Re: WIP: cross column correlation ...
Martijn van Oosterhout <kleptog@svana.org>
From: Martijn van Oosterhout <kleptog@svana.org>
To: Greg Stark <gsstark@mit.edu>
Cc: PostgreSQL - Hans-Jürgen Schönig <postgres@cybertec.at>, Robert Haas <robertmhaas@gmail.com>, Grzegorz Jaskiewicz <gj@pointblue.com.pl>, Bruce Momjian <bruce@momjian.us>, Josh Berkus <josh@agliodbs.com>, pgsql-hackers Hackers <pgsql-hackers@postgresql.org>, Boszormenyi Zoltan <zb@cybertec.at>
Date: 2011-02-26T18:58:20Z
Lists: pgsql-hackers
On Sat, Feb 26, 2011 at 06:44:52PM +0000, Greg Stark wrote: > 2011/2/26 PostgreSQL - Hans-Jürgen Schönig <postgres@cybertec.at>: > > what we are trying to do is to explicitly store column correlations. so, a histogram for (a, b) correlation and so on. > > The problem is that we haven't figured out how to usefully store a > histogram for <a,b>. Consider the oft-quoted example of a > <city,postal-code> -- or <city,zip code> for Americans. A histogram > of the tuple is just the same as a histogram on the city. But there are cases where it can work. Frankly the example you mention is odd because for we can't even build useful 1D histograms for <city> and <zip code>, so the fact that 2D is hard is not surprising. The histograms we do build work fine from > and <, just equality. The 2D will handle the same. Have a nice day, -- Martijn van Oosterhout <kleptog@svana.org> http://svana.org/kleptog/ > Patriotism is when love of your own people comes first; nationalism, > when hate for people other than your own comes first. > - Charles de Gaulle