Elasticsearch AGG

package elasticsearch;

import org.elasticsearch.action.index.IndexResponse;
import org.elasticsearch.action.search.SearchRequestBuilder;
import org.elasticsearch.action.search.SearchResponse;
import org.elasticsearch.client.transport.TransportClient;
import org.elasticsearch.common.settings.Settings;
import org.elasticsearch.common.transport.InetSocketTransportAddress;
import org.elasticsearch.search.aggregations.Aggregation;
import org.elasticsearch.search.aggregations.AggregationBuilders;
import org.elasticsearch.search.aggregations.bucket.terms.StringTerms;
import org.elasticsearch.search.aggregations.bucket.terms.Terms;
import org.elasticsearch.search.aggregations.bucket.terms.TermsAggregationBuilder;
import org.elasticsearch.search.aggregations.metrics.avg.AvgAggregationBuilder;
import org.elasticsearch.search.aggregations.metrics.avg.InternalAvg;
import org.elasticsearch.search.aggregations.metrics.max.InternalMax;
import org.elasticsearch.search.aggregations.metrics.max.MaxAggregationBuilder;
import org.elasticsearch.search.aggregations.metrics.sum.InternalSum;
import org.elasticsearch.search.aggregations.metrics.sum.SumAggregationBuilder;
import org.elasticsearch.transport.client.PreBuiltTransportClient;
import org.junit.Before;
import org.junit.Test;

import java.io.IOException;
import java.net.InetAddress;
import java.util.Iterator;
import java.util.Map;

import static org.elasticsearch.common.xcontent.XContentFactory.jsonBuilder;

public class AGG {

    private TransportClient client = null;
    @Before
    public void init() throws Exception {
        Settings settings = Settings.builder()
                .put("cluster.name", "my-application")
//                .put("client.transport.sniff", true)//可以通过当前指定的节点获取所有es节点的信息
                .build();
        client = new PreBuiltTransportClient(settings).addTransportAddresses(//创建client
                new InetSocketTransportAddress(InetAddress.getByName("192.168.1.241"), 9300));
    }

    /**
     * curl -XPUT 'http://192.168.1.241:9200/user_info/user/1' -d '{ "name": "curry", "age": 29, "salary": 3500,"team": "war", "position": "pg"}'
     * curl -XPUT 'http://192.168.1.241:9200/user_info/user/2' -d '{ "name": "thompson", "age": 26, "salary": 2000,"team": "war", "position": "pg"}'
     * curl -XPUT 'http://192.168.1.241:9200/user_info/user/3' -d '{ "name": "irving", "age": 25, "salary": 2000,"team": "cav", "position": "pg"}'
     * curl -XPUT 'http://192.168.1.241:9200/user_info/user/4' -d '{ "name": "green", "age": 26, "salary": 2000,"team": "war", "position": "pf"}'
     * curl -XPUT 'http://192.168.1.241:9200/user_info/user/5' -d '{ "name": "james", "age": 33, "salary": 4000,"team": "cav", "position": "sf"}'
     */
    @Test
    public void testAdduser() throws IOException {
        IndexResponse response = client.prepareIndex("user_info", "user", "1")
                .setSource(
                        jsonBuilder()
                                .startObject()
                                .field("name", "James")
                                .field("age", 33)
                                .field("salary", 3000)
                                .field("team", "cav")
                                .field("position", "sf")
                                .endObject()
                ).get();
    }

    /**
     * https://elasticsearch.cn/article/102
     * select team, count(*) as user_count from user group by team;
     */
    @Test
    public void testAgg1() {
        //指定索引和type
        SearchRequestBuilder builder = client.prepareSearch("user_info").setTypes("user");
        //按team分组然后聚合,但是并没有指定聚合函数
        TermsAggregationBuilder teamAgg = AggregationBuilders.terms("user_count").field("team");
        //添加聚合器
        builder.addAggregation(teamAgg);
        //触发
        SearchResponse response = builder.execute().actionGet();
        //将返回的结果放入到一个map中
        Map<String, Aggregation> aggMap = response.getAggregations().getAsMap();
        //取出聚合属性
        StringTerms terms = (StringTerms) aggMap.get("user_count");
        Iterator<Terms.Bucket> teamBucketIt = terms.getBuckets().iterator();
        while (teamBucketIt .hasNext()) {
            Terms.Bucket bucket = teamBucketIt.next();
            String team = (String) bucket.getKey();
            long count = bucket.getDocCount();
            System.out.println(team + " " + count);
        }
    }

