Elasticsearch:aggregation介绍

聚合 (aggregation) 功能集是整个 Elasticsearch 产品中最令人兴奋和有益的功能之一,主要是因为它提供了一个非常有吸引力对之前的 facets 的替代。

在本教程中,我们将解释Elasticsearch中的聚合(aggregation)并逐步介绍一些示例。 我们比较了指标聚合和存储桶聚合,并展示了如何利用聚合嵌套(对于 facets 而言这是不可能的)。 欢迎你在本文中复制所有示例代码。

 

关于 Elastic Facets 的一点背景

如果你曾经使用过 Elasticsearch 的 facets,那么你肯定了解它们的实用性。 经过丰富的经验,我们在这里告诉你 Elasticsearch  聚合(aggregations)甚至更好。 facets 使你可以快速计算和汇总查询结果,并且可以将其用于各种任务,例如结果值的动态计数或创建分布直方图。 尽管facets非常强大,但它们在 Elasticsearch 核心中的实现存在一些限制。 由于 facets 仅执行一级深度的计算,因此将它们组合起来并不容易。

聚合(Aggregation)API 解决了这些问题,并且还提供了一种简单的方法在查询时(在单个请求中)进行的非常精确的多级计算。 简而言之:Elasticsearch聚合是对facets的一个更加全面的提高的。

 

准备数据

为了完成我们今天的练习,我们先来准备一些数据。我们想创建一个叫做 sports 的索引。为此,我们先创建一个 mapping

PUT sports
{
  "mappings": {
    "properties": {
      "birthdate": {
        "type": "date",
        "format": "dateOptionalTime"
      },
      "location": {
        "type": "geo_point"
      },
      "name": {
        "type": "keyword"
      },
      "rating": {
        "type": "integer"
      },
      "sport": {
        "type": "keyword"
      }
    }
  }
}

在上面,我们定义了一个 sports 索引的 mapping。在下面,我们通过 bulk API 来把我们想要的数据导入到索引中。

POST _bulk/
{"index":{"_index":"sports"}}
{"name":"Michael","birthdate":"1989-10-1","sport":"Baseball","rating":["5","4"],"location":"46.22,-68.45"}
{"index":{"_index":"sports"}}
{"name":"Bob","birthdate":"1989-11-2","sport":"Baseball","rating":["3","4"],"location":"45.21,-68.35"}
{"index":{"_index":"sports"}}
{"name":"Jim","birthdate":"1988-10-3","sport":"Baseball","rating":["3","2"],"location":"45.16,-63.58"}
{"index":{"_index":"sports"}}
{"name":"Joe","birthdate":"1992-5-20","sport":"Baseball","rating":["4","3"],"location":"45.22,-68.53"}
{"index":{"_index":"sports"}}
{"name":"Tim","birthdate":"1992-2-28","sport":"Baseball","rating":["3","3"],"location":"46.22,-68.85"}
{"index":{"_index":"sports"}}
{"name":"Alfred","birthdate":"1990-9-9","sport":"Baseball","rating":["2","2"],"location":"45.12,-68.35"}
{"index":{"_index":"sports"}}
{"name":"Jeff","birthdate":"1990-4-1","sport":"Baseball","rating":["2","3"],"location":"46.12,-68.55"}
{"index":{"_index":"sports"}}
{"name":"Will","birthdate":"1988-3-1","sport":"Baseball","rating":["4","4"],"location":"46.25,-68.55"}
{"index":{"_index":"sports"}}
{"name":"Mick","birthdate":"1989-10-1","sport":"Baseball","rating":["3","4"],"location":"46.22,-68.45"}
{"index":{"_index":"sports"}}
{"name":"Pong","birthdate":"1989-11-2","sport":"Baseball","rating":["1","3"],"location":"45.21,-68.35"}
{"index":{"_index":"sports"}}
{"name":"Ray","birthdate":"1988-10-3","sport":"Baseball","rating":["2","2"],"location":"45.16,-63.58"}
{"index":{"_index":"sports"}}
{"name":"Ping","birthdate":"1992-5-20","sport":"Baseball","rating":["4","3"],"location":"45.22,-68.53"}
{"index":{"_index":"sports"}}
{"name":"Duke","birthdate":"1992-2-28","sport":"Baseball","rating":["5","2"],"location":"46.22,-68.85"}
{"index":{"_index":"sports"}}
{"name":"Hal","birthdate":"1990-9-9","sport":"Baseball","rating":["4","2"],"location":"45.12,-68.35"}
{"index":{"_index":"sports"}}
{"name":"Charge","birthdate":"1990-4-1","sport":"Baseball","rating":["3","2"],"location":"46.12,-68.55"}
{"index":{"_index":"sports"}}
{"name":"Barry","birthdate":"1988-3-1","sport":"Baseball","rating":["5","2"],"location":"46.25,-68.55"}
{"index":{"_index":"sports"}}
{"name":"Bank","birthdate":"1988-3-1","sport":"Golf","rating":["6","4"],"location":"46.25,-68.55"}
{"index":{"_index":"sports"}}
{"name":"Bingo","birthdate":"1988-3-1","sport":"Golf","rating":["10","7"],"location":"46.25,-68.55"}
{"index":{"_index":"sports"}}
{"name":"James","birthdate":"1988-3-1","sport":"Basketball","rating":["10","8"],"location":"46.25,-68.55"}
{"index":{"_index":"sports"}}
{"name":"Wayne","birthdate":"1988-3-1","sport":"Hockey","rating":["10","10"],"location":"46.25,-68.55"}
{"index":{"_index":"sports"}}
{"name":"Brady","birthdate":"1988-3-1","sport":"Football","rating":["10","10"],"location":"46.25,-68.55"}
{"index":{"_index":"sports"}}
{"name":"Lewis","birthdate":"1988-3-1","sport":"Football","rating":["10","10"],"location":"46.25,-68.55"}

