【ES小结】还在用ElasticSearch做查询？换条思路实现高效数据统计

博客首页：派 大 星

⛳️ 欢迎关注 点赞 收藏 ✏️ 留言

本文由派大星原创编撰

系列专栏：《ES小结》

本系列记录ElasticSearch技术学习历程以及问题解决

---

---

聚合查询

① 什么是聚合查询

聚合是ES除搜索功能外提供的针对ES数据做统计分析的功能，聚合有助于根据搜索查询提供聚合数据，聚合查询是数据库中重要额功能特性，ES作为搜索引擎兼数据库，同样提供了强大的聚合分析功能力，它是基于查询条件来对数据进行分桶、计算的方法，这种很类似与SQL 中的group by 再加上一些函数方法的操作。

在了解聚合查询之前需要注意的一点是：text类型是不支持聚合的，主要是因为text类型本身是分词的，通俗的说，如果一句话分成了多个词然后进行group by操作，那么问题就出现了，到底对哪一个词进行group by操作呢？无法指定！

② Kibana 命令测试聚合查询

创建测试索引

PUT /fruit
{
    "mappings":{
        "properties":{
            "title":"keyword"
        },
        "price":{
            "type":"double"
        },
        "description":{
            "type":"text"
        }
    }
}

存放测试数据

PUT /fruit/_bulk
{"index":{}}
	{"title":"面包","price":19.6,"description":"小面包很便宜"}
{"index":{}}
	{"title":"旺旺牛奶","price":29.6,"description":"旺旺牛奶很好喝"}
{"index":{}}
	{"title":"日本豆","price":9.0,"description":"日本豆很便宜"}
{"index":{}}
	{"title":"大辣条","price":10.6,"description":"大辣条超级好吃"}
{"index":{}}
	{"title":"海苔","price":49.6,"description":"海苔很一般"}
{"index":{}}
	{"title":"小饼干","price":9.6,"description":"小饼干很小"}
{"index":{}}
	{"title":"小葡萄","price":59.6,"description":"小葡萄很好吃"}	
{"index":{}}
	{"title":"小饼干","price":19.6,"description":"小饼干很小"}
{"index":{}}
	{"title":"小饼干","price":59.6,"description":"小饼干很小"}
{"index":{}}
	{"title":"小饼干","price":29.6,"description":"小饼干很小"}
{"index":{}}
	{"title":"小饼干","price":39.6,"description":"小饼干很小"}

③ 聚合操作使用

根据某个字段分组

GET /fruit/_search
{
  "query": {
    "match_all": {
      
    }
  },
  "aggs": {
    "price_group": {
      "terms": {
        "field": "price"
      }
    }
  }
}

求最大值

GET /fruit/_search
{
  "query": {
    "match_all": {}
  },
  "aggs": {
    "max_price": {
      "max": {
        "field": "price"
      }
    }
  }
}

最小值

GET /fruit/_search
{
  "query": {
    "match_all": {}
  },
  "size": 0, 
  "aggs": {
    "min_price": {
      "min": {
        "field": "price"
      }
    }
  }
}

求总数

GET /fruit/_search
{
  "query": {
    "match_all": {}
  },
  "size": 0, 
  "aggs": {
    "min_price": {
      "sum": {
        "field": "price"
      }
    }
  }
}

求平均值

GET /fruit/_search
{
  "query": {
    "match_all": {}
  },
  "size": 0, 
  "aggs": {
    "avg_price": {
      "avg": {
        "field": "price"
      }
    }
  }
}

④ RestHighLevelClient 测试聚合查询

在使用Java API实现上述操作之前，有必要先了解一下实现过程中使用到的某些方法以及工具

常见的聚合查询：

• 统计某个字段的数量

ValueCountBuilder vcb= AggregationBuilders.count(“分组的名称”).field(“字段”);

• 去重统计某个字段的数量（有少量的误差）

CardinalityBuilder cb= AggregationBuilders.cardinality(“分组的名称”).field(“字段”);

• 聚合过滤

FilterAggregationBuilder fab= AggregationBuilders.filter(“分组的名称”).filter(QueryBuilders.queryStringQuery(“字段:过滤值”));

• 按某个字段分组

TermsBuilder tb= AggregationBuilders.terms(“分组的名称”).field(“字段”);

• 求最大值

SumBuilder sumBuilder= AggregationBuilders.max(“分组的名称”).field(“字段”);

• 求最小值

AvgBuilder ab= AggregationBuilders.min(“分组的名称”).field(“字段”);

• 求平均值

MaxBuilder mb= AggregationBuilders.avg(“分组的名称”).field(“字段”);

• 按日期间隔分组

DateHistogramBuilder dhb= AggregationBuilders.dateHistogram(“分组的名称”).field(“字段”);

• 获取聚合里面的结果

TopHitsBuilder thb= AggregationBuilders.topHits(“分组的名称”);

• 嵌套的聚合

NestedBuilder nb= AggregationBuilders.nested(“分组的名称”).path(“字段”);

• 反转嵌套

AggregationBuilders.reverseNested(“分组的名称”).path("字段 ");

