ElasticSearch聚合嵌套字段

Question

I have a product repository with mapping:

settings do
  mapping do
    indexes :name
    indexes :vendor_id,   type: 'integer'
    indexes :category_id, type: 'integer'

    indexes :spec_entries, type: 'nested' do
      indexes :spec_id,     type: 'integer'
      indexes :value_id,    type: 'integer'
      indexes :name,        index: 'no'
      indexes :description, index: 'no'
      indexes :value,       index: 'no'
    end
  end
end

Spec entries is a product specifications (ex: Fork: Air) where Fork is a name and Air is a value. Also there are specification ID, specification value ID, and specification description.

I need to get a aggregations result like this:

[
...
{
  id: 335,
  name: "Fork",
  description: "There are few common types of fork — elastomer, oil and air",
  count: 30,
  values: [{
    id: 645,
    name: "Elastomer",
    count: 17
  }, {
    id: 643,
    name: "Oil",
    count: 10
  }, {
    id: 649,
    name: "Air",
    count: 3
  }, ]
},
...
]

Specs and values should be ordered by count.

What type of aggregation I need to use?

Answer 1

You want to use the nested aggregation (because spec_entries has nested type and then one terms aggregations for the spec_entries.name and a top_hits sub-aggregation to get the top nested spec_entries . Something like this should do:

{
  "query": {
    "match_all": {}
  },
  "aggs": {
    "spec_names": {
      "nested": {
        "path": "spec_entries"
      },
      "aggs": {
        "names": {
          "terms": {
            "field": "spec_entries.name"
          },
          "aggs": {
            "top_entries": {
              "top_hits": {
                "field": "spec_entries.value"
              }
            }
          }
        }
      }
    }
  }
}

And that would yield something very close to what you expect:

{
  ...
  "aggregations" : {
    "spec_names" : {
      "doc_count" : 1,
      "names" : {
        "doc_count_error_upper_bound" : 0,
        "sum_other_doc_count" : 0,
        "buckets" : [ {
          "key" : "fork",     <-------- The spec_entries name
          "doc_count" : 1,
          "top_values" : {
            "hits" : {
              "total" : 1,
              "max_score" : 1.0,
              "hits" : [ {
                "_index" : "tests",
                "_type" : "test1",
                "_id" : "1",
                "_nested" : {
                  "field" : "spec_entries",
                  "offset" : 0
                },
                "_score" : 1.0,
                "_source":{  <-------- For each name, the top spec_entries content (value, id, desc, etc)
                  "name":"Fork",
                  "value":"Air",
                  "description":"desc",
                  "spec_id":1,
                  "value_id":1
                }
              } ]
            }
          }
        } ]
      }
    }
  }
}

Note that using the top_hits aggregation as a sub-aggregation of the nested aggregation will only work from ES 1.5 onward