1 导引

我们在博客《Hadoop: 单词计数(Word Count)的MapReduce实现 》中学习了如何用Hadoop-MapReduce实现单词计数，现在我们来看如何用Spark来实现同样的功能。

2. Spark的MapReudce原理

Spark框架也是MapReduce-like模型，采用“分治-聚合”策略来对数据分布进行分布并行处理。不过该框架相比Hadoop-MapReduce，具有以下两个特点：

• 对大数据处理框架的输入/输出，中间数据进行建模，将这些数据抽象为统一的数据结构命名为弹性分布式数据集(Resilient Distributed Dataset)，并在此数据结构上构建了一系列通用的数据操作，使得用户可以简单地实现复杂的数据处理流程。

• 采用了基于内存的数据聚合、数据缓存等机制来加速应用执行尤其适用于迭代和交互式应用。

Spark社区推荐用户使用Dataset、DataFrame等面向结构化数据的高层API(Structured API)来替代底层的RDD API,因为这些高层API含有更多的数据类型信息（Schema），支持SQL操作，并且可以利用经过高度优化的Spark SQL引擎来执行。不过，由于RDD API更基础，更适合用来展示基本概念和原理，后面我们的代码都使用RDD API。

Spark的RDD/dataset分为多个分区。RDD/Dataset的每一个分区都映射一个或多个数据文件， Spark通过该映射读取数据输入到RDD/dataset中。

因为我们这里采用的本地单机多线程调试模式，默认分区数即为本地机器使用的线程数，若在代码中设置了local[N]（使用N个线程），则默认为N个分区；若设为local[*]（使用本地CPU核数个线程），则默认分区数为本地CPU核数。大家可以通过调用RDD对象的getNumPartitions()查看实际分区个数。

我们下面的流程描述中，假设每个文件对应一个分区。

Spark的Map示意图如下：

Spark的Reduce示意图如下：

3. Word Count的Java实现

项目架构如下图：

Word-Count-Spark
├─ input
│  ├─ file1.txt
│  ├─ file2.txt
│  └─ file3.txt
├─ output
│  └─ result.txt
├─ pom.xml
├─ src
│  ├─ main
│  │  └─ java
│  │     └─ WordCount.java
│  └─ test
└─ target

WordCount.java文件如下：

package com.orion;

import org.apache.spark.api.java.JavaPairRDD;
import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.sql.SparkSession;

import scala.Tuple2;
import java.util.Arrays;
import java.util.List;
import java.util.regex.Pattern;
import java.io.*;
import java.nio.file.*;

public class WordCount {
	private static Pattern SPACE = Pattern.compile(" ");

	public static void main(String[] args) throws Exception {
		if (args.length != 3) {
			System.err.println("Usage: WordCount <intput directory> <output directory> <number of local threads>");
			System.exit(1);
		}
                String input_path = args[0];
                String output_path = args[1];
		int n_threads = Integer.parseInt(args[2]);

		SparkSession spark = SparkSession.builder()
			.appName("WordCount")
			.master(String.format("local[%d]", n_threads))
			.getOrCreate();

		JavaRDD<String> lines = spark.read().textFile(input_path).javaRDD();

		JavaRDD<String> words = lines.flatMap(s -> Arrays.asList(SPACE.split(s)).iterator());
		JavaPairRDD<String, Integer> ones = words.mapToPair(s -> new Tuple2<>(s, 1));
		JavaPairRDD<String, Integer> counts = ones.reduceByKey((i1, i2) -> i1 + i2);

		List<Tuple2<String, Integer>> output = counts.collect();

                String filePath = Paths.get(output_path, "result.txt").toString();
                BufferedWriter out = new BufferedWriter(new FileWriter(filePath));
		for (Tuple2<?, ?> tuple : output) {
			out.write(tuple._1() + ": " + tuple._2() + "\n");
		}
		out.close();
                spark.stop();
	}
}

pom.xml文件配置如下：

<?xml version="1.0" encoding="UTF-8"?>

<project xmlns="http://maven.apache.org/POM/4.0.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
  xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
  <modelVersion>4.0.0</modelVersion>

