CODEWITHSUNDEEP
mongodbmongodb-query
raven
raven

Reputation: 2431

How to group subsequent elements in mongo?

How can I group subsequent elements by a property in mongo. This is my mongo document (Events) and I want to be able to group them sub sequentially by type.

{
    "_id" : ObjectId("5d1b68d708f3870049d9cc37"),
    "controllerId" : "80058c2b-9525-4f7f-8e26-faea4ad92b15",
    "id" : "76800453-e72e-410b-accc-cf47cd2773a1",
    "type" : "controller_connection_status",
    "timestamp" : 1562077399832.0,

}

/* 2 */
{
    "_id" : ObjectId("5d1b68db08f3870049d9cc39"),
    "controllerId" : "80058c2b-9525-4f7f-8e26-faea4ad92b15",
    "id" : "fabd883c-6971-4977-b3fc-31679c2b85dd",
    "type" : "controller_connection_status",
    "timestamp" : 1562077402916.0,
}

/* 3 */
{
    "_id" : ObjectId("5d1b68db08f3870049d9cc3a"),
    "controllerId" : "80058c2b-9525-4f7f-8e26-faea4ad92b15",
    "siteId" : "226168be-866c-11e8-adc0-fa7ae01bbebc",
    "id" : "98decbea-8288-4df5-807d-14e90f929df2",
    "type" : "controller_added",
    "timestamp" : 1562077402920.0,
}

/* 4 */
{
    "_id" : ObjectId("5d1b690908f3870049d9cc3c"),
    "controllerId" : "80058c2b-9525-4f7f-8e26-faea4ad92b15",
    "id" : "b5ad6199-8805-43fd-bd7e-80f0410e744a",
    "type" : "controller_connection_status",
    "timestamp" : 1562077449904.0,
}

/* 5 */
{
    "_id" : ObjectId("5d1b690d08f3870049d9cc3d"),
    "controllerId" : "80058c2b-9525-4f7f-8e26-faea4ad92b15",
    "id" : "276ea325-0eec-47a2-8e0e-3805ed34b80b",
    "type" : "controller_error",
    "timestamp" : 1562077452975.0,
}

/* 6 */
{
    "_id" : ObjectId("5d1b694508f3870049d9cc3f"),
    "controllerId" : "80058c2b-9525-4f7f-8e26-faea4ad92b15",
    "id" : "03ce803b-6b2e-49fe-8f0d-4feee44251e9",
    "type" : "controller_error",
    "timestamp" : 1562077509904.0,
}

/* 7 */
{
    "_id" : ObjectId("5d1b694908f3870049d9cc41"),
    "controllerId" : "80058c2b-9525-4f7f-8e26-faea4ad92b15",
    "id" : "b144a04f-8201-4945-b2c4-faef5b41866e",
    "type" : "controller_connection_status",
    "timestamp" : 1562077512974.0,
}

/* 8 */
{
    "_id" : ObjectId("5d1b698208f3870049d9cc42"),
    "controllerId" : "80058c2b-9525-4f7f-8e26-faea4ad92b15",
    "id" : "235874f3-c017-4ea8-abaf-8c5edf1b317a",
    "type" : "controller_connection_status",
    "timestamp" : 1562077569903.0,
}

/* 9 */
{
    "_id" : ObjectId("5d1b698508f3870049d9cc43"),
    "controllerId" : "80058c2b-9525-4f7f-8e26-faea4ad92b15",
    "id" : "4fb3706f-d195-4ded-87b9-8482c712825c",
    "type" : "controller_connection_status",
    "timestamp" : 1562077572973.0,
    "createdAt" : ISODate("2019-07-02T14:26:13.120Z"),
    "updatedAt" : ISODate("2019-07-02T14:26:13.120Z"),
    "__v" : 0
}

/* 10 */
{
    "_id" : ObjectId("5d1b69bd08f3870049d9cc45"),
    "controllerId" : "80058c2b-9525-4f7f-8e26-faea4ad92b15",
    "id" : "e743baef-1701-436a-baf2-8367a0917c81",
    "type" : "controller_removed",
    "timestamp" : 1562077629903.0,
}

My desired output:

timestamp         type                                               count
(last timestamp)  controller_connection_status                         2
 --               controller_added                                     1
 --               controller_connection_status                         1
 --               controller_error                                     2
 --               controller_connection_status                         3
 --               controller_removed                                   1

What I've tried so far: 

db.getCollection('events').aggregate([
      {
        '$match': {
          'controllerId': '80058c2b-9525-4f7f-8e26-faea4ad92b15'
        }
      },
      {
        '$group': {
          '_id': '$type',
          'type': {
            '$first': '$type'
          },
          'timestamp': {
            '$last': '$timestamp'
          },
          'count': {
            '$sum': 1,
          }
        }
      },
      {
        '$sort': {
          'timestamp': -1
        }
      }
])

My output:

timestamp         type                                               count
(last timestamp)  controller_connection_status                         6
 --               controller_added                                     1
 --               controller_error                                     2
 --               controller_removed                                   1

Upvotes: 4

Views: 194

Answers (1)

Alex Blex
Alex Blex

Reputation: 37048

You can use $graphLookup to group sequential documents into array. It requires a collection to lookup from and in your case it can be a view.

