Re: How to make hash indexes fast - Mailing list pgsql-performance

From Ondrej Ivanič
Subject Re: How to make hash indexes fast
Date
Msg-id CAM6mieL7nSVvvM8m9qOZNBkHGuDBQf5inMPuMZ-JkWrd_1_Ymg@mail.gmail.com
Whole thread Raw
In response to How to make hash indexes fast  (Craig James <craig_james@emolecules.com>)
Responses Re: How to make hash indexes fast
List pgsql-performance
Hi,

On 19 September 2011 11:14, Craig James <craig_james@emolecules.com> wrote:
> DBsig for a hash-collision chain is always the bitwise OR of every record in
> that hash-collision chain.  When you add a record to the hash table, you do
> a bitwise OR of its signature into the existing DBsig.  If you delete a
> record, you erase DBsig and rebuild it by recomputing the signatures of each
> record in the hash-collision chain and ORing them together again.

Sound like a Bloom filter [1] to me.

BTW, Does Postgres use Bloom filter anywhere?

[1] http://en.wikipedia.org/wiki/Bloom_filter

--
Ondrej Ivanic
(ondrej.ivanic@gmail.com)

pgsql-performance by date:

Previous
From: Craig James
Date:
Subject: How to make hash indexes fast
Next
From: Claudio Freire
Date:
Subject: Re: How to make hash indexes fast