通过上面的 bulk API 接口,我们可以把我们想要的数据输入到 sports 的索引中。我们可以通过如下的接口来获得我多少条数据:

GET sports/_count

显示结果:

{
  "count" : 22,
  "_shards" : {
    "total" : 1,
    "successful" : 1,
    "skipped" : 0,
    "failed" : 0
  }
}

在这个数据库里,我们有可以看到有22条的数据。

 

动手实践

聚合的两个主要系列是指标聚合(metric aggregations)和存储桶聚合(bucket aggregation)。 指标聚合计算一组文档中的某些值(例如平均值); 存储桶聚合将文档分组到存储桶中。 在详细介绍之前,让我们看一下聚合请求的一般结构。除此之前,聚合还有 Matrix 及 Pipleline 聚合。

Aggregation结构

"aggregations" : {
    "<aggregation_name>" : {
        "<aggregation_type>" : { 
            <aggregation_body>
        },
        ["aggregations" : { [<sub_aggregation>]* } ]
    }
    [,"<aggregation_name_2>" : { ... } ]*
}

请求 json 中的聚合(你也可以改用 aggs)对象包含聚合名称,类型和主体。 <aggregation_name> 是用户定义的名称(不带括号),该名称将唯一标识响应中的聚合名称/键。

<aggregation_type> 通常是聚合中的第一个键。 它可以是 terms,stats 或 geo-distance 聚合,但这是它的起点。 在我们的 <aggregation_type> 中,我们有一个 <aggregation_body>。 在 <aggregation_body> 中,我们指定聚合所需的属性。 可用属性取决于聚合的类型。

你可以选择提供子聚合,以将一个聚合元素的结果嵌套到另一个聚合元素中。 此外,你可以在查询中输入多个聚合(aggregation_name_2),以具有更多单独的顶级聚合。 尽管对嵌套级别没有限制,但是你不能将度量标准嵌套在度量标准聚合中,原因如下所述。 在研究可以聚合的不同类型的值之后,我们将了解桶聚合和度量聚合之间的区别。

 

例子

一些聚合使用从聚合文档中获取的值。 这些值可以从指定的文档字段(field)中获取,也可以从随每个文档生成值的脚本中获取。 下面的第一个示例在名称字段上提供了术语聚合(terms aggregation),在子聚合 rating_avg 值上给出了顺序。 如你所见,我们使用嵌套的指标聚合对存储桶聚合的结果进行排序。

