Elastic实战：spring-data-elasticsearch聚合查询指南｜spring-data-elasticsearch实现各类聚合

0.引言

elasticsearch支持各种类型的聚合查询，给我们做数据统计、数据分析时提供了强大的处理能力，但是作为java开发者，如何在java client中实现这些聚合呢？

我们知道spring-data-elasticsearch提供了针对整合spring的es java client，但是在elastic、spring-data官方文档中都没有详细说明聚合查询在java client中如何实现。

所以本期，我们的目标就是一篇将这些聚合操作一网打尽！

为了更好的将这些聚合讲解清楚，我们结合es官方文档的结构，将三种类型的聚合一一讲解。但不会将每种小类型都演示一遍，相信经过几种常用类型的演示，大家自己也能推敲出其他类型的用法。如果实在写不出来的可以留言博主，我们一起来探讨

1. 运行环境

本次演示基于以下环境

spring-data-elasticsearch3.2.12.RELEASE

基础环境的搭建可参考这篇文章：

从零搭建springboot+spring data elasticsearch3.x环境

在开始讲解之前，我们先声明我们的索引结构，方便大家后续理解我们的案例

# 订单索引，一个订单下有多个商品
PUT order_test
{
  "mappings": {
    "properties": {
      // 订单状态 0未付款 1未发货 2运输中 3待签收 4已签收 
      "status": {
        "type": "integer"
      },
      // 订单编号
      "no": {
        "type": "keyword"
      },
      // 下单时间
      "create_time": {
        "type": "date",
        "format": "yyyy-MM-dd HH:mm:ss"
      },
      // 订单金额
      "amount": {
        "type": "double"
      },
      // 创建人
      "creator":{
        "type": "keyword"
      },
      // 商品信息
      "product":{
        "type": "nested",
        "properties": {
          // 商品ID
          "id": {
            "type": "keyword"
          },
          // 商品名称
          "name":{
            "type": "keyword"
          },
          // 商品价格
          "price": {
            "type": "double"
          },
          // 商品数量
          "quantity": {
            "type": "integer"
          }
        }
      }
    }
  }
}

测试数据，供大家跟练

POST order_test/_bulk
{"index":{}}
{"status":0,"no":"DD202205280001","create_time":"2022-05-01 12:00:00","amount":100.0,"creator":"张三","product":[{"id":"1","name":"苹果","price":20.0,"quantity":5}]}
{"index":{}}
{"status":0,"no":"DD202205280002","create_time":"2022-05-01 12:00:00","amount":100.0,"creator":"李四","product":[{"id":"2","name":"香蕉","price":20.0,"quantity":5}]}
{"index":{}}
{"status":1,"no":"DD202205280003","create_time":"2022-05-02 12:00:00","amount":100.0,"creator":"张三","product":[{"id":"2","name":"香蕉","price":20.0,"quantity":5}]}
{"index":{}}
{"status":2,"no":"DD202205280004","create_time":"2022-05-01 12:00:00","amount":150.0,"creator":"王二","product":[{"id":"1","name":"苹果","price":30.0,"quantity":5}]}
{"index":{}}
{"status":2,"no":"DD202205280005","create_time":"2022-05-03 12:00:00","amount":100.0,"creator":"55555","product":[{"id":"2","name":"香蕉","price":20.0,"quantity":5}]}
{"index":{}}
{"status":3,"no":"DD202205280006","create_time":"2022-05-04 12:00:00","amount":150.0,"creator":"李四","product":[{"id":"3","name":"榴莲","price":150.0,"quantity":1}]}
{"index":{}}
{"status":4,"no":"DD202205280007","create_time":"2022-05-04 12:00:00","amount":100.0,"creator":"张三","product":[{"id":"2","name":"香蕉","price":20.0,"quantity":5}]}
{"index":{}}
{"status":3,"no":"DD202205280008","create_time":"2022-05-01 12:00:00","amount":200.0,"creator":"王二","product":[{"id":"1","name":"苹果","price":40.0,"quantity":5}]}
{"index":{}}
{"status":4,"no":"DD202205280009","create_time":"2022-05-03 12:00:00","amount":100.0,"creator":"55555","product":[{"id":"2","name":"香蕉","price":20.0,"quantity":5}]}

