flink有隐式转换吗 flink应用

转载

archangle 2024-08-17 16:09:18

文章标签 flink有隐式转换吗 java git github 文章分类 架构后端开发

在《Flink1.7从安装到体验》一文中，我们安装和体验了Flink，今天就用java来一起开发一个简单的Flink应用；

步骤列表

本次实战经历以下步骤：

创建应用；
编码；
构建；
提交任务到Flink，验证功能；

环境信息

Flink：1.7；
Flink所在机器的操作系统：CentOS Linux release 7.5.1804；
开发环境JDK：1.8.0_181；
开发环境Maven：3.5.0；

应用功能简介

在《Flink1.7从安装到体验》一文中，我们在Flink运行SocketWindowWordCount.jar，实现的功能是从socket读取字符串，将其中的每个单词的数量统计出来，今天我们就来编码开发这个应用，实现此功能；

创建应用

应用基本代码是通过mvn命令创建的，在命令行输入以下命令：

<span style="color:#000000"><code class="language-shell">mvn archetype:generate -DarchetypeGroupId<span style="color:#669900">=</span>org.apache.flink -DarchetypeArtifactId<span style="color:#669900">=</span>flink-quickstart-java -DarchetypeVersion<span style="color:#669900">=</span>1.7.0
</code></span>

按控制台的提示输入groupId、artifactId、version、package等信息，一路回车确认后，会生成一个和你输入的artifactId同名的文件夹，里面是个maven工程：

<span style="color:#000000"><code class="language-shell">Define value <span style="color:#c678dd">for</span> property <span style="color:#669900">'groupId'</span><span style="color:#c678dd">:</span> com.bolingcavalry
Define value <span style="color:#c678dd">for</span> property <span style="color:#669900">'artifactId'</span><span style="color:#c678dd">:</span> socketwordcountdemo
Define value <span style="color:#c678dd">for</span> property <span style="color:#669900">'version'</span> 1.0-SNAPSHOT: <span style="color:#c678dd">:</span>
Define value <span style="color:#c678dd">for</span> property <span style="color:#669900">'package'</span> com.bolingcavalry: <span style="color:#c678dd">:</span>
Confirm properties configuration:
groupId: com.bolingcavalry
artifactId: socketwordcountdemo
version: 1.0-SNAPSHOT
package: com.bolingcavalry
</code></span>

用IEDA导入这个maven工程，如下图，已经有了两个类：BatchJob和StreamingJob，BatchJob是用于批处理的，本次实战用不上，因此可以删除，只保留流处理的StreamingJob：

应用创建成功，接下来可以开始编码了；

编码

您可以选择直接从GitHub下载这个工程的源码，地址和链接信息如下表所示：

名称	链接	备注
项目主页	https://github.com/zq2599/blog_demos	该项目在GitHub上的主页
git仓库地址(https)	https://github.com/zq2599/blog_demos.git	该项目源码的仓库地址，https协议
git仓库地址(ssh)	git@github.com:zq2599/blog_demos.git	该项目源码的仓库地址，ssh协议

这个git项目中有多个文件夹，本章源码在socketwordcountdemo这个文件夹下，如下图红框所示：

flink有隐式转换吗 flink应用_java

接下来开始编码：

在StreamingJob类中添加静态内部类WordWithCount，这是个PoJo，用来保存一个具体的单词及其出现频率：

<span style="color:#000000"><code class="language-java">	 <span style="color:#5c6370">/**
	 * 记录单词及其出现频率的Pojo
	 */</span>
	<span style="color:#c678dd">public</span> <span style="color:#c678dd">static</span> <span style="color:#c678dd">class</span> WordWithCount <span style="color:#999999">{</span>
		<span style="color:#5c6370">/**
		 * 单词内容
		 */</span>
		<span style="color:#c678dd">public</span> String word<span style="color:#999999">;</span>

		<span style="color:#5c6370">/**
		 * 出现频率
		 */</span>
		<span style="color:#c678dd">public</span> <span style="color:#c678dd">long</span> count<span style="color:#999999">;</span>

		<span style="color:#c678dd">public</span> <span style="color:#61aeee">WordWithCount</span><span style="color:#999999">(</span><span style="color:#999999">)</span> <span style="color:#999999">{</span>
			<span style="color:#c678dd">super</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
		<span style="color:#999999">}</span>

		<span style="color:#c678dd">public</span> <span style="color:#61aeee">WordWithCount</span><span style="color:#999999">(</span>String word<span style="color:#999999">,</span> <span style="color:#c678dd">long</span> count<span style="color:#999999">)</span> <span style="color:#999999">{</span>
			<span style="color:#c678dd">this</span><span style="color:#999999">.</span>word <span style="color:#669900">=</span> word<span style="color:#999999">;</span>
			<span style="color:#c678dd">this</span><span style="color:#999999">.</span>count <span style="color:#669900">=</span> count<span style="color:#999999">;</span>
		<span style="color:#999999">}</span>

		<span style="color:#5c6370">/**
		 * 将单词内容和频率展示出来
		 * @return
		 */</span>
		<span style="color:#999999">@Override</span>
		<span style="color:#c678dd">public</span> String <span style="color:#61aeee">toString</span><span style="color:#999999">(</span><span style="color:#999999">)</span> <span style="color:#999999">{</span>
			<span style="color:#c678dd">return</span> word <span style="color:#669900">+</span> <span style="color:#669900">" : "</span> <span style="color:#669900">+</span> count<span style="color:#999999">;</span>
		<span style="color:#999999">}</span>
	<span style="color:#999999">}</span>
</code></span>