The view aggregates documents in previous-next pairs using $zip operator:

db.createView("events-view", "original_collection", [
    { $sort: { timestamp: 1 } },
    { $group: { _id: null, docs: { $push: "$$ROOT" } } },
    { $project: {
        pair: { $zip: {
            inputs:[ { $concatArrays: [ [false], "$docs" ]} , "$docs" ]            
        } }
    } },
    { $unwind: "$pair" },
    { $project: {
        prev: { $arrayElemAt: [ "$pair", 0 ] },
        next: { $arrayElemAt: [ "$pair", 1 ] }
    } },
    { $project: {
        _id: "$prev._id",
        prev: 1,
        next: 1,
        sameType: { $eq: ["$prev.type", "$next.type"] } 
    } },
]);

It should look like following: 

{
    "prev" : false,
    "next" : {
        "_id" : ObjectId("5d1b68d708f3870049d9cc37"),
        "controllerId" : "80058c2b-9525-4f7f-8e26-faea4ad92b15",
        "id" : "76800453-e72e-410b-accc-cf47cd2773a1",
        "type" : "controller_connection_status",
        "timestamp" : 1562077399832.0
    },
    "sameType" : false
},
{
    "prev" : {
        "_id" : ObjectId("5d1b68d708f3870049d9cc37"),
        "controllerId" : "80058c2b-9525-4f7f-8e26-faea4ad92b15",
        "id" : "76800453-e72e-410b-accc-cf47cd2773a1",
        "type" : "controller_connection_status",
        "timestamp" : 1562077399832.0
    },
    "next" : {
        "_id" : ObjectId("5d1b68db08f3870049d9cc39"),
        "controllerId" : "80058c2b-9525-4f7f-8e26-faea4ad92b15",
        "id" : "fabd883c-6971-4977-b3fc-31679c2b85dd",
        "type" : "controller_connection_status",
        "timestamp" : 1562077402916.0
    },
    "_id" : ObjectId("5d1b68d708f3870049d9cc37"),
    "sameType" : true
},
{
    "prev" : {
        "_id" : ObjectId("5d1b68db08f3870049d9cc39"),
        "controllerId" : "80058c2b-9525-4f7f-8e26-faea4ad92b15",
        "id" : "fabd883c-6971-4977-b3fc-31679c2b85dd",
        "type" : "controller_connection_status",
        "timestamp" : 1562077402916.0
    },
    "next" : {
        "_id" : ObjectId("5d1b68db08f3870049d9cc3a"),
        "controllerId" : "80058c2b-9525-4f7f-8e26-faea4ad92b15",
        "siteId" : "226168be-866c-11e8-adc0-fa7ae01bbebc",
        "id" : "98decbea-8288-4df5-807d-14e90f929df2",
        "type" : "controller_added",
        "timestamp" : 1562077402920.0
    },
    "_id" : ObjectId("5d1b68db08f3870049d9cc39"),
    "sameType" : false
},
etc...

Then you can query the view grouping documents by type and latest timestamp until "sameType" condition holds. Longest chain of the documents is the count you are looking for:

db.getCollection("events-view").aggregate([
   { $graphLookup: {
       from: "events-view",
       startWith: "$next._id",
       connectFromField: "next._id",
       connectToField: "_id",
       restrictSearchWithMatch: { "sameType": true },
       as: "chain"
   } },
   { $project: {
       _id: "$next._id",
       type: "$next.type",
       chain: { $concatArrays: [ [{ next: "$next" }], "$chain" ] }
   } },
   { $addFields: {
       chainLength: { $size: "$chain" },
       timestamp: { $max: { $map: {
           input: "$chain",
           in: "$$this.next.timestamp"
       } } }
   } },
   { $group: {
       _id: {type: "$type", timestamp: "$timestamp"},
       count: {$max: "$chainLength"}
   } },
   { $sort: { "_id.timestamp": 1 } }, 
   { $project: {
       _id: 0,
       timestamp: "$_id.timestamp",
       type: "$_id.type",
       count: 1
   } }
])

It should be mentioned that the query will be quite slow. The longer chains the worse performance. Also keep in mind that $graphLookup stage must stay within the 100 megabyte memory limit. For larger collections you should set allowDiskUse option to true.

Upvotes: 1

Related Questions