Mailing-List: contact user-help@couchdb.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@couchdb.apache.org
Received-SPF: pass (athena.apache.org: domain of alon.keren@gmail.com
 designates 74.125.83.52 as permitted sender)
MIME-Version: 1.0
In-Reply-To: 
 <CAHZBNKabA4NaV7984HBnNwr2B3VeaufR7fT2DwtRxivvzP9Qqg@mail.gmail.com>
References: 
 <CAPTWwjAMmDhv1uaCM25J-xoOiRZRgJRLC-p4WgtJ3_Bje2Q1yA@mail.gmail.com>
	<CAHZBNKabA4NaV7984HBnNwr2B3VeaufR7fT2DwtRxivvzP9Qqg@mail.gmail.com>
Date: Wed, 14 Dec 2011 11:01:53 +0200
Message-ID: 
 <CABzyt1G3wBhZqfceGSLyV1FjOVXo80Z7rkp+9ko_NxkeuTOn0Q@mail.gmail.com>
Subject: Re: iterative map reduce
From: Alon Keren <alon.keren@gmail.com>
To: user@couchdb.apache.org
Content-Type: multipart/alternative; boundary=0015174c3e3479a3d204b4099e7f

--0015174c3e3479a3d204b4099e7f
Content-Type: text/plain; charset=ISO-8859-1

I've been working on an open-source tool with a similar approach chaining
map-reduce:
https://github.com/afters/Couch-Incarnate

The need to pay the bills withheld me from putting more time into it
lately, so I can't say it's production quality.
Hopefully, this will change soon.

Dominic - the method I currently use in production is to poll the changes
on my DB every few seconds and update a 'summary' DB accordingly. It's not
pretty, but it works. I hope to replace it with Couch-Incarnate asap.

  Alon

On 14 December 2011 10:50, Dave Cottlehuber <dave@muse.net.nz> wrote:

> On 14 December 2011 05:46, Dominic Tarr <dominic.tarr@gmail.com> wrote:
> > hi,
> >
> > I need a iterative map reduce, prefurably, in couchdb.
> >
> > I want to be able to do a map reduce to generate unique items, then
> another
> > map reduce to generate stats about those unique items.
> >
> > from what I know about couchdb at the moment, it seems like id have to do
> > the first step at the document creation step, which is just ugly.
> >https://mail.google.com/mail/?shva=1#inbox
> > has anyone else dealt with this problem?
>
> Hi Dom,
>
> Cloudant do this in BigCouch which they released as open source a
> while back http://support.cloudant.com/kb/views/chained-mapreduce-views
> there are a few other posts on chained MR around but I don't know of a
> better Couch-style implementation. Their basic approach is to copy the
> view output into a new db which you'd then re-MR as required. Rinse
> and repeat until complete.
>
> Depending on your data it might be possible to implement their
> view->copy to db -> new view approach yourself in native couch. It
> would be interesting to do this as a couch extension using that nifty
> plugin interface that hasn't been developed yet!
>
> A+
> Dave
>

--0015174c3e3479a3d204b4099e7f--