  <groupId>com.WordCount</groupId>
  <artifactId>WordCount</artifactId>
  <version>1.0-SNAPSHOT</version>

  <name>WordCount</name>
  <!-- FIXME change it to the project's website -->
  <url>http://www.example.com</url>

  <!-- 集中定义版本号 -->
  <properties>
    <scala.version>2.12.10</scala.version>
    <scala.compat.version>2.12</scala.compat.version>
    <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
    <project.reporting.outputEncoding>UTF-8</project.reporting.outputEncoding>
    <project.timezone>UTC</project.timezone>
    <java.version>11</java.version>
    <scoverage.plugin.version>1.4.0</scoverage.plugin.version>
    <site.plugin.version>3.7.1</site.plugin.version>
    <scalatest.version>3.1.2</scalatest.version>
    <scalatest-maven-plugin>2.0.0</scalatest-maven-plugin>
    <scala.maven.plugin.version>4.4.0</scala.maven.plugin.version>
    <maven.compiler.plugin.version>3.8.0</maven.compiler.plugin.version>
    <maven.javadoc.plugin.version>3.2.0</maven.javadoc.plugin.version>
    <maven.source.plugin.version>3.2.1</maven.source.plugin.version>
    <maven.deploy.plugin.version>2.8.2</maven.deploy.plugin.version>
    <nexus.staging.maven.plugin.version>1.6.8</nexus.staging.maven.plugin.version>
    <maven.help.plugin.version>3.2.0</maven.help.plugin.version>
    <maven.gpg.plugin.version>1.6</maven.gpg.plugin.version>
    <maven.surefire.plugin.version>2.22.2</maven.surefire.plugin.version>
    <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
    <maven.compiler.source>11</maven.compiler.source>
    <maven.compiler.target>11</maven.compiler.target>
    <spark.version>3.2.1</spark.version>
  </properties>

  <dependencies>
    <dependency>
      <groupId>junit</groupId>
      <artifactId>junit</artifactId>
      <version>4.11</version>
      <scope>test</scope>
    </dependency>
    <!--======SCALA======-->
    <dependency>
        <groupId>org.scala-lang</groupId>
        <artifactId>scala-library</artifactId>
        <version>${scala.version}</version>
        <scope>provided</scope>
    </dependency>
    <!-- https://mvnrepository.com/artifact/org.apache.spark/spark-core -->
    <dependency>
        <groupId>org.apache.spark</groupId>
        <artifactId>spark-core_2.12</artifactId>
        <version>${spark.version}</version>
    </dependency>
    <!-- https://mvnrepository.com/artifact/org.apache.spark/spark-core -->
    <dependency> <!-- Spark dependency -->
        <groupId>org.apache.spark</groupId>
        <artifactId>spark-sql_2.12</artifactId>
        <version>${spark.version}</version>
        <scope>provided</scope>
    </dependency>
  </dependencies>


