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 bchesneau@gmail.com
 designates 72.14.220.152 as permitted sender)
DomainKey-Signature: a=rsa-sha1; c=nofws;
        d=gmail.com; s=gamma;
        h=mime-version:date:message-id:subject:from:to:content-type
         :content-transfer-encoding;
        b=pDkueG3MAz7Map2xHfFUK+eT/YOp5OZfMVL1jPmP4mWVzk4IgP953bD7LmkmVVgx5C
         g7JX9VrBqLfn8Q0fxt3dGz8r+WcVNHa7/M9MnIRYr3pfJ1mbkC2S/YK5P/15V8i5z20J
         bBwxkceA6kCLeK7TmQOMyd4MBZSJ4Hc+zjst0=
MIME-Version: 1.0
Date: Wed, 10 Jun 2009 12:25:41 +0200
Message-ID: <b7cd8ed10906100325m2aba0b9cuf9efba013b02af1a@mail.gmail.com>
Subject: calulate doc popularity with couchdb map/reduce
From: Benoit Chesneau <bchesneau@gmail.com>
To: user@couchdb.apache.org
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

Hi all,

I'm trying to calculate popularity of a doc wiyth couchdb but for now
have "reduce_overflow_error" error. What i do is saving each vote as a
doc :

{
  docid: id of document where vote happened,
  v :  value of doc,
  d: date when docid have been created

}

Then I do map/reduce over it :

map :
function(doc) {
  if (doc.type =3D=3D "vote") {
    emit(doc.itemid, {d: doc.d, v: doc.v});
  }
}

reduce :
function(keys, values, rereduce) {
  // !code vendor/couchapp/date.js

  /* we use reddit algorithm to calculate scores
  http://blog.linkibol.com/post/How-to-Build-a-Popularity-Algorithm-You-can=
-be-Proud-of.aspx
  http://news.ycombinator.com/item?id=3D231168 */

  // all started at this date

  var scores =3D {};
  var points =3D {};
  if (!rereduce) {
    var B =3D new Date("Thu May 28 11:16:49 2009 +0200").valueOf();
    for (var k in keys) {
      if (!scores[keys[k][0]]) {
        scores[keys[k][0]] =3D 0;
        points[keys[k][0]] =3D {
          A: new Date().setRFC3339(values[k].d).valueOf(),
          x: 0
        }
      }

      var t =3D points[keys[k][0]].A - B;
      points[keys[k][0]].x +=3D values[k].v;

      if (points[keys[k][0]].x  > 0) {
        y =3D 1;
      } else if (points[keys[k][0]].x  =3D=3D 0) {
        y =3D 0;
      } else {
        y =3D -1;
      }

      z =3D (Math.abs(points[keys[k][0]].x) >=3D1 && points[keys[k][0]].x  =
|| 1);
      scores[keys[k][0]] =3D Math.log(z) + (y*t)/45000;

    }
    lastkey =3D keys[keys.length-1][0];

  } else {
    scores =3D values[0];
    for(var v =3D 1; v < values.length; v++) {
      for(var s in values[v]) {
        if (scores[s])
          scores[s] +=3D values[v][s];
        else
          scores[s] =3D values[v][s];
      }
    }
  }

  var top =3D [];
  for (k in scores)
    top[top.length] =3D [k, scores[k]];
  top.sort(function(a, b) { return a[1] - b[1] });

  for(var n =3D 20; n < top.length; n++)
    if(top[n][0] !=3D lastkey) scores[top[n][0]] =3D undefined;

  return scores;
}

Is this kind of reduce ok  with couchdb ? The other solution would be
saving scores without sorting them in a temporary db and make another
map/reuce over it. Maybe I should go that way, but I wonder if
somepone have a better idee ?

- beno=EEt