3.3 聚合搜索

聚合是搜索中很重要的一个功能，本章对一些基础的分桶、指标计算做一个简单的讲解。

本章开始前先创建以下索引：

PUT test_aggs_index
{
  "mappings": {
    "properties": {
      "name": {
        "type": "keyword"
      },
      "course": {
        "type": "keyword"
      },
      "class": {
        "type": "keyword"
      },
      "score": {
        "type": "float"
      }
    }
  }
}

并添加样例数据：

PUT _bulk
{"create":{"_index":"test_aggs_index","_id":1}}
{"name":"刘一","course": "java", "class": "1班", "score": 98.4}
{"create":{"_index":"test_aggs_index","_id":2}}
{"name":"陈二","course": "php", "class": "3班", "score": 37}
{"create":{"_index":"test_aggs_index","_id":3}}
{"name":"张三","course": "php", "class": "3班", "score": 67}
{"create":{"_index":"test_aggs_index","_id":4}}
{"name":"李四","course": "java", "class": "2班", "score": 48}
{"create":{"_index":"test_aggs_index","_id":5}}
{"name":"王五","course": "php", "class": "1班", "score": 80}
{"create":{"_index":"test_aggs_index","_id":6}}
{"name":"赵六","course": "java", "class": "1班", "score": 99}
{"create":{"_index":"test_aggs_index","_id":7}}
{"name":"孙七","course": "python", "class": "3班", "score": 94}
{"create":{"_index":"test_aggs_index","_id":8}}
{"name":"周八","course": "python", "class": "2班", "score": 64}
{"create":{"_index":"test_aggs_index","_id":9}}
{"name":"吴九","course": "python", "class": "1班", "score": 87}
{"create":{"_index":"test_aggs_index","_id":10}}
{"name":"郑十","course": "java", "class": "2班", "score": 68}

3.3.1 分桶聚合

分桶聚合类似MySQL的Group By查询，按照指定条件，进行分组统计

1. 单字段聚合

查询各个科目考试的有多少人

1.声明字段聚合

GET test_aggs_index/_search
{
  "size": 0,
  "aggs": {
    "my_aggs_name": {
      "terms": {
        "field": "course",
        "size": 10,
        "order": {
          "_count": "asc"
        }
      }
    }
  }
}

2. 使用 script 聚合

可以用脚本聚合

GET test_aggs_index/_search
{
  "size": 0,
  "aggs": {
    "my_aggs_name": {
      "terms": {
        "script": "doc['course'].value",
        "size": 10,
        "order": {
          "_count": "asc"
        }
      }
    }
  }
}

参数介绍： size：结果条数，只要聚合结果设置0 aggs：聚合函数
my_aggs_name：这个位置写自定义名称，查询结果会以这个名称为 key 返回结果
terms：指定字段
field：字段名称
size：返回数量
order：聚合结果排序，当前设置按照数量排序，默认 _key 按照名称排序

2. 多字段聚合

如果要汇总 每个班级各科目考试人数 ，就需要用到多字段统计，本章讲 script 实现。

ES的更高版本 7.12 及以上有 multi_terms 语法支持，可以自行阅读。

GET test_aggs_index/_search
{
  "size": 0,
  "aggs": {
    "my_aggs_name": {
      "terms": {
        "script": "doc['class'].value+'|'+doc['course'].value",
        "size": 10,
        "order": {
          "_count": "asc"
        }
      }
    }
  }
}

3. 条件聚合

只查询 1班 的聚合结果

GET test_aggs_index/_search
{
  "size":0,
  "query": {
    "term": {
      "class": {
        "value": "1班"
      }
    }
  },
  "aggs": {
    "my_aggs_name": {
      "terms": {
        "field": "course",
        "size": 10,
        "order": {
          "_count": "asc"
        }
      }
    }
  }
}

3.3.2 指标聚合

在分桶后进行指标计算，例如Avg(平均值)、Max(最大值)、Min(最小值)、Sum(求和)、Cardinality(去重)等。

1. Avg、Max、Min、Sum

查询每个科目成绩的平均分、最高分、最低分、总分

GET test_aggs_index/_search
{
  "size": 0,
  "aggs": {
    "my_aggs_name": {
      "terms": {
        "field": "course",
        "size": 10
      },
      "aggs": {
        "avg_score": {
          "avg": {
            "field": "score"
          }
        },
        "max_score": {
          "max": {
            "field": "score"
          }
        },
        "min_score": {
          "min": {
            "field": "score"
          }
        },
        "sum_score": {
          "sum": {
            "field": "score"
          }
        }
      }
    }
  }
}

该语句包含嵌套聚合，在科目分桶基础上，对每个科目计算指标

2. Cardinality 去重

对字段去重统计数量，类似MySQL的 Distinct + Count

查询科目数量

GET test_aggs_index/_search
{
  "size": 0,
  "aggs": {
    "count": {
      "cardinality": {
        "field": "course"
      }
    }
  }
}

3. 指标统计汇总

有一个函数 stats ，可以把一般的聚合指标计算后返回

GET test_aggs_index/_search
{
  "size": 0,
  "aggs": {
    "my_aggs_name": {
      "terms": {
        "field": "course",
        "size": 10
      },
      "aggs": {
        "my_stats": {
          "stats": {
            "field": "score"
          }
        }
      }
    }
  }
}

4. 百分位统计

对数据进行百分位的统计，函数为：percentiles

GET test_aggs_index/_search
{
  "size": 0,
  "aggs": {
    "my_aggs_name": {
      "terms": {
        "field": "course",
        "size": 10
      },
      "aggs": {
        "my_percentiles": {
          "percentiles": {
            "field": "score"
          }
        }
      }
    }
  }
}

会输出 1~99 的分布数值

可以指定指标统计范围，例如只想要 1、50、99的数据

GET test_aggs_index/_search
{
  "size": 0,
  "aggs": {
    "my_aggs_name": {
      "terms": {
        "field": "course",
        "size": 10
      },
      "aggs": {
        "my_percentiles": {
          "percentiles": {
            "field": "score",
            "percents": [1, 50, 99]
          }
        }
      }
    }
  }
}

该函数更多操作请参考官方文档