  <build>
    <pluginManagement><!-- lock down plugins versions to avoid using Maven defaults (may be moved to parent pom) -->
      <plugins>
        <!-- clean lifecycle, see https://maven.apache.org/ref/current/maven-core/lifecycles.html#clean_Lifecycle -->
        <plugin>
          <artifactId>maven-clean-plugin</artifactId>
          <version>3.1.0</version>
        </plugin>
        <!-- default lifecycle, jar packaging: see https://maven.apache.org/ref/current/maven-core/default-bindings.html#Plugin_bindings_for_jar_packaging -->
        <plugin>
          <artifactId>maven-resources-plugin</artifactId>
          <version>3.0.2</version>
        </plugin>
        <plugin>
          <artifactId>maven-compiler-plugin</artifactId>
          <version>3.8.0</version>
        </plugin>
        <plugin>
          <artifactId>maven-surefire-plugin</artifactId>
          <version>2.22.1</version>
        </plugin>
        <plugin>
          <artifactId>maven-jar-plugin</artifactId>
          <version>3.0.2</version>
        </plugin>
        <plugin>
          <artifactId>maven-install-plugin</artifactId>
          <version>2.5.2</version>
        </plugin>
        <plugin>
          <artifactId>maven-deploy-plugin</artifactId>
          <version>2.8.2</version>
        </plugin>
        <!-- site lifecycle, see https://maven.apache.org/ref/current/maven-core/lifecycles.html#site_Lifecycle -->
        <plugin>
          <artifactId>maven-site-plugin</artifactId>
          <version>3.7.1</version>
        </plugin>
        <plugin>
          <artifactId>maven-project-info-reports-plugin</artifactId>
          <version>3.0.0</version>
        </plugin>
        <plugin>
          <artifactId>maven-compiler-plugin</artifactId>
          <version>3.8.0</version>
          <configuration>
              <source>11</source>
              <target>11</target>
              <fork>true</fork>
              <executable>/Library/Java/JavaVirtualMachines/jdk-11.0.15.jdk/Contents/Home/bin/javac</executable>
          </configuration>
        </plugin>
      </plugins>
    </pluginManagement>
  </build>
</project>

记得配置输入参数input、output、3分别代表输入目录、输出目录和使用本地线程数（在VSCode中在launch.json文件中配置）。编译运行后可在output目录下查看result.txt：

Tom: 1
Hello: 3
Goodbye: 1
World: 2
David: 1

可见成功完成了单词计数功能。

4. Word Count的Python实现

先使用pip按照pyspark==3.8.2：

pip install pyspark==3.8.2

注意PySpark只支持Java 8/11，请勿使用更高级的版本。这里我使用的是Java 11。运行java -version可查看本机Java版本。

(base) orion-orion@MacBook-Pro ~ % java -version
java version "11.0.15" 2022-04-19 LTS
Java(TM) SE Runtime Environment 18.9 (build 11.0.15+8-LTS-149)
Java HotSpot(TM) 64-Bit Server VM 18.9 (build 11.0.15+8-LTS-149, mixed mode)

项目架构如下：

Word-Count-Spark
├─ input
│  ├─ file1.txt
│  ├─ file2.txt
│  └─ file3.txt
├─ output
│  └─ result.txt
├─ src
│  └─ word_count.py

word_count.py编写如下：

from pyspark.sql import SparkSession
import sys
import os
from operator import add

if len(sys.argv) != 4:
    print("Usage: WordCount <intput directory> <output directory> <number of local threads>", file=sys.stderr)
    exit(1)
     
input_path, output_path, n_threads = sys.argv[1], sys.argv[2], int(sys.argv[3])

spark = SparkSession.builder.appName("WordCount").master("local[%d]" % n_threads).getOrCreate()

lines = spark.read.text(input_path).rdd.map(lambda r: r[0])

counts = lines.flatMap(lambda s: s.split(" "))\
    .map(lambda word: (word, 1))\
    .reduceByKey(add)

output = counts.collect()

with open(os.path.join(output_path, "result.txt"), "wt") as f:
    for (word, count) in output:
        f.write(str(word) +": " + str(count) + "\n")

spark.stop()

使用python word_count.py input output 3运行后，可在output中查看对应的输出文件result.txt：

Hello: 3
World: 2
Goodbye: 1
David: 1
Tom: 1

可见成功完成了单词计数功能。

参考

• [1] Spark官方文档： Quick Start
• [2] 许利杰，方亚芬. 大数据处理框架Apache Spark设计与实现[M]. 电子工业出版社, 2021.
• [3] GiHub: Spark官方Java样例
• [4] similarface： Spark数据分区数量的原理