MongoDB order by a sum on a subset

Question

I have the following collection:

error_reports
[    
    {
       "_id":{
          "$oid":"5184de1261"
       },
       "date":"29/04/2013",
       "errors":[
          {
             "_id":"10",
             "failures":2,
             "alerts":1,
          },
          {
             "_id":"11",
             "failures":7,
             "alerts":4,
          }
       ]
    },
    {
       "_id":{
          "$oid":"5184de1262"
       },
       "date":"30/04/2013",
       "errors":[
          {
             "_id":"15",
             "failures":3,
             "alerts":2,
          },
          {
             "_id":"16",
             "failures":9,
             "alerts":1,
          }
       ]
    }
]

Is it possible to retrieve the list of documents with failures and alerts sum sorted by failures in descending order? I am new to mongodb, I have been searching for 2 days but I can't figure out what is the proper query...

I tried something like this :

db.error_reports.aggregate(
    { $sort : { failures: -1} },
    { $group:
        { _id: "$_id",
        failures: { "$sum": "$errors.failures" }
        }
    }
);

But it didn't work, I think it is because of the $sum: $errors.failures thing, I would like to sum this attribute on every item of the day_hours subcollection but I don't know of to do this in a query...

Answer 1

You were very close with your attempt. The only thing missing is the $unwind aggregation operator. $unwind basically splits each document out based on a sub-document. So before you group the failures and alerts, you unwind the errors, like so:

db.error_reports.aggregate(
  { $unwind : '$errors' },
  { $group : {
    _id : '$_id',
    'failures' : { $sum : '$errors.failures' },
    'alerts' : { $sum : '$errors.alerts' }
  } },
  { $sort : { 'failures': -1 } }
);

Which gives you the follow result:

{
    "result" : [
        {
            "_id" : ObjectId("5184de1262"),
            "failures" : 12,
            "alerts" : 3
        },
        {
            "_id" : ObjectId("5184de1261"),
            "failures" : 9,
            "alerts" : 5
        }
    ],
    "ok" : 1
}