尽管我们使用上面给出的索引,但是我们鼓励你运行此查询(以及下面的其他查询)。 你可以从工作中获得直接结果,然后对其进行修改以匹配你的数据集。

另外,请仔细查看我们是否包含 “size”:0,因为我们的重点是聚合结果,而不是文档结果。这里设置为0,表示我们不想得到任何的文档。

GET sports/_search
{
  "size": 0, 
  "aggregations": {
    "the_name": {
      "terms": {
        "field": "name",
        "order": {
          "rating_avg": "desc"
        }
      },
      "aggregations": {
        "rating_avg": {
          "avg": {
            "field": "rating"
          }
        }
      }
    }
  }
}

显示的结果为:

{
  "took" : 1,
  "timed_out" : false,
  "_shards" : {
    "total" : 1,
    "successful" : 1,
    "skipped" : 0,
    "failed" : 0
  },
  "hits" : {
    "total" : {
      "value" : 22,
      "relation" : "eq"
    },
    "max_score" : null,
    "hits" : [ ]
  },
  "aggregations" : {
    "the_name" : {
      "doc_count_error_upper_bound" : 0,
      "sum_other_doc_count" : 12,
      "buckets" : [
        {
          "key" : "Brady",
          "doc_count" : 1,
          "rating_avg" : {
            "value" : 10.0
          }
        },
        {
          "key" : "Lewis",
          "doc_count" : 1,
          "rating_avg" : {
            "value" : 10.0
          }
        },
        {
          "key" : "Wayne",
          "doc_count" : 1,
          "rating_avg" : {
            "value" : 10.0
          }
        },
        {
          "key" : "James",
          "doc_count" : 1,
          "rating_avg" : {
            "value" : 9.0
          }
        },
        {
          "key" : "Bingo",
          "doc_count" : 1,
          "rating_avg" : {
            "value" : 8.5
          }
        },
        {
          "key" : "Bank",
          "doc_count" : 1,
          "rating_avg" : {
            "value" : 5.0
          }
        },
        {
          "key" : "Michael",
          "doc_count" : 1,
          "rating_avg" : {
            "value" : 4.5
          }
        },
        {
          "key" : "Will",
          "doc_count" : 1,
          "rating_avg" : {
            "value" : 4.0
          }
        },
        {
          "key" : "Barry",
          "doc_count" : 1,
          "rating_avg" : {
            "value" : 3.5
          }
        },
        {
          "key" : "Bob",
          "doc_count" : 1,
          "rating_avg" : {
            "value" : 3.5
          }
        }
      ]
    }
  }
}

上面的结果显示:我们得到了按照每个人来进行分类的聚合,而他们的顺序是按照 rating_avg 聚合所获得平均分数来排序的。

我们还可以提供一个 script 脚本来生成聚合所使用的值:

GET sports/_search
{
  "size": 0,
  "aggs": {
    "age_range": {
      "range": {
        "script": {
          "source": 
            """
            def dob = doc['birthdate'].value;
            return params.now - dob.getYear()
            """
            ,
          "params": {
            "now": 2019
          }
        },
        "ranges": [
          {
            "from": 30,
            "to": 31
          }
        ]
      }
    }
  }
}

 在上面,我们通过脚本生成 value source,并对它做出统计。

显示的结果是:

{
  "took" : 0,
  "timed_out" : false,
  "_shards" : {
    "total" : 1,
    "successful" : 1,
    "skipped" : 0,
    "failed" : 0
  },
  "hits" : {
    "total" : {
      "value" : 22,
      "relation" : "eq"
    },
    "max_score" : null,
    "hits" : [ ]
  },
  "aggregations" : {
    "age_range" : {
      "buckets" : [
        {
          "key" : "30.0-31.0",
          "from" : 30.0,
          "to" : 31.0,
          "doc_count" : 4
        }
      ]
    }
  }
}

上面显示在 30 至 31 岁之间的有 4 个人。

 

Metric Aggregations

指标聚合类型用于计算整个文档集的指标。 有单值指标聚合(例如avg)和多值指标聚合(例如stats)。 指标聚合的一个简单示例是 value_count 聚合,它仅返回已为给定字段建立索引的值的总数。 要在运动员数据集中的 “sport” 字段中找到值的数量,我们可以使用以下查询:

GET sports/_search
{
  "size": 0,
  "aggs": {
    "sport_count": {
      "value_count": {
        "field": "sport"
      }
    }
  }
}

显示结果:

