Aggregate to collect data in array

[masoud_darvishi]

hello
this is my data:

[
   {
      type:  "a",  // value is "a" or "b"
      number:  15  // number between 1 to 100
  },
  {
      type:  "b",
      number:  56
  },
  {
      type:  "a",
      number:  40
  },
  {
      type:  "b",
      number:  2
  }
]

is there aggregation that is can group by "type" field and collect "number" values in array? i want aggregation result be something like this :

result: [
    {
       key:  "a",
       doc_count: 2
       numbers: {
            value: [15,40]
       }
     },
     {
        key:  "b",
        doc_count: 2
        numbers: {
             value: [56,2]
        }
     },
]

[dadoonet (David Pilato)]

If you want the unique values of the numbers, you can use this:

DELETE test
POST test/_doc
{ "type":  "a", "number":  15 }
POST test/_doc
{ "type":  "a", "number":  15 }
POST test/_doc
{ "type":  "b", "number":  56 }
POST test/_doc
{ "type":  "a", "number":  40 }
POST test/_doc
{ "type":  "b", "number":  2 }
GET test/_search
{
  "size": 0,
  "aggs": {
    "by_type": {
      "terms": {
        "field": "type.keyword"
      },
      "aggs": {
        "by_number": {
          "terms": {
            "field": "number"
          }
        }
      }
    }
  }
}

This gives:

{
  "took": 3,
  "timed_out": false,
  "_shards": {
    "total": 1,
    "successful": 1,
    "skipped": 0,
    "failed": 0
  },
  "hits": {
    "total": {
      "value": 5,
      "relation": "eq"
    },
    "max_score": null,
    "hits": []
  },
  "aggregations": {
    "by_type": {
      "doc_count_error_upper_bound": 0,
      "sum_other_doc_count": 0,
      "buckets": [
        {
          "key": "a",
          "doc_count": 3,
          "by_number": {
            "doc_count_error_upper_bound": 0,
            "sum_other_doc_count": 0,
            "buckets": [
              {
                "key": 15,
                "doc_count": 2
              },
              {
                "key": 40,
                "doc_count": 1
              }
            ]
          }
        },
        {
          "key": "b",
          "doc_count": 2,
          "by_number": {
            "doc_count_error_upper_bound": 0,
            "sum_other_doc_count": 0,
            "buckets": [
              {
                "key": 2,
                "doc_count": 1
              },
              {
                "key": 56,
                "doc_count": 1
              }
            ]
          }
        }
      ]
    }
  }
}

But if you want to get all the values, even though some of them are already collected, you could do that instead:

GET test/_search?filter_path=aggregations.by_type.buckets.key,aggregations.by_type.buckets.all_numbers.hits.hits.fields.number
{
  "size": 0,
  "aggs": {
    "by_type": {
      "terms": {
        "field": "type.keyword"
      },
      "aggs": {
        "all_numbers": {
          "top_hits": {
            "size": 1000,
            "docvalue_fields": ["number"],
            "_source": false
          }
        }
      }
    }
  }
}

{
  "aggregations": {
    "by_type": {
      "buckets": [
        {
          "key": "a",
          "all_numbers": {
            "hits": {
              "hits": [
                {
                  "fields": {
                    "number": [
                      15
                    ]
                  }
                },
                {
                  "fields": {
                    "number": [
                      15
                    ]
                  }
                },
                {
                  "fields": {
                    "number": [
                      40
                    ]
                  }
                }
              ]
            }
          }
        },
        {
          "key": "b",
          "all_numbers": {
            "hits": {
              "hits": [
                {
                  "fields": {
                    "number": [
                      56
                    ]
                  }
                },
                {
                  "fields": {
                    "number": [
                      2
                    ]
                  }
                }
              ]
            }
          }
        }
      ]
    }
  }
}