Re: [PATCH] bitmap indexes

From: Jeff Janes <jeff(dot)janes(at)gmail(dot)com>
To: Alvaro Herrera <alvherre(at)2ndquadrant(dot)com>
Cc: Abhijit Menon-Sen <ams(at)2ndquadrant(dot)com>, pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>, bausch(at)dvs(dot)tu-darmstadt(dot)de
Subject: Re: [PATCH] bitmap indexes
Date: 2013-09-26 15:39:05
Message-ID: CAMkU=1zUjrG=5RRLJSELrmTXXDg177aLy-2264sXXOD7zQiM0A@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Wed, Sep 25, 2013 at 3:20 PM, Alvaro Herrera <alvherre(at)2ndquadrant(dot)com>wrote:

> Hi,
>
> Here are some quick items while skimming this patch. I am looking at
> commit 6448de29d from your github repo, branch bmi.
>
> What's with the pg_bitmapindex stuff in pg_namespace.h? It doesn't seem
> to be used anywhere.
>
> This led me to research how these indexes are stored. I note that what
> we're doing here is to create another regular table and a btree index on
> top of it, and those are the ones that actually store the index data.
> This seems grotty and completely unlike the way we do things elsewhere
> (compare GIN indexes which have rbtrees inside them).

+1 on that. I don't know about grottiness, but it certainly seems like it
would deadlock like crazy. Which another product's bitmap indexes are
infamous for, but since we don't need to store visibility information in
our indexes, hopefully we can do better.

Cheers,

Jeff

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Robert Haas 2013-09-26 16:44:28 Re: FW: REVIEW: Allow formatting in log_line_prefix
Previous Message Robert Haas 2013-09-26 15:18:21 Re: record identical operator