Aggregate objects' nested array occurrences count - mongodb

Question

I'm trying to aggregate logs in that way, so I can get count of how many times keywords were favorited by particular user. What I came up is following query:
db.a.aggregate([
{$unwind: "$keywords"},
{$group : {_id : {word : "$keywords", user : "$favorited_by"}, count : {$sum : 1}}}
]);
But it produces output:
{ "_id" : { "word" : "another", "user" : "too_creepy" }, "count" : 1 }
{ "_id" : { "word" : "test", "user" : "too_creepy" }, "count" : 2 }
Whilst I want to get something like this:
INPUT
{
_id: ObjectId("5475cf117ccee624583ba94a"),
favorited_by: "too_creepy",
keywords: [
"test"
]
},
{
_id: ObjectId("5475cf117ccee624583ba949"),
favorited_by: "too_creepy",
keywords: [
"test"
]
},
{
_id: ObjectId("5475cf117ccee624583ba949"),
favorited_by: "too_creepy",
keywords: [
"anotherone"
]
},
{
_id: ObjectId("5475cf117ccee624583ba09a"),
favorited_by: "hello_world",
keywords: [
"test"
]
}
OUTPUT
{
favorited_by: "too_creepy",
keywords: [
{keyword: "test", count: 2},
{keyword: "anotherone", count: 1}
]
},
{
favorited_by: "hello_world",
keywords: [
{keyword: "test", count: 1}
]
}
Any ideas how can to write this query if it's even possible?

Answer 1

You can do that by adding a second $group to your pipeline followed up with a final $project to reshape the output a bit:
db.a.aggregate([
{$unwind: "$keywords"},
{$group: {_id: {word: "$keywords", user: "$favorited_by"}, count: {$sum: 1}}},
// Group again on just user, and use $push to assemble an array of their keywords
{$group: {
_id: '$_id.user',
keywords: {$push: {keyword: '$_id.word', count: '$count'}}
}},
// Reshape the output
{$project: {favorited_by: '$_id', keywords: 1, _id: 0}}
]);
Output:
{
"keywords" : [
{
"keyword" : "anotherone",
"count" : 1
},
{
"keyword" : "test",
"count" : 2
}
],
"favorited_by" : "too_creepy"
},
{
"keywords" : [
{
"keyword" : "test",
"count" : 1
}
],
"favorited_by" : "hello_world"
}