Aggregating only top x nodes

Users Dgraph
1

Is any way to do a groupby of a predicate but only for the top x nodes on each group.

For example if I have 2 nodes: Donations and State

The Donations node has an amount attribute.

I would like to query the sum of the top 10 donations per state.

Any help will be appreciated.

Thanks,
Marcelo

2

Yes,

you can use multiple blocks: https://docs.dgraph.io/query-language#multiple-query-blocks

With Var Block: https://docs.dgraph.io/query-language/#var-blocks

with Variables: https://docs.dgraph.io/query-language/#query-variables

Aggregation itself: https://docs.dgraph.io/query-language/#max

and Math https://docs.dgraph.io/query-language/#math-on-value-variables

3

I have been working with all those concepts you outline but I can’t figure how to pick the top highest values for each group. Once the aggregation is done, you can’t do any filtering on it. So I guess I will have to filter before that but I want to do filtering with grouping at the same time and that does not seem to work.

Is any way you can provide a sample.

Thanks,
Marcelo

4

have you tried these

https://docs.dgraph.io/query-language/#pagination

https://docs.dgraph.io/query-language/#sorting

5

Yes. I have done sorting of all the donations but I have to groupby state but I don’t know how to indicate to group only the top 10 of each group.

Once the aggregation is done, I don’t see a way to filter the components making up the aggregation.

Thanks,
Marcelo

6

when you did groupby and you can also add sorting too, this can get top k group, if you want to get top k in each group, you need to use sort again in each group

7

That is my problem. Once you do the aggregation (grouping), you can’t do any other filtering on it.

Thanks

8

yeah, it might concern about you real data, and then think about the query

9

i tried this, it works. hope it helps.

  1. mutate my data like:

     _:class1 <name> "class1" .
 _:class2 <name> "class2" .

 _:st_sype1 <name> "strudent_type_one" .
 _:st_sype2 <name> "strudent_type_two" .

 _:st1 <name> "student1" .
 _:st1 <type> _:st_sype1 .
 _:st1 <grade> "3.8" . 
 _:st2 <name> "student2" .
 _:st2 <type> _:st_sype1 .
 _:st2 <grade> "4.8" .
 _:st3 <name> "student3" .
 _:st3 <type> _:st_sype1 .
 _:st3 <grade> "5.8" .

 _:st4 <name> "student4" .
 _:st4 <type> _:st_sype2 .
 _:st4 <grade> "6.8" .
 _:st5 <name> "student5" .
 _:st5 <type> _:st_sype2 .
 _:st5 <grade> "7.8" .
 _:st6 <name> "student6" .
 _:st6 <type> _:st_sype2 .
 _:st6 <grade> "8.8" .	

 _:class1 <student>  _:st1 .
 _:class1 <student>  _:st2 .
 _:class1 <student>  _:st3 .

 _:class2 <student>  _:st4 .
 _:class2 <student>  _:st5 .
 _:class2 <student>  _:st6 .

  2. groupby type and get top k student in each group

{
  var(func:uid(0x7, 0x2)) {
    student @groupby(type) {
      a as count(uid)
    }
  }
  
  byTypequery(func:uid(a)) {
    number : val(a)
    name
    ~type (orderdesc: grade, first:2) {
      name
      grade
      type {
        name
      }
   	}
  }
}
  1. result of this query
{
  "data": {
    "byTypequery": [
      {
        "number": 3,
        "name": "strudent_type_one",
        "~type": [
          {
            "name": "student3",
            "grade": 5.8,
            "type": [
              {
                "name": "strudent_type_one"
              }
            ]
          },
          {
            "name": "student2",
            "grade": 4.8,
            "type": [
              {
                "name": "strudent_type_one"
              }
            ]
          }
        ]
      },
      {
        "number": 3,
        "name": "strudent_type_two",
        "~type": [
          {
            "name": "student6",
            "grade": 8.8,
            "type": [
              {
                "name": "strudent_type_two"
              }
            ]
          },
          {
            "name": "student5",
            "grade": 7.8,
            "type": [
              {
                "name": "strudent_type_two"
              }
            ]
          }
        ]
      }
    ]
  },
  "extensions": {
    "server_latency": {
      "parsing_ns": 96397,
      "processing_ns": 4301956,
      "encoding_ns": 1172687
    },
    "txn": {
      "start_ts": 38,
      "lin_read": {
        "ids": {
          "1": 13
        }
      }
    }
  }
}

PS: class1’s uid is 0x7 and class2’s 0x2 . i am sure you have noticed this.

10

@shanghai-Jerry Thanks. I will give a try.