elasticsearch使用spark sql来实现join

elastic search是一个分布式的搜索引擎，支持对数亿的数据进行秒级的查询操作。其底层使用了lucene来进行存储，对lucene进行了分布式的封装，同时在数据进入时进行了translog以实现fail over。

在将elastic search时当做数据库使用时，必然会遇到join操作。

这里提供spark sql来实现join的一种思路。

spark是一个通用的分布式处理框架，包括但不限于数据的读、写、流式计算等操作。使用spark，可以将自己的业务逻辑，由spark框架分布在多台机器上来执行。

elastic search在2.1+的版本上，对spark提供了支持。

下面提供一些读写的例子，绝大部分内容取自于参考资料中的官方文档。

需要注意的是，spark是使用scala语言来开发的，而elasticsearch是使用java语言来开发的。本文中的例子是使用java语言来调用spark的库，官方有更简捷的使用spark语言的例子。

1. 使用spark来读写elasticsearch

import java.util.Map;
import java.util.Map.Entry;

import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaPairRDD;
import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.api.java.function.Function;
import org.elasticsearch.spark.rdd.api.java.JavaEsSpark;

import com.google.common.collect.ImmutableList;
import com.google.common.collect.ImmutableMap;

public class WriteToEs {

	public static void main(String[] args) {
		SparkConf conf = new SparkConf().setAppName("Test").setMaster("local");
		conf.set("es.index.auto.create", "true");

		JavaSparkContext jsc = new JavaSparkContext(conf);

		// 写入es
		{
			Map<String, ?> numbers = ImmutableMap.of("one", 1, "two", 2);
			Map<String, ?> airports = ImmutableMap.of("OTP", "Otopeni", "SFO",
					"San Fran");

			JavaRDD<Map<String, ?>> javaRDD = jsc.parallelize(ImmutableList.of(
					numbers, airports));
			JavaEsSpark.saveToEs(javaRDD, "spark/docs");
		}

		// 从es中读取
		{
			JavaPairRDD<String, Map<String, Object>> esRDD = JavaEsSpark.esRDD(
					jsc, "spark/docs");

			for (Entry<String, Map<String, Object>> entry : esRDD
					.collectAsMap().entrySet()) {
				System.out.println(entry.getKey());
				System.out.println(entry.getValue());
			}
		}

		// 从es中读取
		{
			JavaRDD<Map<String, Object>> esRDD = JavaEsSpark.esRDD(jsc,
					"spark/docs", "?q=one:1").values();
			// JavaRDD<Map<String, Object>> esRDD = JavaEsSpark.esRDD(jsc,
			// "spark/docs").values();

			Function<Map<String, Object>, Boolean> filter = new Function<Map<String, Object>, Boolean>() {
				public Boolean call(Map<String, Object> map) throws Exception {
					return map.containsKey("one");
				}
			};
			JavaRDD<Map<String, Object>> filtered = esRDD.filter(filter);
			for (Map<String, Object> map : filtered.collect()) {
				System.out.println(map);
			}
		}

		jsc.close();
	}
}

2. 使用spark sql来读写elasticsearch

此例子为了演示join而写，没有实际的意义。

import java.io.Serializable;
import java.util.ArrayList;
import java.util.List;

import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.sql.DataFrame;
import org.apache.spark.sql.Row;
import org.apache.spark.sql.SQLContext;
import org.elasticsearch.spark.sql.api.java.JavaEsSparkSQL;

public class SparkSql implements Serializable {

	private static final long serialVersionUID = -8843264837345502547L;

	public static class People {
		private int id;
		private String name;
		private String surname;
		private int age;

		public People(int id, String name, String surname, int age) {
			this.id = id;
			this.name = name;
			this.surname = surname;
			this.age = age;
		}

		public People() {
		}

		public int getId() {
			return id;
		}

		public void setId(int id) {
			this.id = id;
		}

		public String getName() {
			return name;
		}

		public void setName(String name) {
			this.name = name;
		}

		public String getSurname() {
			return surname;
		}

		public void setSurname(String surname) {
			this.surname = surname;
		}

		public int getAge() {
			return age;
		}

		public void setAge(int age) {
			this.age = age;
		}

	}

	public static void main(String[] args) {
		SparkConf conf = new SparkConf().setAppName("Test").setMaster("local");
		conf.set("es.index.auto.create", "true");

		JavaSparkContext jsc = new JavaSparkContext(conf);

		// 写入
		{
			SQLContext sc = new SQLContext(jsc);

			List<People> data = new ArrayList<People>();
			data.add(new People(1, "Micheal", "Mike", 18));
			data.add(new People(2, "Flowaters", "fw", 18));

			DataFrame people = sc.createDataFrame(data, People.class);

			JavaEsSparkSQL.saveToEs(people, "spark/person");
		}

		// 读取
		{
			SQLContext sql = new SQLContext(jsc);

			// 注册表people
			DataFrame people = sql.read().format("es").load("spark/people");
			people.registerTempTable("people");

			// 注册表person
			DataFrame person = sql.read().format("es").load("spark/person");
			person.registerTempTable("person");

			// 查看表的schema
			// df.printSchema();

			// 执行sql
			DataFrame df = sql
					.sql("SELECT ps.id, p.name FROM people as p INNER JOIN person as ps ON p.name = ps.name");

			// 输出执行的结果
			for (Row row : df.collectAsList()) {
				System.out.println(row.toString());
			}
		}

		jsc.close();
	}
}

参考资料：

1. elasticsearch官方网站：https://www.elastic.co/

2. elasticsearch官方对spark的支持：https://www.elastic.co/guide/en/elasticsearch/hadoop/current/spark.html