Re: prefix btree implementation

From: "Qingqing Zhou" <zhouqq(at)cs(dot)toronto(dot)edu>
To: pgsql-hackers(at)postgresql(dot)org
Subject: Re: prefix btree implementation
Date: 2005-10-05 19:56:15
Message-ID: di10jf$2jdr$1@news.hub.org
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers


"Bricklen Anderson" <BAnderson(at)PresiNET(dot)com> wrote
>
> Oracle implements something similar called index compression, but I
> believe it
> is only for common column values. I haven't checked in versions>9r1 so
> maybe
> there are other options implemented by now.
>
> Jonathan Lewis describes some pros and cons here:
> http://www.jlcomp.demon.co.uk/faq/compress_ind.html
>

Oracle 9 uses the grammar like this:

CREATE INDEX ... [ COMPRESS <number_of_first_columns> ]

So it gives the flexibility of choosing optimal number of coulumns to the
user. The script mentioned in the article guesses the optimal number by
estimating the size of each choice. But I am thinking we can do it better:
(1) we don't require that the compressed number of columns on each page are
the same; (2) when we build up index bottom-up, we can determine this number
for each page automatically by maximizing the number of items within a page.

Regards,
Qingqing

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Gaetano Mendola 2005-10-05 20:13:21 Re: wrong optimization ( postgres 8.0.3 )
Previous Message Jonah H. Harris 2005-10-05 19:54:42 Re: [PERFORM] A Better External Sort?