ElasticSearch 聚合嵌套字段

Question

我有一个带有映射的产品存储库：

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

规格条目是产品规格（例如：Fork：Air），其中 Fork 是名称，Air 是值。还有规格ID、规格值ID、规格说明。

我需要得到这样的聚合结果：

[
...
{
  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
  }, ]
},
...
]

规格和值应按数量排序。

我需要使用什么类型的聚合？

Answer 1

您想使用nested 聚合（因为spec_entries 具有nested 类型，然后是spec_entries.name 的一个terms 聚合和一个top_hits 子聚合以获得顶部嵌套的spec_entries . 应该这样做：

{
  "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"
              }
            }
          }
        }
      }
    }
  }
}

这将产生非常接近您期望的结果：

{
  ...
  "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
                }
              } ]
            }
          }
        } ]
      }
    }
  }
}

请注意，使用 top_hits 聚合作为 nested 聚合的子聚合仅适用于 from ES 1.5 onward