suspect that since
> 8.4 has been out for ages and 9.0 is almost released, they've probably
> lost interest and dropped the work.
>

I looked a bit in the latest patch. There's a lot of FIXME's and TODO's in
it. It adds at least one plan node, so it's probably hard to rewrite into
contrib form. That'll probably make it harder to get accepted for mainline
postgres, also because the skyline queries can be translated into 'normal'
sql (referenced pdf, section 3.1), though the authors claim the performance
using skyline is better. Since 8.3 CTE's have been introduced and maybe that
makes writing these kinds of queries easier. Also the knngist ordering of
'interesting points' might be a solution for problems in the same ballpark.

regards,
Yeb Havinga

--00163628528cc912d0048d50095e
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

<div class=3D"gmail_quote">On Sun, Aug 8, 2010 at 5:35 AM, Craig Ringer <sp=
an dir=3D"ltr">&lt;<a href=3D"mailto:craig@postnewspapers.com.au">craig@pos=
tnewspapers.com.au</a>&gt;</span> wrote:<br><blockquote class=3D"gmail_quot=
e" style=3D"border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt =
0.8ex; padding-left: 1ex;">
<br>
The masters thesis page does contain a patch against PostgreSQL 8.3:<br>
 =A0<a href=3D"http://skyline.dbai.tuwien.ac.at/" target=3D"_blank">http://=
skyline.dbai.tuwien.ac.at/</a><br></blockquote><div><br>That page has a nic=
e web page and examples where queries can be run. &#39;The skyline operator=
&#39; is described in <a href=3D"http://www.dbis.ethz.ch/research/publicati=
ons/38.pdf">http://www.dbis.ethz.ch/research/publications/38.pdf</a>. It re=
minds me a bit of the staircase join operator, something on my personal wis=
h list for postgres as well (latest patch seems to have got lost on the int=
ernet).<br>
</div><blockquote class=3D"gmail_quote" style=3D"border-left: 1px solid rgb=
(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<br>
They say they&#39;re porting the patch to 8.4devel, but I suspect that sinc=
e<br>
8.4 has been out for ages and 9.0 is almost released, they&#39;ve probably<=
br>
lost interest and dropped the work.<br></blockquote><div><br>I looked a bit=
 in the latest patch. There&#39;s a lot of FIXME&#39;s and TODO&#39;s in it=
. It adds at least one plan node, so it&#39;s probably hard to rewrite into=
 contrib form. That&#39;ll probably make it harder to get accepted for main=
line postgres, also because the skyline queries can be translated into &#39=
;normal&#39; sql (referenced pdf, section 3.1), though the authors claim th=
e performance using skyline is better. Since 8.3 CTE&#39;s have been introd=
uced and maybe that makes writing these kinds of queries easier. Also the k=
nngist ordering of &#39;interesting points&#39; might be a solution for pro=
blems in the same ballpark.<br>
<br>regards,<br>Yeb Havinga<br></div></div><br>

--00163628528cc912d0048d50095e--