Combine multiple groups in an aggregation in mongodb

问题

If I have a collection like this:

{
    "store" : "XYZ",
    "total" : 100
},
{
    "store" : "XYZ",
    "total" : 200
},
{
    "store" : "ABC",
    "total" : 300
},
{
    "store" : "ABC",
    "total" : 400
}

I can get the $sum of orders in the collection by aggregation:

db.invoices.aggregate([{$group: { _id: null, total: { $sum: "$total"}}}])

{
    "result": [{
            "_id": null,
            "total": 1000
        }
    ],
    "ok": 1
}

And I can get the $sum of orders grouped by store:

db.invoices.aggregate([{$group: { _id: "$store", total: { $sum: "$total"}}}])

{
    "result": [{
            "_id": "ABC",
            "total": 700
        }, {
            "_id": "XYZ",
            "total": 300
        }
    ],
    "ok": 1
}

But how can I do this in one query?

回答1:

You could aggregate as below:

• $group by the store field, calculate the subtotal.

• $project a field doc to keep the subtotal group in tact, during the next group.

• $group by null and accumulate the net total.

Code:

db.invoices.aggregate([{
            $group: {
                "_id": "$store",
                "subtotal": {
                    $sum: "$total"
                }
            }
        }, {
            $project: {
                "doc": {
                    "_id": "$_id",
                    "total": "$subtotal"
                }
            }
        }, {
            $group: {
                "_id": null,
                "total": {
                    $sum: "$doc.total"
                },
                "result": {
                    $push: "$doc"
                }
            }
        }, {
            $project: {
                "result": 1,
                "_id": 0,
                "total": 1
            }
        }
    ])

Output:

{
    "total": 1000,
    "result": [{
            "_id": "ABC",
            "total": 700
        }, {
            "_id": "XYZ",
            "total": 300
        }
    ]
}

回答2:

Another approach would be using the $facet aggregation stage.

• $facet allows you to do multiple nested sub-aggregations within your main aggregation.
• Each sub-aggregation has its own pipeline.
• For each result of a sub-aggregation we define another field.

Like this, for example:

db.invoices.aggregate([
    {
        $facet: {
            total: [
                {
                    $group: {
                        _id: null,
                        total: { $sum: "$total"}
                    }
                }
            ],
            store_totals: [
                {
                    $group: {
                        _id: "$store",
                        total: { $sum: "$total"}
                    }
                }
            ]
        }
    },{
        $unwind: "$total"
    },{
        $project: {
            _id: 0,
            total: "$total.total",
            store_totals: "$store_totals"
        }
    }
]

@BatScream wrote, that an $unwind stage might be costly. However we're unwinding an array of length 1 here. So I'm curious which approach is more efficient under which circumstances. If someone can compare those with console.time(), I'd be happy to include the results.

---

Output

Should be the same as in the accepted answer.

来源：https://stackoverflow.com/questions/28420631/combine-multiple-groups-in-an-aggregation-in-mongodb