Question

I am trying to compute an average value from a collection using the mongodb java driver, like this:

DBObject condition = 
    new BasicDBObject("pluginIdentifier", plugin.getIdentifier());

DBObject initial = new BasicDBObject();

initial.put("count", 0);
initial.put("totalDuration", 0);
String reduce = "function(duration, out) { out.count++; 
    out.totalDuration+=duration.floatApprox; }";
String finalize = "function(out) { out.avg = out.totalDuration.floatApprox / 
    out.count; }";

DBObject avg = durationEntries.group(
    new BasicDBObject("pluginIdentifier", true), 
    condition, initial, reduce, finalize);

System.out.println(avg);

“duration” is a NumberLong (in java, it is a Long, probably the java driver converts it).
I figured out after some searching that in order to extract the number, using .floatApprox was one way to go, and this also works in the mongodb console:

> db.DurationEntries.findOne().duration.floatApprox
5

However, running the above java code won’t compute an average, but returns this instead

[{"pluginIdentifier":"dummy", "count":7.0, "totalDuration":NaN, "avg":NaN}]

I tried several variations, with and without .floatApprox, but have only been able to obtain some weird string concatenations until now.

My question is: what am I doing wrong / how should I go about to calculate the average of one NumberLong column?

Answer 1

If you’re having problems with map/reduce you should probably drop down into the mongodb console, work it out there and then translate that into your driver.

Take, for example, the following documents:

db.tasks.find()
{ "_id" : ObjectId("4dd51c0a3f42cc01ab0e6506"), "duration" : 10, "name" : "StartProcess", "date" : "20110501" }
{ "_id" : ObjectId("4dd51c0e3f42cc01ab0e6507"), "duration" : 11, "name" : "StartProcess", "date" : "20110502" }
{ "_id" : ObjectId("4dd51c113f42cc01ab0e6508"), "duration" : 12, "name" : "StartProcess", "date" : "20110503" }

You would write the mapReduce to calculate the average duration of StartProcess as follows:

m = function (){
  emit( this.name , { totalDuration : this.duration , num : 1 } );
};

r = function (name, values){
  var n = {totalDuration : 0, num : 0};
  for ( var i=0; i<values.length; i++ ){
    n.totalDuration += values[i].totalDuration;
    n.num += values[i].num;
  }
  return n;
};

f = function(who, res){
  res.avg = res.totalDuration / res.num;
  return res;
};

Then, assuming you’re using MongoDB 1.7 or above:

db.tasks.mapReduce( m, r, { finalize : f, out : {inline : 1} });

Would give you the following answer:

"results" : [
  {
    "_id" : "StartProcess",
      "value" : {
        "totalDuration" : 33,
        "num" : 3,
        "avg" : 11
      }
  }
]

If this doesn’t help, can you post your map function and document structure.