elasticsearc使用指南之ES管道聚合(Pipeline Aggregation)

管道聚合處理來自其他聚合而不是文檔集的輸出，將信息添加到輸出樹中。

注：關於腳本聚合目前在本文中暫時不會涉及。

主要有如下兩種管道聚合方式：

parent
sibling

下面一一介紹ES定義的管道聚合。

1、Avg Bucket Aggregation

同級管道聚合，它計算同級聚合中指定度量的平均值。同級聚合必須是多桶聚合，針對的是度量聚合(metric Aggregation)。
示例如下：

{
    "avg_bucket": {
        "buckets_path": "the_sum"  // @1
    }
}

buckets_path：指定聚合的名稱，支持多級嵌套聚合。
其他參數：
gap_policy
當管道聚合遇到不存在的值，有點類似於term等聚合的(missing)時所採取的策略，可選擇值爲：skip、insert_zeros。
skip：此選項將丟失的數據視爲bucket不存在。它將跳過桶並使用下一個可用值繼續計算。
insert_zeros：默認使用0代替。
format
用於格式化聚合桶的輸出(key)。

示例如下：

POST /_search
{
  "size": 0,
  "aggs": {
    "sales_per_month": {                  // @1
           "date_histogram": {
            "field": "date",
            "interval": "month"
      },
      "aggs": {                                    // @2
        "sales": {
          "sum": {
            "field": "price"
          }
        }
      }
    },
    "avg_monthly_sales": {             // @3
      "avg_bucket": {
        "buckets_path": "sales_per_month>sales" 
      }
    }
  }
}

代碼@1：首先定義第一級聚合（按月）直方圖聚合。
代碼@2：定義第二級聚合，在按月聚合的基礎上，對每個月的文檔求sum。
代碼@3：對上面的聚合求平均值。

其返回結果如下：

{
    ... // 省略
   "aggregations": {
      "sales_per_month": {
         "buckets": [
            {
               "key_as_string": "2015/01/01 00:00:00",
               "key": 1420070400000,
               "doc_count": 3,
               "sales": {
                  "value": 550.0
               }
            },
            {
               "key_as_string": "2015/02/01 00:00:00",
               "key": 1422748800000,
               "doc_count": 2,
               "sales": {
                  "value": 60.0
               }
            }
         ]
      },
      "avg_monthly_sales": {   // 這是對二級聚合的結果再進行一次求平均值聚合。
          "value": 328.33333333333333
      }
   }
}

對應的JAVA示例如下：

public static void test_pipeline_avg_buncket_aggregation() {
        RestHighLevelClient client = EsClient.getClient();
        try {
            SearchRequest searchRequest = new SearchRequest();
            searchRequest.indices("aggregations_index02");
            SearchSourceBuilder sourceBuilder = new SearchSourceBuilder();
            AggregationBuilder aggregationBuild = AggregationBuilders.terms("seller_agg")
                                                        .field("sellerId")
                                                        .subAggregation(AggregationBuilders.sum("seller_num_agg")
                                                                            .field("num")
                                                        )
                                                  ;
            sourceBuilder.aggregation(aggregationBuild);
            
            // 添加 avg bucket pipeline
            sourceBuilder.aggregation(new AvgBucketPipelineAggregationBuilder("seller_num_agg_av", "seller_agg>seller_num_agg"));
            sourceBuilder.size(0);
            
            searchRequest.source(sourceBuilder);
            SearchResponse result = client.search(searchRequest, RequestOptions.DEFAULT);
            System.out.println(result);
        } catch (Throwable e) {
            e.printStackTrace();
        } finally {
            EsClient.close(client);
        }
    }

2、Percentiles Bucket Aggregation

同級管道聚合，百分位管道聚合。其JAVA示例如下：

public static void test_Percentiles_buncket_aggregation() {
        RestHighLevelClient client = EsClient.getClient();
        try {
            SearchRequest searchRequest = new SearchRequest();
            searchRequest.indices("aggregations_index02");
            SearchSourceBuilder sourceBuilder = new SearchSourceBuilder();
            AggregationBuilder aggregationBuild = AggregationBuilders.terms("seller_agg")
                                                        .field("sellerId")
                                                        .subAggregation(AggregationBuilders.sum("seller_num_agg")
                                                                            .field("num")
                                                        )
                                                  ;
            sourceBuilder.aggregation(aggregationBuild);
            
            // 添加 avg bucket pipeline
            sourceBuilder.aggregation(new PercentilesBucketPipelineAggregationBuilder("seller_num_agg_av", "seller_agg>seller_num_agg"));
            sourceBuilder.size(0);
            
            searchRequest.source(sourceBuilder);
            SearchResponse result = client.search(searchRequest, RequestOptions.DEFAULT);
            System.out.println(result);
        } catch (Throwable e) {
            e.printStackTrace();
        } finally {
            EsClient.close(client);
        }
    }

