Dynamic keys after $group by

问题

I have following collection

{
    "_id" : ObjectId("5b18d14cbc83fd271b6a157c"),
    "status" : "pending",
    "description" : "You have to complete the challenge...",
}
{
    "_id" : ObjectId("5b18d31a27a37696ec8b5773"),
    "status" : "completed",
    "description" : "completed...",
}
{
    "_id" : ObjectId("5b18d31a27a37696ec8b5775"),
    "status" : "pending",
    "description" : "pending...",
}
{
    "_id" : ObjectId("5b18d31a27a37696ec8b5776"),
    "status" : "inProgress",
    "description" : "inProgress...",
}

I need to group by status and get all the keys dynamically which are in status

[
  {
    "completed": [
      {
        "_id": "5b18d31a27a37696ec8b5773",
        "status": "completed",
        "description": "completed..."
      }
    ]
  },
  {
    "pending": [
      {
        "_id": "5b18d14cbc83fd271b6a157c",
        "status": "pending",
        "description": "You have to complete the challenge..."
      },
      {
        "_id": "5b18d31a27a37696ec8b5775",
        "status": "pending",
        "description": "pending..."
      }
    ]
  },
  {
    "inProgress": [
      {
        "_id": "5b18d31a27a37696ec8b5776",
        "status": "inProgress",
        "description": "inProgress..."
      }
    ]
  }
]

回答1:

Not that I think it's a good idea and mostly because I don't see any "aggregation" here at all is that after "grouping" to add to an array you similarly $push all that content into array by the "status" grouping key and then convert into keys of a document in a $replaceRoot with $arrayToObject:

db.collection.aggregate([
  { "$group": {
    "_id": "$status",
    "data": { "$push": "$$ROOT" }
  }},
  { "$group": {
    "_id": null,
    "data": {
      "$push": {
        "k": "$_id",
        "v": "$data"
      }
    }
  }},
  { "$replaceRoot": {
    "newRoot": { "$arrayToObject": "$data" }
  }}
])

Returns:

{
        "inProgress" : [
                {
                        "_id" : ObjectId("5b18d31a27a37696ec8b5776"),
                        "status" : "inProgress",
                        "description" : "inProgress..."
                }
        ],
        "completed" : [
                {
                        "_id" : ObjectId("5b18d31a27a37696ec8b5773"),
                        "status" : "completed",
                        "description" : "completed..."
                }
        ],
        "pending" : [
                {
                        "_id" : ObjectId("5b18d14cbc83fd271b6a157c"),
                        "status" : "pending",
                        "description" : "You have to complete the challenge..."
                },
                {
                        "_id" : ObjectId("5b18d31a27a37696ec8b5775"),
                        "status" : "pending",
                        "description" : "pending..."
                }
        ]
}

That might be okay IF you actually "aggregated" beforehand, but on any practically sized collection all that is doing is trying force the whole collection into a single document, and that's likely to break the BSON Limit of 16MB, so I just would not recommend even attempting this without "grouping" something else before this step.

Frankly, the same following code does the same thing, and without aggregation tricks and no BSON limit problem:

var obj = {};

// Using forEach as a premise for representing "any" cursor iteration form
db.collection.find().forEach(d => {
  if (!obj.hasOwnProperty(d.status))
    obj[d.status] = [];
  obj[d.status].push(d);
})

printjson(obj);

Or a bit shorter:

var obj = {};

// Using forEach as a premise for representing "any" cursor iteration form
db.collection.find().forEach(d => 
  obj[d.status] = [ 
    ...(obj.hasOwnProperty(d.status)) ? obj[d.status] : [],
    d
  ]
)

printjson(obj);

Aggregations are used for "data reduction" and anything that is simply "reshaping results" without actually reducing the data returned from the server is usually better handled in client code anyway. You're still returning all data no matter what you do, and the client processing of the cursor has considerably less overhead. And NO restrictions.

来源：https://stackoverflow.com/questions/50735135/dynamic-keys-after-group-by