如何訪問 elasticsearch 中子聚合中的 date_histogram 鍵字段？

Question

我想對date_histogram生成的桶響應應用一些過濾器，該過濾器取決於date_histogram output 桶的鍵。

假設我有以下數據

{
   "entryTime":"",
   "soldTime:""
}

彈性查詢是這樣的

{
  "aggs": {
    "date": {
      "date_histogram": {
        "field": "entryTime",
        "interval": "month",
        "keyed": true
      },
      "aggs": {
        "filter_try": {
          "filter": {
            "bool": {
              "must": [
                {
                  "range": {
                    "entryTime": {
                      "lte": 1588840533000
                    }
                  }
                },
                {
                  "bool": {
                    "should": [
                      {
                        "bool": {
                          "must": [
                            {
                              "exists": {
                                "field": "soldTime"
                              }
                            },
                            {
                              "range": {
                                "soldTime": {
                                  "gt": 1588840533000
                                }
                              }
                            }
                          ]
                        }
                      },
                      {
                        "bool": {
                          "must_not": [
                            {
                              "exists": {
                                "field": "soldTime"
                              }
                            }
                          ]
                        }
                      }
                    ]
                  }
                }
              ]
            }
          }
        }
      }
    }
  }
}

因此，在該布爾查詢中，我想在兩個范圍子句中使用date_histogram聚合為特定存儲桶生成的日期，而不是硬編碼的紀元時間。

即使我們可以使用腳本訪問也可以。

為了進一步說明，這是 boolean 查詢，在查詢中想用date_histogram桶鍵替換這個"DATE" 。

# (entryTime < DATE) 
# AND 
# (
#    (soldTime != null AND soldTime > DATE) 
#          OR 
#      (soldTime == NULL)
#  )

考慮以下我擁有的 10 個文檔：

"hits" : [
      {
        "_index" : "vi_test",
        "_type" : "_doc",
        "_id" : "1",
        "_score" : 1.0,
        "_source" : {
          "deaerId" : "4",
          "entryTime" : "1577869200000",
          "soldTime" : "1578646800000"
        }
      },
      {
        "_index" : "vi_test",
        "_type" : "_doc",
        "_id" : "2",
        "_score" : 1.0,
        "_source" : {
          "deaerId" : "4",
          "entryTime" : "1578214800000"
        }
      },
      {
        "_index" : "vi_test",
        "_type" : "_doc",
        "_id" : "3",
        "_score" : 1.0,
        "_source" : {
          "deaerId" : "4",
          "entryTime" : "1578560400000",
          "soldTime" : "1579942800000"
        }
      },
      {
        "_index" : "vi_test",
        "_type" : "_doc",
        "_id" : "4",
        "_score" : 1.0,
        "_source" : {
          "deaerId" : "4",
          "entryTime" : "1579683600000",
          "soldTime" : "1581325200000"
        }
      },
      {
        "_index" : "vi_test",
        "_type" : "_doc",
        "_id" : "5",
        "_score" : 1.0,
        "_source" : {
          "deaerId" : "4",
          "entryTime" : "1580893200000"
        }
      },
      {
        "_index" : "vi_test",
        "_type" : "_doc",
        "_id" : "6",
        "_score" : 1.0,
        "_source" : {
          "deaerId" : "4",
          "entryTime" : "1582189200000",
          "soldTime" : "1582362000000"
        }
      },
      {
        "_index" : "vi_test",
        "_type" : "_doc",
        "_id" : "7",
        "_score" : 1.0,
        "_source" : {
          "deaerId" : "4",
          "entryTime" : "1582621200000",
          "soldTime" : "1584349200000"
        }
      },
      {
        "_index" : "vi_test",
        "_type" : "_doc",
        "_id" : "8",
        "_score" : 1.0,
        "_source" : {
          "deaerId" : "4",
          "entryTime" : "1583053200000",
          "soldTime" : "1583830800000"
        }
      },
      {
        "_index" : "vi_test",
        "_type" : "_doc",
        "_id" : "9",
        "_score" : 1.0,
        "_source" : {
          "deaerId" : "4",
          "entryTime" : "1584262800000"
        }
      },
      {
        "_index" : "vi_test",
        "_type" : "_doc",
        "_id" : "10",
        "_score" : 1.0,
        "_source" : {
          "deaerId" : "4",
          "entryTime" : "1585472400000"
        }
      }
    ]