其返回值如下：

{
  ...  // 省略其他屬性
    "aggregations":{
        "lterms#seller_agg":{
            "doc_count_error_upper_bound":0,
            "sum_other_doc_count":12,
            "buckets":[
                {
                    "key":45,
                    "doc_count":567,
                    "sum#seller_num_agg":{
                        "value":911
                    }
                },
                {
                    "key":31,
                    "doc_count":324,
                    "sum#seller_num_agg":{
                        "value":353
                    }
                } // 省略其他桶的顯示
            ]
        },
        "percentiles_bucket#seller_num_agg_av":{
            "values":{
                "1.0":5,
                "5.0":5,
                "25.0":10,
                "50.0":20,
                "75.0":290,
                "95.0":911,
                "99.0":911
            }
        }
    }
}

3、Cumulative Sum Aggregation

累積管道聚合，就是就是依次將每個管道的sum聚合進行累加。

其語法(restfull)如下：

{
    "cumulative_sum": {
        "buckets_path": "the_sum"
    }
}

支持的參數說明：

buckets_path
桶聚合名稱，作爲管道聚合的輸入信息。
format
格式化key。

使用示例如下：

POST /sales/_search
{
    "size": 0,
    "aggs" : {
        "sales_per_month" : {
            "date_histogram" : {
                "field" : "date",
                "interval" : "month"
            },
            "aggs": {
                "sales": {
                    "sum": {
                        "field": "price"
                    }
                },
                "cumulative_sales": {
                    "cumulative_sum": {
                        "buckets_path": "sales" 
                    }
                }
            }
        }
    }
}

其返回結果如下：

{
   "took": 11,
   "timed_out": false,
   "_shards": ...,
   "hits": ...,
   "aggregations": {
      "sales_per_month": {
         "buckets": [
            {
               "key_as_string": "2015/01/01 00:00:00",
               "key": 1420070400000,
               "doc_count": 3,
               "sales": {
                  "value": 550.0
               },
               "cumulative_sales": {
                  "value": 550.0
               }
            },
            {
               "key_as_string": "2015/02/01 00:00:00",
               "key": 1422748800000,
               "doc_count": 2,
               "sales": {
                  "value": 60.0
               },
               "cumulative_sales": {
                  "value": 610.0
               }
            },
            {
               "key_as_string": "2015/03/01 00:00:00",
               "key": 1425168000000,
               "doc_count": 2,
               "sales": {
                  "value": 375.0
               },
               "cumulative_sales": {
                  "value": 985.0
               }
            }
         ]
      }
   }
}

從結果可知，cumulative_sales的值等於上一個cumulative_sales + 當前桶的sum聚合。

對應的JAVA示例如下：

{
    "aggregations":{
        "date_histogram#createTime_histogram":{
            "buckets":{
                "2015-12-01 00:00:00":{
                    "key_as_string":"2015-12-01 00:00:00",
                    "key":1448928000000,
                    "doc_count":6,
                    "sum#seller_num_agg":{
                        "value":16
                    },
                    "simple_value#Cumulative_Seller_num_agg":{
                        "value":16
                    }
                },
                "2016-01-01 00:00:00":{
                    "key_as_string":"2016-03-01 00:00:00",
                    "key":1456790400000,
                    "doc_count":10,
                    "sum#seller_num_agg":{
                        "value":11
                    },
                    "simple_value#Cumulative_Seller_num_agg":{
                        "value":31
                    }
                }
                // ... 忽略
            }
        }
    }
}

4、Bucket Sort Aggregation

一種父管道聚合，它對其父多桶聚合的桶進行排序。並可以指定多個排序字段。每個bucket可以根據它的_key、_count或子聚合進行排序。此外，可以設置from和size的參數，以便截斷結果桶。

使用語法如下：

{
    "bucket_sort": {
        "sort": [
            {"sort_field_1": {"order": "asc"}},
            {"sort_field_2": {"order": "desc"}},
            "sort_field_3"
        ],
        "from": 1,
        "size": 3
    }
}

支持的參數說明如下：

sort
定義排序結構。
from
用與對父聚合的桶進行截取，該值之前的所有桶將忽略，也就是不參與排序，默認爲0。
size
返回的桶數。默認爲父聚合的所有桶。
gap_policy
當管道聚合遇到不存在的值，有點類似於term等聚合的(missing)時所採取的策略，可選擇值爲：skip、insert_zeros。
skip：此選項將丟失的數據視爲bucket不存在。它將跳過桶並使用下一個可用值繼續計算。
insert_zeros：默認使用0代替。

官方示例如下：

POST /sales/_search
{
    "size": 0,
    "aggs" : {
        "sales_per_month" : {
            "date_histogram" : {
                "field" : "date",
                "interval" : "month"
            },
            "aggs": {
                "total_sales": {
                    "sum": {
                        "field": "price"
                    }
                },
                "sales_bucket_sort": {
                    "bucket_sort": {
                        "sort": [
                          {"total_sales": {"order": "desc"}}
                        ],
                        "size": 3
                    }
                }
            }
        }
    }
}