{
  "took" : 2,
  "timed_out" : false,
  "_shards" : {
    "total" : 1,
    "successful" : 1,
    "skipped" : 0,
    "failed" : 0
  },
  "hits" : {
    "total" : {
      "value" : 22,
      "relation" : "eq"
    },
    "max_score" : null,
    "hits" : [ ]
  },
  "aggregations" : {
    "sport_count" : {
      "value" : 22
    }
  }
}

请注意,这将返回该字段的值总数,而不是唯一值的数目。 因此,在这种情况下(由于每个文档在 “sport” 字段中都有一个单词值),结果仅等于索引中的文档数。

Bucket Aggregations

存储桶聚合是用于对文档进行分组的机制。 每种类型的存储桶聚合都有自己的分割文档集的方法。 也许最简单的类型是术语聚合。 这个功能非常像术语方面,返回给定字段索引的唯一术语以及匹配文档的数量。 如果我们想在数据集中的 “sport” 字段中找到所有值,则可以使用以下方法:

GET sports/_search
{
  "size": 0,
  "aggs": {
    "sport": {
      "terms": {
        "field": "sport",
        "size": 10
      }
    }
  }
}

返回值:

{
  "took" : 0,
  "timed_out" : false,
  "_shards" : {
    "total" : 1,
    "successful" : 1,
    "skipped" : 0,
    "failed" : 0
  },
  "hits" : {
    "total" : {
      "value" : 22,
      "relation" : "eq"
    },
    "max_score" : null,
    "hits" : [ ]
  },
  "aggregations" : {
    "sport" : {
      "doc_count_error_upper_bound" : 0,
      "sum_other_doc_count" : 0,
      "buckets" : [
        {
          "key" : "Baseball",
          "doc_count" : 16
        },
        {
          "key" : "Football",
          "doc_count" : 2
        },
        {
          "key" : "Golf",
          "doc_count" : 2
        },
        {
          "key" : "Basketball",
          "doc_count" : 1
        },
        {
          "key" : "Hockey",
          "doc_count" : 1
        }
      ]
    }
  }
}

你可能会发现 geo_distance 聚合更具吸引力。 尽管它有许多选项,但在最简单的情况下,它取一个原点和一个距离范围,然后根据给定的 geo_point 字段计算圆中有多少文档。

假设我们需要知道多少个运动员居住在距离地理位置“ 46.12,-68.55” 20英里范围内。 我们可以使用以下聚合:
 

GET sports/_search
{
  "size": 0,
  "aggregations": {
    "baseball_player_ring": {
      "geo_distance": {
        "field": "location",
        "origin": "46.12,-68.55",
        "unit": "mi",
        "ranges": [
          {
            "from": 0,
            "to": 20
          }
        ]
      }
    }
  }
}

返回结果:

{
  "took" : 4,
  "timed_out" : false,
  "_shards" : {
    "total" : 1,
    "successful" : 1,
    "skipped" : 0,
    "failed" : 0
  },
  "hits" : {
    "total" : {
      "value" : 22,
      "relation" : "eq"
    },
    "max_score" : null,
    "hits" : [ ]
  },
  "aggregations" : {
    "baseball_player_ring" : {
      "buckets" : [
        {
          "key" : "*-20.0",
          "from" : 0.0,
          "to" : 20.0,
          "doc_count" : 14
        }
      ]
    }
  }
}

内嵌 Bucket Aggregations

许多开发人员会同意,桶聚合的最强大方面是嵌套它们的能力。 你可以定义顶级存储桶聚合,并在其内部定义对每个结果存储桶进行操作的第二级聚合。 此嵌套可以根据需要扩展到多个级别。

继续我们的示例,我们可以使用按年龄划分的嵌套范围聚合(根据脚本的“出生日期”计算得出)来进一步细分 geo_distance 聚合的结果。 假设我们想知道属于两个年龄段的每个运动员中有多少运动员(他们生活在上一节中定义的圈子内)。 我们可以使用以下聚合来获取此信息:

