MongoDB Fancy Ass Grouping

So lately I’ve been playing with Mongodb for CrowdFlower. We need to calculate statistics from the information that we gather. Mongo provides support for this but has little documentation. I thought it would be appropriate to describe how I got some of the fancier group statements working. Let’s take this document format as an example:

    {
     "state": "finished",
     "data": {
       "everythings_good": "yes"
     }
   }
  

The key property alone of the group function can not be used because our key is in an embedded member. We have to use the keyf property of the group command. I’m using the Mongo ruby library and it currently doesn’t support the use of this property. Instead I issue the raw db_command:

    db = Mongo::Connection.new.db('docs')
    db.db_command({
      "group" => {
        "ns" => "ratings",
        "$reduce" => Mongo::Code.new(<<-JS),
          function(obj,prev){ prev.count++; }
        JS
        "$keyf" => Mongo::Code.new(<<-JS),
          function(obj){
            var g = {}
            g[obj.data.everythings_good] = true
            return g
          }
        JS
        "cond" => {},
        "initial" => {:count => 0}
      }
    })
  

This command returns the following value:

    {
      "retval"=>[
        {"yes"=>1.0, "count"=>17355.0}, 
        {"no"=>1.0, "count"=>6395.0}
      ], 
      "count"=>23750.0, 
      "keys"=>2, 
      "ok"=>1.0
    }
  

Note: the keyf property was not working properly until version 1.1.0 of mongodb