2. 分桶聚合 Bucket aggregations 2.1 Terms aggregation

分桶聚合中最常用的就是terms聚合了，它可以按照指定字段将数据分组聚合，类似mysql中的group by

• 案例

要求统计各种状态的单数

• DSL

GET order_test/_search
{
  "size": 0,
  "aggs": {
    "status_bucket": {
      "terms": {
        "field": "status"
      }
    }
  }
}

• 执行结果

• java实现

	public void termsAgg(){
        String aggName = "status_bucket";
        NativeSearchQueryBuilder queryBuilder =  new NativeSearchQueryBuilder();
        queryBuilder.withPageable(PageRequest.of(0,1));
        TermsAggregationBuilder termsAgg = AggregationBuilders.terms(aggName).field("status");
        queryBuilder.addAggregation(termsAgg);
        Aggregations aggregations = restTemplate.query(queryBuilder.build(), SearchResponse::getAggregations);
        Terms terms = aggregations.get(aggName);
        List buckets = terms.getBuckets();
        HashMap statusRes = new HashMap();
        buckets.forEach(bucket -> {
            statusRes.put(bucket.getKeyAsString(),bucket.getDocCount());
        });
        System.out.println("---聚合结果---");
        System.out.println(statusRes);
    }

2.2 Date histogram aggregation

日期分组聚合可以按照日期进行分组，常用到一些日期趋势统计中

• 案例

统计每天的下单量

• DSL

GET order_test/_search
{
  "size": 0,
  "aggs": {
    "date": {
      "date_histogram": {
        "field": "create_time",
        "calendar_interval": "day",
        "format": "yyyy-MM-dd"
      }
    }
  }
}

• 执行结果

• java

	public void dateHistogramAgg(){
        String aggName = "date";
        DateHistogramAggregationBuilder dateHistogramAggregation = AggregationBuilders.dateHistogram(aggName).field("create_time")
                .calendarInterval(DateHistogramInterval.days(1)).format("yyyy-MM-dd");
        NativeSearchQueryBuilder queryBuilder =  new NativeSearchQueryBuilder();
        queryBuilder.withPageable(PageRequest.of(0,1)).addAggregation(dateHistogramAggregation);
        Aggregations aggregations = restTemplate.query(queryBuilder.build(), SearchResponse::getAggregations);
        ParsedDateHistogram terms = aggregations.get(aggName);
        List buckets = terms.getBuckets();
        HashMap resultMap = new HashMap();
        buckets.forEach(bucket -> {
            resultMap.put(bucket.getKeyAsString(),bucket.getDocCount());
        });
        System.out.println("---聚合结果---");
        System.out.println(resultMap);
    }

拓展：

这里大家会发现使用的是ParsedDateHistogram来承接结果，与上述的Term不一致，那么我们怎么知道什么时候该用哪个呢？实际上可以通过断点来判断

我们通过把断点截取到restTemplate.query的执行结果aggregations之后，会发现该aggregations中的元素已经标明了其类型为ParsedDateHistogram，所以大家只需要跟着用就可以了。

2.3 Range aggregation

范围分组聚合可以帮助我们按照指定的数值范围进行分组

• 案例

统计订单金额在0～100，100～200，200+ 这几个区间的订单数量

• DSL

GET order_test/_search
{
  "size": 0,
  "aggs": {
    "date_range": {
      "range": {
        "field": "amount",
        "ranges": [
          {
            "to": "100"
          },
          {
            "from": "100",
            "to": "200"
          },
          {
            "from": "200"
          }
        ]
      }
    }
  }
}

• 执行结果

• java

 	public void rangeAgg(){
        String aggName = "range";
        RangeAggregationBuilder agg = AggregationBuilders.range(aggName).field("amount").addUnboundedTo(100).addRange(100, 200).addUnboundedFrom(200);
        NativeSearchQueryBuilder queryBuilder =  new NativeSearchQueryBuilder();
        queryBuilder.withPageable(PageRequest.of(0,1)).addAggregation(agg);
        Aggregations aggregations = restTemplate.query(queryBuilder.build(), SearchResponse::getAggregations);
        ParsedRange terms = aggregations.get(aggName);
        List buckets = terms.getBuckets();
        HashMap resultMap = new HashMap();
        buckets.forEach(bucket -> {
            resultMap.put(bucket.getKeyAsString(),bucket.getDocCount());
        });
        System.out.println("---聚合结果---");
        System.out.println(resultMap);
    }

