MongoDB group by type and by date

Question

I have a Collection with a simple Document to store impressions and conversions with the following structure:

/* 1 */
{
    "_id" : ObjectId("566f1ef857c1e6dd3123050a"),
    "path_id" : ObjectId("55944c1ebe244fd19cbf510b"),
    "data_type" : "impression",
    "created_at" : ISODate("2015-12-14T19:56:40.100Z"),
    "__v" : 0
}

/* 2 */
{
    "_id" : ObjectId("566f1fc9ac964e6f327c55d6"),
    "path_id" : ObjectId("55944c1ebe244fd19cbf510b"),
    "data_type" : "conversion",
    "created_at" : ISODate("2015-12-14T20:00:09.972Z"),
    "__v" : 0
}

/* 3 */
{
    "_id" : ObjectId("566f2896739f6afa4485f327"),
    "path_id" : ObjectId("562e594315ef3d8c3f05d219"),
    "data_type" : "impression",
    "created_at" : ISODate("2015-12-14T20:37:42.139Z"),
    "__v" : 0
}

/* 4 */
{
    "_id" : ObjectId("566f28e5739f6afa4485f328"),
    "path_id" : ObjectId("562e594315ef3d8c3f05d219"),
    "data_type" : "impression",
    "created_at" : ISODate("2015-12-14T20:39:01.233Z"),
    "__v" : 0
}

I'm able to group and count by data_type, but what I need to do is group by date and then count the data_type in order to get the following result:

[
    {
        '_id': 'Y',
        'conversions': 20,
        'impressions': 2703,
        'date': '2015-12-14'
    },
    {
        '_id': 'Z',
        'conversions': 10,
        'impressions': 1703,
        'date': '2015-12-13'
    } 
]

The code I have right now is the following, but it only groups by data_type. I'm trying to add a project to regroup by date with no luck so far.

var path_id = new mongoose.Types.ObjectId( req.body.path_id );
var match = {
    'path_id': {
        $eq: path_id
    }
};

var group = {
    '_id': '$data_type',
    'count': {
        '$sum': 1
    }
}

Hit.aggregate( [ {
    $match: match
}, {
    $group: group
} ], function( err, res ) {
    console.log( res );
} );

The result is

POST /api/hits/bypath 200 30ms - 15b
[ { _id: 'conversion', count: 2 },
  { _id: 'impression', count: 2703 } ]

Answer 1

To do nested group by date, you have to use to Date aggregation operator $dateToString.

Here is query

db.hits.aggregate([
    {
      "$project": {
        "created_at": {
          "$dateToString": {
            "format": "%Y-%m-%d",
            "date": "$created_at"
          }
        },
        "data_type": true
      }
    },
    {
      "$group": {
        "_id": {
          "data_type": "$data_type",
          "created_at": "$created_at"
        },
        "count": {
          "$sum": 1
        }
      }
    },
    {
      "$group": {
        "_id": {
          "data_type": "$_id.data_type"
        },
        "data":{ "$addToSet" : { count: "$count", date: "$_id.created_at" } }
      }
    }
  ])

If you want to match before group by operation based on condition, Add as following in the query

{
  "$match": {
    "path_id": {
      "$eq": "<path_id>"
    }
  }
}

Answer 2

you can use Date Aggregation Operators to project the day/month/year fields and then group by them

{
  "$project": {
    "y": {
      "$year": "$created_at"
    },
    "m": {
      "$month": "$created_at"
    },
    "d": {
      "$dayOfMonth": "$created_at"
    },
    "data_type" : 1
  }
},
{
  "$group": {
    "_id": {
      "year": "$y",
      "month": "$m",
      "day": "$d",
      "data_type": "$data_type"
    },
    count: {
      "$sum": 1
    }
  }
}

and will output in this format:

    "_id": {
      "year": 2015,
      "month": 10,
      "day": 5,
      "data_type": "impression"
    },
    count: 10

and then group again by date to combine the types in one document

{
  "$group": {
    "_id": {
      "year": "$_id.year",
      "month": "$_id.month",
      "day": "$_id.day"
    },
    types: {"$push":"$_id.data_type"},
    counters: {"$push":"$count"}
  }
}

which will result in this:

    "_id": {
      "year": 2015,
      "month": 10,
      "day": 5
    },
    types: ["impression", "conversion"]
    counters: [10, 5]

there might be a more elegant or faster (with 1 group) way to do this though, i am not sure.