count multiple distinct fields by group with Mongo

Question

I have a data set looks as

{"BrandId":"a","SessionId":100,"UserName":"tom"}
{"BrandId":"a","SessionId":200,"UserName":"tom"}
{"BrandId":"b","SessionId":300,"UserName":"mike"}

I would like to count distinct session and username group by brandid, the sample sql is like:

select brandid,count_distinct(sessionid),count_distinct(username)
from data
group by brandid

I tried to write Mongo DB, my current code is as following and it does not work. Is there anyway to make it work?

db.logs.aggregate([ 
    {$group:{
        _id:{brand:"$BrandId",user:"$UserName",session:"$SessionId"},  
        count:{$sum:1}}}, 
    {$group:{
        _id:"$_id.brand",
        users:{$sum:"$_id.user"},
        sessions:{$sum:"$_id.session"}
    }}
])

for the certain example, the expected count is

{"BrandId:"a","countSession":2,"countUser":1}
{"BrandId:"b","countSession":1,"countUser":1}

if you know SQL, the expect result is as same as the SQL I mentioned.

Answer 1

You can do this by using $addToSet to accumulate the distinct set of SessionId and UserName values during the $group, and then adding a $project stage to your pipeline that uses the $size operator to get the size of each set:

db.logs.aggregate([
    {$group: {
        _id: '$BrandId', 
        sessionIds: {$addToSet: '$SessionId'}, 
        userNames: {$addToSet: '$UserName'}
    }},
    {$project: {
        _id: 0, 
        BrandId: '$_id', 
        countSession: {$size: '$sessionIds'}, 
        countUser: {$size: '$userNames'}
    }}
])

Result:

{
    "BrandId" : "b",
    "countSession" : 1,
    "countUser" : 1
}, 
{
    "BrandId" : "a",
    "countSession" : 2,
    "countUser" : 1
}