    /**
     * select team, position, count(*) as pos_count from user group by team, position;
     */
    @Test
    public void testAgg2() {
        SearchRequestBuilder builder = client.prepareSearch("user_info").setTypes("user");
        //指定别名和分组的字段
        TermsAggregationBuilder teamAgg = AggregationBuilders.terms("team_name").field("team");
        TermsAggregationBuilder posAgg= AggregationBuilders.terms("pos_count").field("position");
        //添加两个聚合构建器
        builder.addAggregation(teamAgg.subAggregation(posAgg));
        //执行查询
        SearchResponse response = builder.execute().actionGet();
        //将查询结果放入map中
        Map<String, Aggregation> aggMap = response.getAggregations().getAsMap();
        //根据属性名到map中查找
        StringTerms teams = (StringTerms) aggMap.get("team_name");
        //循环查找结果
        for (Terms.Bucket teamBucket : teams.getBuckets()) {
            //先按球队进行分组
            String team = (String) teamBucket.getKey();
            Map<String, Aggregation> subAggMap = teamBucket.getAggregations().getAsMap();
            StringTerms positions = (StringTerms) subAggMap.get("pos_count");
            //因为一个球队有很多位置,那么还要依次拿出位置信息
            for (Terms.Bucket posBucket : positions.getBuckets()) {
                //拿到位置的名字
                String pos = (String) posBucket.getKey();
                //拿出该位置的数量
                long docCount = posBucket.getDocCount();
                //打印球队,位置,人数
                System.out.println(team + " " + pos + " " + docCount);
            }
        }
    }

    /**
     * select team, max(age) as max_age from user group by team;
     */
    @Test
    public void testAgg3() {
        SearchRequestBuilder builder = client.prepareSearch("user_info").setTypes("user");
        //指定安球队进行分组
        TermsAggregationBuilder teamAgg = AggregationBuilders.terms("team_name").field("team");
        //指定分组求最大值
        MaxAggregationBuilder maxAgg = AggregationBuilders.max("max_age").field("age");
        //分组后求最大值
        builder.addAggregation(teamAgg.subAggregation(maxAgg));
        //查询
        SearchResponse response = builder.execute().actionGet();
        Map<String, Aggregation> aggMap = response.getAggregations().getAsMap();
        //根据team属性,获取map中的内容
        StringTerms teams = (StringTerms) aggMap.get("team_name");
        for (Terms.Bucket teamBucket : teams.getBuckets()) {
            //分组的属性名
            String team = (String) teamBucket.getKey();
            //在将聚合后取最大值的内容取出来放到map中
            Map<String, Aggregation> subAggMap = teamBucket.getAggregations().getAsMap();
            //取分组后的最大值
            InternalMax ages = (InternalMax)subAggMap.get("max_age");
            double max = ages.getValue();
            System.out.println(team + " " + max);
        }
    }

    /**
     * select team, avg(age) as avg_age, sum(salary) as total_salary from user group by team;
     */
    @Test
    public void testAgg4() {
        SearchRequestBuilder builder = client.prepareSearch("user_info").setTypes("user");
        //指定分组字段
        TermsAggregationBuilder termsAgg = AggregationBuilders.terms("team_name").field("team");
        //指定聚合函数是求平均数据
        AvgAggregationBuilder avgAgg = AggregationBuilders.avg("avg_age").field("age");
        //指定另外一个聚合函数是求和
        SumAggregationBuilder sumAgg = AggregationBuilders.sum("total_salary").field("salary");
        //分组的聚合器关联了两个聚合函数
        builder.addAggregation(termsAgg.subAggregation(avgAgg).subAggregation(sumAgg));
        SearchResponse response = builder.execute().actionGet();
        Map<String, Aggregation> aggMap = response.getAggregations().getAsMap();
        //按分组的名字取出数据
        StringTerms teams = (StringTerms) aggMap.get("team_name");
        for (Terms.Bucket teamBucket : teams.getBuckets()) {
            //获取球队名字
            String team = (String) teamBucket.getKey();
            Map<String, Aggregation> subAggMap = teamBucket.getAggregations().getAsMap();
            //根据别名取出平均年龄
            InternalAvg avgAge = (InternalAvg)subAggMap.get("avg_age");
            //根据别名取出薪水总和
            InternalSum totalSalary = (InternalSum)subAggMap.get("total_salary");
            double avgAgeValue = avgAge.getValue();
            double totalSalaryValue = totalSalary.getValue();
            System.out.println(team + " " + avgAgeValue + " " + totalSalaryValue);
        }
    }


    /**
     * select team, sum(salary) as total_salary from user group by team order by total_salary desc;
     */
    @Test
    public void testAgg5() {
        SearchRequestBuilder builder = client.prepareSearch("user_info").setTypes("user");
        //按team进行分组,然后指定排序规则
        TermsAggregationBuilder termsAgg = AggregationBuilders.terms("team_name").field("team").order(Terms.Order.aggregation("total_salary ", true));
        SumAggregationBuilder sumAgg = AggregationBuilders.sum("total_salary").field("salary");
        builder.addAggregation(termsAgg.subAggregation(sumAgg));
        SearchResponse response = builder.execute().actionGet();
        Map<String, Aggregation> aggMap = response.getAggregations().getAsMap();
        StringTerms teams = (StringTerms) aggMap.get("team_name");
        for (Terms.Bucket teamBucket : teams.getBuckets()) {
            String team = (String) teamBucket.getKey();
            Map<String, Aggregation> subAggMap = teamBucket.getAggregations().getAsMap();
            InternalSum totalSalary = (InternalSum)subAggMap.get("total_salary");
            double totalSalaryValue = totalSalary.getValue();
            System.out.println(team + " " + totalSalaryValue);
        }
    }

}

评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值