一、使用cardinality去重统计

一般我们在使用elasticsearch进行去重是通过在聚合里使用cardinality对统计结果的去重，比如有个字段“one_account.one_account_no”,有两个文档的“one_account.one_account_no”值都是111，那么对“one_account.one_account_no“”去重后结果是1。
dsl语句：

POST user_onoffline_log/_search
{
  "query": {
    "match_all": {}
  },
  "size": 0,
  "aggs": {
    "one_account_no_aggs": {
      "cardinality": {
        "field": "one_account.one_account_no"
      }
    }
  }
}

结果：

{
  "took": 565,
  "timed_out": false,
  "_shards": {
    "total": 5,
    "successful": 5,
    "failed": 0
  },
  "hits": {
    "total": 21,
    "max_score": 0,
    "hits": []
  },
  "aggregations": {
    "one_account_no_aggs": {
      "value": 14
    }
  }
}

可以看到使用cardinality对one_account.one_account_no字段去重后的计数值为14。

二、collapse对内容去重

上面的使用cardinality去重是作为统计来使用，如果我们想查询所有去除重复后的one_account_no有哪些而不是仅仅得到一个数字14的话，可以使用collapse对内容去重。
dsl语句：

GET customer/_search
{
  "from": 0,
  "size": 5,
  "query": {
    "bool": {
      "filter": [
        {
          "exists": {
            "field": "one_account.one_account_no",
            "boost": 1
          }
        }
      ]
    }
  },
  "collapse": {
    "field": "one_account.one_account_no"
  },
  "_source": {
    "includes": [
      "one_account.one_account_no"
    ],"excludes": []
  },
  "aggregations": {
    "count": {
      "cardinality": {
        "field": "one_account.one_account_no"
      }
    }
  }
}

说明：exists筛选出字段one_account.one_account_no存在的，collapse对多个相同的只取其中一个展示，_source要返回展示的字段，cardinality去重统计one_account.one_account_no的数量

结果：

{
  "took" : 2,
  "timed_out" : false,
  "_shards" : {
    "total" : 6,
    "successful" : 6,
    "skipped" : 0,
    "failed" : 0
  },
  "hits" : {
    "total" : 21,
    "max_score" : 0.0,
    "hits" : [
      {
        "_index" : "customer_v2025",
        "_type" : "customer_info",
        "_id" : "105100015130",
        "_score" : 0.0,
        "_source" : {
          "one_account" : {
            "one_account_no" : "105100015130"
          }
        },
        "fields" : {
          "one_account.one_account_no" : [
            "105100015130"
          ]
        }
      },
      {
        "_index" : "customer_v2025",
        "_type" : "customer_info",
        "_id" : "99522458",
        "_score" : 0.0,
        "_source" : {
          "one_account" : {
            "one_account_no" : "99522458"
          }
        },
        "fields" : {
          "one_account.one_account_no" : [
            "99522458"
          ]
        }
      },
      {
        "_index" : "customer_v2025",
        "_type" : "customer_info",
        "_id" : "105500032110",
        "_score" : 0.0,
        "_source" : {
          "one_account" : {
            "one_account_no" : "105500032110"
          }
        },
        "fields" : {
          "one_account.one_account_no" : [
            "105500032110"
          ]
        }
      },
      {
        "_index" : "customer_v2025",
        "_type" : "customer_info",
        "_id" : "110600001247",
        "_score" : 0.0,
        "_routing" : "110600001247",
        "_source" : {
          "one_account" : {
            "one_account_no" : "110600001248"
          }
        },
        "fields" : {
          "one_account.one_account_no" : [
            "110600001248"
          ]
        }
      },
      {
        "_index" : "customer_v2025",
        "_type" : "customer_info",
        "_id" : "110600000858",
        "_score" : 0.0,
        "_source" : {
          "one_account" : {
            "one_account_no" : "110600000858"
          }
        },
        "fields" : {
          "one_account.one_account_no" : [
            "110600000858"
          ]
        }
      }
    ]
  },
  "aggregations" : {
    "count" : {
      "value" : 14
    }
  }
}

可以看出总的含有one_account.one_account_no的有21个，去除重复的后有14个，
因为采用了from-size分页，hits里只返回了5条，要把14条都显示出来，可以把from改为5或10，来查看第2页和第3页的。

Java代码：

SearchRequest searchRequest = new SearchRequest("customer");
searchRequest.types("customer_info");
SearchSourceBuilder searchSourceBuilder = new SearchSourceBuilder();
BoolQueryBuilder boolQueryBuilder = QueryBuilders.boolQuery();
boolQueryBuilder.filter(QueryBuilders.existsQuery("one_account.one_account_no"));
searchSourceBuilder.collapse(new CollapseBuilder("one_account.one_account_no"));
searchSourceBuilder.aggregation(AggregationBuilders.cardinality("count").field("one_account.one_account_no"));
searchSourceBuilder.from(0);
searchSourceBuilder.size(5);
searchSourceBuilder.query(boolQueryBuilder);
searchRequest.source(searchSourceBuilder);
      
SearchResponse  searchresponse = restHighLevelClient.search(searchRequest);

参考：
Elasticsearch去重查询