對應的JAVA示例如下：

public static void test_bucket_sort_Aggregation() {
        RestHighLevelClient client = EsClient.getClient();
        try {
            
            //構建日期直方圖聚合  時間間隔，示例中按月統計
            DateHistogramInterval interval = new DateHistogramInterval("1M"); 
            SearchRequest searchRequest = new SearchRequest();
            searchRequest.indices("aggregations_index02");
            SearchSourceBuilder sourceBuilder = new SearchSourceBuilder();
            AggregationBuilder aggregationBuild = AggregationBuilders.dateHistogram("createTime_histogram")
                                                                        .field("createTime")
                                                                        .dateHistogramInterval(interval)
                                                                        .keyed(true)
                                                                        .subAggregation(AggregationBuilders.sum("seller_num_agg")
                                                                                .field("num")
                                                                        )
                                                                        .subAggregation(new  BucketSortPipelineAggregationBuilder("seller_num_agg_sort", Arrays.asList(
                                                                                new FieldSortBuilder("seller_num_agg").order(SortOrder.ASC)))
                                                                                .from(0)
                                                                                .size(3))
                                                                        //  BucketSortPipelineAggregationBuilder(String name, List<FieldSortBuilder> sorts)
                                                                        .subAggregation(new CumulativeSumPipelineAggregationBuilder("Cumulative_Seller_num_agg", "seller_num_agg"))
                                                                    //    .format("yyyy-MM-dd") // 對key的格式化
                                                  ;
            sourceBuilder.aggregation(aggregationBuild);
            sourceBuilder.size(0);
            sourceBuilder.query(
                    QueryBuilders.termQuery("sellerId", 24)
            );
            searchRequest.source(sourceBuilder);
            SearchResponse result = client.search(searchRequest, RequestOptions.DEFAULT);
            System.out.println(result);
        } catch (Throwable e) {
            e.printStackTrace();
        } finally {
            EsClient.close(client);
        }
    }

返回值：

{
    "aggregations":{
        "date_histogram#createTime_histogram":{
            "buckets":{
                "2016-04-01 00:00:00":{
                    "key_as_string":"2016-04-01 00:00:00",
                    "key":1459468800000,
                    "doc_count":2,
                    "sum#seller_num_agg":{
                        "value":2
                    },
                    "simple_value#Cumulative_Seller_num_agg":{
                        "value":2
                    }
                },
                "2017-05-01 00:00:00":{
                    "key_as_string":"2017-05-01 00:00:00",
                    "key":1493596800000,
                    "doc_count":3,
                    "sum#seller_num_agg":{
                        "value":3
                    },
                    "simple_value#Cumulative_Seller_num_agg":{
                        "value":5
                    }
                },
                "2017-02-01 00:00:00":{
                    "key_as_string":"2017-02-01 00:00:00",
                    "key":1485907200000,
                    "doc_count":4,
                    "sum#seller_num_agg":{
                        "value":4
                    },
                    "simple_value#Cumulative_Seller_num_agg":{
                        "value":9
                    }
                }
            }
        }
    }

5、Max Bucket Aggregation

與 avg類似。

6、Min Bucket Aggregation

與 avg類似。

7、Sum Bucket Aggregation

與 avg類似。

8、Stats Bucket Aggregation

與 avg類似。

本節詳細介紹了ES Pipeline Aggregation 管道聚合的使用方法，重點介紹了Avg Bucket Aggregation、Percentiles Bucket Aggregation、Cumulative Sum Aggregation、Bucket Sort Aggregation、Max Bucket Aggregation、Min Bucket Aggregation、Sum Bucket Aggregation、Stats Bucket Aggregation。

原文發佈時間爲：2019-03-18
本文作者：丁威，《RocketMQ技術內幕》作者。
本文來自中間件興趣圈，瞭解相關信息可以關注中間件興趣圈。

elasticsearc使用指南之ES管道聚合(Pipeline Aggregation)

1、Avg Bucket Aggregation

2、Percentiles Bucket Aggregation

3、Cumulative Sum Aggregation

4、Bucket Sort Aggregation

5、Max Bucket Aggregation

6、Min Bucket Aggregation

7、Sum Bucket Aggregation

8、Stats Bucket Aggregation

Kafka 和 RocketMQ 之性能對比

程序員如何提高影響力

Elasticsearch Index Aliases詳解

elasticsearc使用指南之ES管道聚合(Pipeline Aggregation)

Elasticsearch Index Setting一覽表

https://yachay.unat.edu.pe/blog/index.php?comment_area=format_blog&comment_component=blog&comment_co

linux以太網驅動總結