2.4 Nested aggregation

nested聚合专用于json型子对象进行聚合，比如上述案例中product是json型数组，如果当我们想通过商品中的属性来聚合统计时就需要用到nested聚合，直接使用product.name来聚合其结果则不会是我们预期的，这主要与es针对数组的存储形式有关。

• 案例

统计每种货物的订单数

• DSL

GET order_test/_search
{
  "size": 0,
  "aggs": {
    "product_nested": {
      "nested": {
        "path": "product"
      },
      "aggs": {
        "name_bucket": {
          "terms": {
            "field": "product.name"
          }
        }
      }
    }
  }
}

• 执行结果

• java：这里我们涉及到要设置一个嵌套聚合，可以通过subAggregation方法来定义子聚合

 	public void nestedAgg(){
        String aggName = "product_nested";
        String termsAggName = "name_bucket";
        NestedAggregationBuilder aggregationBuilder = AggregationBuilders.nested(aggName, "product").subAggregation(AggregationBuilders.terms(termsAggName).field("product.name"));
        NativeSearchQueryBuilder queryBuilder =  new NativeSearchQueryBuilder()
                .withPageable(PageRequest.of(0,1))
                .addAggregation(aggregationBuilder);
        Aggregations aggregations = restTemplate.query(queryBuilder.build(), SearchResponse::getAggregations);
        ParsedNested nestedRes = aggregations.get(aggName);
        Terms terms = nestedRes.getAggregations().get(termsAggName);
        List buckets = terms.getBuckets();
        HashMap resMap = new HashMap();
        buckets.forEach(bucket -> {
            resMap.put(bucket.getKeyAsString(),bucket.getDocCount());
        });
        System.out.println("---聚合结果---");
        System.out.println(resMap);
    }

3. 数值聚合 Metrics aggregations 3.1 Sum aggregations

求和聚合是常用的聚合之一，经常与分组聚合配合使用，用来统计出各组下的合计

• 案例

求5月1日销售总额

• DSL：这里我们添加了一个query语句，用来限定聚合范围是5.1日的订单

GET order_test/_search
{
  "query": {
    "range": {
      "create_time": {
        "format": "yyyy-MM-dd",
        "from": "2022-05-01",
        "to": "2022-05-01"
      }
    }
  }, 
  "size": 0,
  "aggs": {
    "sum_amount": {
      "sum": {
        "field": "amount"
      }
    }
  }
}

• 执行结果

• java

public void sumAgg(){
        String aggName = "sumAmount";
        SumAggregationBuilder agg = AggregationBuilders.sum(aggName).field("amount");
        NativeSearchQueryBuilder queryBuilder =  new NativeSearchQueryBuilder()
                .withPageable(PageRequest.of(0,1))
                .withQuery(QueryBuilders.rangeQuery("create_time").format("yyyy-MM-dd").from("2022-05-01").to("2022-05-01"))
                .addAggregation(agg);
        Aggregations aggregations = restTemplate.query(queryBuilder.build(), SearchResponse::getAggregations);
        ParsedSum metric = aggregations.get(aggName);
        double value = metric.getValue();
        System.out.println("---聚合结果---");
        System.out.println(value);
    }

3.2 Script aggregation

脚本聚合支持我们通过脚本语言来自定义聚合的数值，es中脚本默认的语言为painless。需要注意的是脚本语法非常影响性能，我们一般是尽量避免使用。同时es中还提供了专门的脚本数值聚合 script metric aggregation，但因为不太常用，所以我们这里以更加常用的聚合脚本来讲解

• 案例

求所有货物平均单价

• DSL：注意这里不能直接用product.price。因为product是数组，里面可能包含多种货物，所以应该用订单总金额除以所有订单的货物数量

GET order_test/_search
{
  "size": 0,
  "aggs": {
    "total_amount":{
      "sum": {
        "field": "amount"
      }
    },
    "total_quantity":{
      "sum": {
        "script": {
          "source": """
            int total = 0;
            for(int i=0; i