Question

I’ve run into some strange differences between the mongodb running on MongoHQ and the version running on my own development machine. Specifically, when calling .toString() on an object id inside a MapReduce map function, the results vary:

On my own machine:

ObjectId('foo').toString()  // => 'foo'

On MongoHQ:

ObjectId('foo').toString()  // => 'ObjectId(\'foo\')'

Note: The id’s I use are actual mongodb id’s – not just ‘foo’ etc. as in these examples

I would expect .toString() to behave like on my own machine – not how it’s behaving on MongoHQ. How come it’s not?

My local OSX version of MongoDB is installed using Homebrew and is version 2.0.1-x86_64

To show what’s actually going on, I’ve build a little test case. Let’s assume that we have a users collection with a friends attribute, being an array of user ids:

> db.users.find()
{ _id: ObjectId('a'), friends: [ObjectId('b'), ObjectId('c')] },
{ _id: ObjectId('b'), friends: [] },
{ _id: ObjectId('c'), friends: [] }

As you can see a is friends with b and c where as b and c isn’t friends with anybody.

Now let’s look at a working test-algorithm:

var map = function() {
  this.friends.forEach(function(f) {
    emit(f, { friends: 1, user: user, friend: f.toString() });
  });
};

var reduce = function(k, vals) {
  var result = { friends: 0, user: [], friend: [] };

  vals.forEach(function(val) {
    result.friends += val.friends;
    result.user.push(val.user);
    result.friend.push(val.friend);
  });

  return result;
};

var id = ObjectId('50237c6d5849260996000002');

var query = {
  query   : { friends: id },
  out     : { inline: 1 },
  scope   : { user: id.toString() },
  jsMode  : true,
  verbose : true
};

db.users.mapReduce(map, reduce, query);

Assuming id is set to an id of a user who is a friend of someone in the users collection, then the output returned by the mapReduce method on MongoHQ will look like this:

{
  "results" : [
    {
      "_id" : ObjectId("50237c555849260996000001"),
      "value" : {
        "friends" : 1,
        "user" : "50237c6d5849260996000002",
        "friend" : "ObjectId(\"50237c555849260996000001\")"
      }
    },
    {
      "_id" : ObjectId("50237c74c271be07f6000002"),
      "value" : {
        "friends" : 1,
        "user" : "50237c6d5849260996000002",
        "friend" : "ObjectId(\"50237c74c271be07f6000002\")"
      }
    }
  ],
  "timeMillis" : 0,
  "timing" : {
    "mapTime" : 0,
    "emitLoop" : 0,
    "reduceTime" : 0,
    "mode" : "mixed",
    "total" : 0
  },
  "counts" : {
    "input" : 1,
    "emit" : 2,
    "reduce" : 0,
    "output" : 2
  },
  "ok" : 1,
}

As you can see, the friend attribute in each result is not just a string containing the id, but a string containing the actual method call.

Did I run this on my own machine, the results array would have been:

{
  "_id" : ObjectId("50237c555849260996000001"),
  "value" : {
    "friends" : 1,
    "user" : "50237c6d5849260996000002",
    "friend" : "50237c555849260996000001"
  }
},
{
  "_id" : ObjectId("50237c74c271be07f6000002"),
  "value" : {
    "friends" : 1,
    "user" : "50237c6d5849260996000002",
    "friend" : "50237c74c271be07f6000002"
  }
}

Answer 1

MongoHQ is running a different version of MongoDB than you are.

To get the behavior of your homebrew version, try changing your map function:

var map = function() {
  this.friends.forEach(function(f) {
    emit(f, { friends: 1, user: user.str, friend: f.str });
  });
};