GET sports/_search
{
   "size": 0,
   "aggregations": {
      "baseball_player_ring": {
         "geo_distance": {
            "field": "location",
            "origin": "46.12,-68.55",
            "unit": "mi",
            "ranges": [
               {
                  "from": 0,
                  "to": 20
               }
            ]
         },
         "aggregations": {
            "ring_age_ranges": {
               "range": {
                 "script": {
                    "source": 
                    """
                    ZonedDateTime dob = doc['birthdate'].value;
                    return params.now - dob.getYear()
                    """
                  ,
                  "params": {
                    "now": 2019
                  }                 
                 }, 
                  "ranges": [
                      { "from": 30, "to": 31 },
                      { "from": 31, "to": 32 }
                  ]
               }
            }
         }
      }
   }
}

显示的结果为:

{
  "took" : 0,
  "timed_out" : false,
  "_shards" : {
    "total" : 1,
    "successful" : 1,
    "skipped" : 0,
    "failed" : 0
  },
  "hits" : {
    "total" : {
      "value" : 22,
      "relation" : "eq"
    },
    "max_score" : null,
    "hits" : [ ]
  },
  "aggregations" : {
    "baseball_player_ring" : {
      "buckets" : [
        {
          "key" : "*-20.0",
          "from" : 0.0,
          "to" : 20.0,
          "doc_count" : 14,
          "ring_age_ranges" : {
            "buckets" : [
              {
                "key" : "30.0-31.0",
                "from" : 30.0,
                "to" : 31.0,
                "doc_count" : 2
              },
              {
                "key" : "31.0-32.0",
                "from" : 31.0,
                "to" : 32.0,
                "doc_count" : 8
              }
            ]
          }
        }
      ]
    }
  }
}

现在,让我们使用 stats(多值指标汇总器)来计算最内部结果的一些统计数据。 对于居住在我们圈子中的运动员以及两个年龄段的每个年龄段,我们现在都希望根据结果文档计算 “rating” 字段的统计信息:

GET sports/_search
{
   "size": 0,
   "aggregations": {
      "baseball_player_ring": {
         "geo_distance": {
            "field": "location",
            "origin": "46.12,-68.55",
            "unit": "mi",
            "ranges": [
               {
                  "from": 0,
                  "to": 20
               }
            ]
         },
         "aggregations": {
            "ring_age_ranges": {
               "range": {
                 "script": {
                    "source": 
                    """
                    ZonedDateTime dob = doc['birthdate'].value;
                    return params.now - dob.getYear()
                    """
                  ,
                  "params": {
                    "now": 2019
                  }                 
                 }, 
                  "ranges": [
                      { "from": 30, "to": 31 },
                      { "from": 31, "to": 32 }
                  ]
               },
              "aggregations": {
                "rating_stats": {
                  "stats": {
                      "field": "rating"
                    }
                }
              }
            }
         }
      }
   }
}

我们得到一个我们需要的统计信息的响应:

{
  "took" : 0,
  "timed_out" : false,
  "_shards" : {
    "total" : 1,
    "successful" : 1,
    "skipped" : 0,
    "failed" : 0
  },
  "hits" : {
    "total" : {
      "value" : 22,
      "relation" : "eq"
    },
    "max_score" : null,
    "hits" : [ ]
  },
  "aggregations" : {
    "baseball_player_ring" : {
      "buckets" : [
        {
          "key" : "*-20.0",
          "from" : 0.0,
          "to" : 20.0,
          "doc_count" : 14,
          "ring_age_ranges" : {
            "buckets" : [
              {
                "key" : "30.0-31.0",
                "from" : 30.0,
                "to" : 31.0,
                "doc_count" : 2,
                "rating_stats" : {
                  "count" : 4,
                  "min" : 3.0,
                  "max" : 5.0,
                  "avg" : 4.0,
                  "sum" : 16.0
                }
              },
              {
                "key" : "31.0-32.0",
                "from" : 31.0,
                "to" : 32.0,
                "doc_count" : 8,
                "rating_stats" : {
                  "count" : 16,
                  "min" : 2.0,
                  "max" : 10.0,
                  "avg" : 7.5,
                  "sum" : 120.0
                }
              }
            ]
          }
        }
      ]
    }
  }
}

如你所见,你可以创建一个包含多个存储更多存储桶的大存储桶。 你还可以获取每个存储分区的指标(metrics),以及不断提高的复杂性。 通过这些简单的构建块,你可以使用嵌套聚合从数据中获得深刻而复杂的见解。

已标记关键词 清除标记