elasticsearch - Elasticsearch聚合和最大计数类型的操作

Question

我在Elasticsearch中存储了一些文档，如下所示:

{
          "date" : 1,
          "field1" : 0.2,
          "field2" : 0.5,
          "field3" : 0.3
},
{
          "date" : 1,
          "field1" : 0.9,
          "field2" : 0.5,
          "field3" : 0.1
},
{
          "date" : 2,
          "field1" : 0.2,
          "field2" : 0.6,
          "field3" : 0.7
}

我想得到的是对每个文档中field1，field2或field3中每个字段最大的次数进行计数，即按日期分组。期望结果是这样的:

{
          "date" : 1,
          "field1-greatest" : 1,
          "field2-greatest" : 1,
          "field3-greatest" : 0
},
{
          "date" : 2,
          "field1-greatest" : 0,
          "field2-greatest" : 0,
          "field3-greatest" : 1
}

我在日期上使用术语聚合，但不确定如何使用Elasticsearch聚合比较不同字段以执行此最大和计数类型操作。有什么建议么？

Answer 1

您可以执行以下操作:

{
  "size": 0,
  "aggs": {
    "by_date": {
      "terms": {
        "field": "date"
      },
      "aggs": {
        "field1_greatest": {
          "max": {
            "field": "field1"
          }
        },
        "field2_greatest": {
          "max": {
            "field": "field2"
          }
        },
        "field3_greatest": {
          "max": {
            "field": "field3"
          }
        }
      }
    }
  }
}

提示:确保将 field*属性映射为 double类型，而不是 float类型，因为 field1的最大agg可能会生成 0.8999999761581421而不是 0.9。

更正
这是一个不平凡的用例，因此您可能需要 use a script。以下是一些入门知识:

{
  "size": 0,
  "aggs": {
    "by_date": {
      "terms": {
        "field": "date"
      },
      "aggs": {
        "by_greatest": {
          "scripted_metric": {
            "init_script": """
              state.field1_greatest = 0;
              state.field2_greatest = 0;
              state.field3_greatest = 0;
            """,
            "map_script": """
              def v1 = doc['field1'].value;
              def v2 = doc['field2'].value;
              def v3 = doc['field3'].value;
              
              // your comparison logic
            """,
            "combine_script": "state",
            "reduce_script": "states"
          }
        }
      }
    }
  }
}