現在 2020 年 1 月結束的紀元是 -> 1580515199000

因此，如果我申請上述布爾查詢，

將獲得 output 作為

"hits" : [
      {
        "_index" : "vi_test",
        "_type" : "_doc",
        "_id" : "4",
        "_score" : 3.0,
        "_source" : {
          "deaerId" : "4",
          "entryTime" : "1579683600000",
          "soldTime" : "1581325200000"
        }
      },
      {
        "_index" : "vi_test",
        "_type" : "_doc",
        "_id" : "2",
        "_score" : 1.0,
        "_source" : {
          "deaerId" : "4",
          "entryTime" : "1578214800000"
        }
      }
    ]

由於 ID 為 4 的文檔滿足(soldTime != null AND soldTime > DATE)並且 ID 為 2 的文檔滿足 OR 部分的(soldTime == null)條件。

現在對於相同的布爾請求如果我使用 2020 年二月結束的日期 -> 1583020799000 ，將獲得如下點擊

"hits" : [
      {
        "_index" : "vi_test",
        "_type" : "_doc",
        "_id" : "7",
        "_score" : 3.0,
        "_source" : {
          "deaerId" : "4",
          "entryTime" : "1582621200000",
          "soldTime" : "1584349200000"
        }
      },
      {
        "_index" : "vi_test",
        "_type" : "_doc",
        "_id" : "2",
        "_score" : 1.0,
        "_source" : {
          "deaerId" : "4",
          "entryTime" : "1578214800000"
        }
      },
      {
        "_index" : "vi_test",
        "_type" : "_doc",
        "_id" : "5",
        "_score" : 1.0,
        "_source" : {
          "deaerId" : "4",
          "entryTime" : "1580893200000"
        }
      }
    ]

• ID 7：2 月進入，但 3 月售出，因此有 2020 年 2 月的庫存
• ID 2：1月入庫，未售出即有庫存
• ID 5：2月入庫，未售出即有庫存

現在同樣的數據需要全年每個月底來plot的走勢。

謝謝

Answer 1

我找不到使用普通查詢的方法，因為父聚合鍵在子聚合中不可用。 我為此編寫了一個腳本，它選擇 soldTime 為 null 或不與 entryTime 同月的文檔

詢問：

{
  "query": {
    "script": {
      "script": """
         ZonedDateTime entry;
         ZonedDateTime sold;
         if(doc['entryTime'].size()>0)
         {
           entry= doc['entryTime'].value;
         }
         if(doc['soldTime'].size()>0) 
         {
           sold = doc['soldTime'].value;
         }
         if(sold==null || ( entry.getMonthValue()!==sold.getMonthValue()|| entry.getYear()!==sold.getYear()))
         {
           return true;
         }
         else false;
"""
    }
  },
  "size": 10,
  "aggs": {
    "monthly_trend": {
      "date_histogram": {
        "field": "entryTime",
        "interval": "month"
      },
      "aggs": {
        "docs": {
          "top_hits": {
            "size": 10
          }
        }
      }
    }
  }
}

結果：

    "hits" : [
      {
        "_index" : "index22",
        "_type" : "_doc",
        "_id" : "55Kv83EB8a54AbXfngYU",
        "_score" : 1.0,
        "_source" : {
          "deaerId" : "4",
          "entryTime" : "1578214800000"
        }
      }
    ]
  },
  "aggregations" : {
    "monthly_trend" : {
      "buckets" : [
        {
          "key_as_string" : "2020-01-01T00:00:00.000Z",
          "key" : 1577836800000,
          "doc_count" : 1,
          "docs" : {
            "hits" : {
              "total" : {
                "value" : 1,
                "relation" : "eq"
              },
              "max_score" : 1.0,
              "hits" : [
                {
                  "_index" : "index22",
                  "_type" : "_doc",
                  "_id" : "55Kv83EB8a54AbXfngYU",
                  "_score" : 1.0,
                  "_source" : {
                    "deaerId" : "4",
                    "entryTime" : "1578214800000"
                  }
                }
              ]
            }
          }
        }
      ]
    }
  }