使用Java API实现上述在Kibana中的各项操作

根据某个字段分组

public class RestHighLevelClientForAggs {
    public static void main(String[] args) {
        RestHighLevelClient esClient = Client.getClient();
        //基于terms 类型聚合 基于字段进行分组聚合
        SearchRequest request = new SearchRequest("fruit");
        SearchSourceBuilder sourceBuilder = new SearchSourceBuilder();
        sourceBuilder
            .query(QueryBuilders.matchAllQuery())//查询条件
            //用来设置聚合处理
         	.aggregation(AggregationBuilders.terms("price_group").field("price"))
            .size(0);
        request.source(sourceBuilder);
        SearchResponse response = null;
        try {
            response = esClient.search(request, RequestOptions.DEFAULT);
            //处理聚合的结果
            Aggregations aggregations = response.getAggregations();
            ParsedDoubleTerms doubleTerms = aggregations.get("price_group");
            List<? extends Terms.Bucket> buckets = doubleTerms.getBuckets();
            for (Terms.Bucket bucket : buckets) {
                System.out.println(bucket.getKey()+" "+bucket.getDocCount());
            }
        }catch (Exception e){
            e.printStackTrace();
        }
    }
}

求最大值

public class AggregationForMax {
    public static void main(String[] args) {
        RestHighLevelClient client = Client.getClient();


        SearchRequest request = new SearchRequest("fruit");
        SearchSourceBuilder sourceBuilder = new SearchSourceBuilder();
        sourceBuilder
                .query(QueryBuilders.matchAllQuery())
                .aggregation(AggregationBuilders.max("max_price").field("price"))
                .size(0);
        request.source(sourceBuilder);
        try {
            SearchResponse searchResponse =
            client.search(request,RequestOptions.DEFAULT);
            Aggregations aggregations = searchResponse.getAggregations();
            ParsedMax maxPrice = aggregations.get("max_price");
            System.out.println(maxPrice.getValueAsString());
        } catch (IOException e) {
            e.printStackTrace();
        }
    }
}

注意： 在最终获取分组中的数据时，首先判断所求得的结果是否是Key-Value的结果，比如上述根据某个字段分组的示例从Kibana中就可以看出是Key-Value的形式，所以aggregations.get("分组名称");返回的结果应该为ParsedXXXXTerms类型，如果像求最大值、平均值、最小值等在执行到该aggregations.get("分组名称");返回的结果应该为ParsedXXX类型

求最小值

public class AggregationForMin {
    public static void main(String[] args) {
        RestHighLevelClient client = Client.getClient();
        SearchRequest searchRequest = new SearchRequest("fruit");
        SearchSourceBuilder sourceBuilder = new SearchSourceBuilder();

        sourceBuilder
                .query(QueryBuilders.matchAllQuery())
                .aggregation(AggregationBuilders.min("min_price").field("price"))
                .size(0);
        searchRequest.source(sourceBuilder);
        try {
            SearchResponse searchResponse = 
                client.search(searchRequest, RequestOptions.DEFAULT);
            Aggregations aggregations = searchResponse.getAggregations();
            ParsedMin minPrice = aggregations.get("min_price");
            System.out.println(minPrice.getValueAsString());
        } catch (IOException e) {
            e.printStackTrace();
        }
    }
}

等等一系列需求的演示和模拟，使用ES来完成数据的统计。

⑤ 子聚合

先从需求展开，先按照title进行分组，然后再对每一个分组中的成员对价格price进行降序排序

先使用命令在Kibana中实现该操作，其次再根据实现的命令转换为Java代码实现

使用命令操作进行实现

GET /fruit/_search
{
  "query": {
    "match_all": {}
  },
  "size": 0, 
  "aggs": {
    "title_group": {
      "terms": {
        "field": "title"
      },
      "aggs": {
        "sort_price": {
          "terms": {
            "field": "price",
            "order": {
              "_key": "desc"
            }
          }
        }
      }
    }
  }
}

将实现的命令转换为Java流程

public class AggregationForSub {
    public static void main(String[] args) {
        RestHighLevelClient client = Client.getClient();
        SearchRequest searchRequest = new SearchRequest("fruit");
        SearchSourceBuilder sourceBuilder = new SearchSourceBuilder();
        TermsAggregationBuilder termsAggregationBuilder = 
        AggregationBuilders.terms("title_group").field("title");
        TermsAggregationBuilder subAggregationBuilder = 
        AggregationBuilders.terms("price_sort").field("price").order(BucketOrder.count(false));
        //subAggregation 为子聚合
        termsAggregationBuilder.subAggregation(subAggregationBuilder);
        sourceBuilder
                .query(QueryBuilders.matchAllQuery())
                .aggregation(termsAggregationBuilder)
                .size(0);
        searchRequest.source(sourceBuilder);
        try {
            SearchResponse searchResponse = client.search(searchRequest, RequestOptions.DEFAULT);
            Aggregations aggregations = searchResponse.getAggregations();
            ParsedStringTerms titleGroup = aggregations.get("title_group");
            for (Terms.Bucket bucket : titleGroup.getBuckets()) {
                System.out.println(bucket.getKey()+"--"+bucket.getDocCount());
                Aggregations bucketAggregations = bucket.getAggregations();
                ParsedDoubleTerms priceSort = bucketAggregations.get("price_sort");
                for (Terms.Bucket priceSortBucket : priceSort.getBuckets()) {
                    System.out.println(priceSortBucket.getKey()+"--"+priceSortBucket.getDocCount());
                }
            }
        } catch (IOException e) {
            e.printStackTrace();
        }
    }
}