Group count with MongoDB using aggregation framework

Let's say my MongoDB schema looks like this:

{car_id: "...", owner_id: "..."}

This is a many-to-many relationship. For example, the data might look like this:

+-----+----------+--------+
| _id | owner_id | car_id |
+-----+----------+--------+
|   1 |        1 |      1 |
|   2 |        1 |      2 |
|   3 |        1 |      3 |
|   4 |        2 |      1 |
|   5 |        2 |      2 |
|   6 |        3 |      4 |
|   7 |        3 |      5 |
|   8 |        3 |      6 |
|   9 |        3 |      7 |
|  10 |        1 |      1 | <-- not unique
+-----+----------+--------+

I want to get the number of cars owned by each owner. In SQL, this might look like:

SELECT owner_id, COUNT(*) AS cars_owned
FROM (SELECT owner_id FROM car_owners GROUP BY owner_id, car_id) AS t
GROUP BY owner_id;

In this case, the result would look like this:

+----------+------------+
| owner_id | cars_owned |
+----------+------------+
|        1 |          3 |
|        2 |          2 |
|        3 |          4 |
+----------+------------+

How can I accomplish this same thing using MongoDB using the aggregation framework?

Solution 1:

To accommodate the potential duplicates, you need to use two $group operations:

db.test.aggregate([
    { $group: {
        _id: { owner_id: '$owner_id', car_id: '$car_id' }
    }},
    { $group: {
        _id: '$_id.owner_id',
        cars_owned: { $sum: 1 }
    }},
    { $project: {
        _id: 0,
        owner_id: '$_id',
        cars_owned: 1
    }}]
    , function(err, result){
        console.log(result);
    }
);

Gives a result with a format of:

[ { cars_owned: 2, owner_id: 10 },
  { cars_owned: 1, owner_id: 11 } ]