把所有业务逻辑写在StreamJob类的main方法中，如下所示，关键位置都加了中文注释：

<span style="color:#000000"><code class="language-java"><span style="color:#c678dd">public</span> <span style="color:#c678dd">static</span> <span style="color:#c678dd">void</span> <span style="color:#61aeee">main</span><span style="color:#999999">(</span>String<span style="color:#999999">[</span><span style="color:#999999">]</span> args<span style="color:#999999">)</span> <span style="color:#c678dd">throws</span> Exception <span style="color:#999999">{</span>

		<span style="color:#5c6370">//环境信息</span>
		<span style="color:#c678dd">final</span> StreamExecutionEnvironment env <span style="color:#669900">=</span> StreamExecutionEnvironment<span style="color:#999999">.</span><span style="color:#61aeee">getExecutionEnvironment</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">;</span>

		<span style="color:#5c6370">//数据来源是本机9999端口，换行符分隔，您也可以考虑将hostname和port参数通过main方法的入参传入</span>
		DataStream<span style="color:#61aeee"><span style="color:#999999"><</span>String<span style="color:#999999">></span></span> text <span style="color:#669900">=</span> env<span style="color:#999999">.</span><span style="color:#61aeee">socketTextStream</span><span style="color:#999999">(</span><span style="color:#669900">"localhost"</span><span style="color:#999999">,</span> <span style="color:#98c379">9999</span><span style="color:#999999">,</span> <span style="color:#669900">"\n"</span><span style="color:#999999">)</span><span style="color:#999999">;</span>

		<span style="color:#5c6370">//通过text对象转换得到新的DataStream对象，</span>
		<span style="color:#5c6370">//转换逻辑是分隔每个字符串，取得的所有单词都创建一个WordWithCount对象</span>
		DataStream<span style="color:#61aeee"><span style="color:#999999"><</span>WordWithCount<span style="color:#999999">></span></span> windowCounts <span style="color:#669900">=</span> text<span style="color:#999999">.</span><span style="color:#61aeee">flatMap</span><span style="color:#999999">(</span><span style="color:#c678dd">new</span> FlatMapFunction<span style="color:#61aeee"><span style="color:#999999"><</span>String<span style="color:#999999">,</span> WordWithCount<span style="color:#999999">></span></span><span style="color:#999999">(</span><span style="color:#999999">)</span> <span style="color:#999999">{</span>
			<span style="color:#999999">@Override</span>
			<span style="color:#c678dd">public</span> <span style="color:#c678dd">void</span> <span style="color:#61aeee">flatMap</span><span style="color:#999999">(</span>String s<span style="color:#999999">,</span> Collector<span style="color:#61aeee"><span style="color:#999999"><</span>WordWithCount<span style="color:#999999">></span></span> collector<span style="color:#999999">)</span> <span style="color:#c678dd">throws</span> Exception <span style="color:#999999">{</span>
				<span style="color:#c678dd">for</span><span style="color:#999999">(</span>String word <span style="color:#669900">:</span> s<span style="color:#999999">.</span><span style="color:#61aeee">split</span><span style="color:#999999">(</span><span style="color:#669900">"\\s"</span><span style="color:#999999">)</span><span style="color:#999999">)</span><span style="color:#999999">{</span>
					collector<span style="color:#999999">.</span><span style="color:#61aeee">collect</span><span style="color:#999999">(</span><span style="color:#c678dd">new</span> WordWithCount<span style="color:#999999">(</span>word<span style="color:#999999">,</span> <span style="color:#98c379">1</span>L<span style="color:#999999">)</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
				<span style="color:#999999">}</span>
			<span style="color:#999999">}</span>
		<span style="color:#999999">}</span><span style="color:#999999">)</span>
		<span style="color:#999999">.</span><span style="color:#61aeee">keyBy</span><span style="color:#999999">(</span><span style="color:#669900">"word"</span><span style="color:#999999">)</span><span style="color:#5c6370">//key为word字段</span>
		<span style="color:#999999">.</span><span style="color:#61aeee">timeWindow</span><span style="color:#999999">(</span>Time<span style="color:#999999">.</span><span style="color:#61aeee">seconds</span><span style="color:#999999">(</span><span style="color:#98c379">5</span><span style="color:#999999">)</span><span style="color:#999999">)</span>	<span style="color:#5c6370">//五秒一次的翻滚时间窗口</span>
		<span style="color:#999999">.</span><span style="color:#61aeee">reduce</span><span style="color:#999999">(</span><span style="color:#c678dd">new</span> ReduceFunction<span style="color:#61aeee"><span style="color:#999999"><</span>WordWithCount<span style="color:#999999">></span></span><span style="color:#999999">(</span><span style="color:#999999">)</span> <span style="color:#999999">{</span> <span style="color:#5c6370">//reduce策略</span>
			<span style="color:#999999">@Override</span>
			<span style="color:#c678dd">public</span> WordWithCount <span style="color:#61aeee">reduce</span><span style="color:#999999">(</span>WordWithCount a<span style="color:#999999">,</span> WordWithCount b<span style="color:#999999">)</span> <span style="color:#c678dd">throws</span> Exception <span style="color:#999999">{</span>
				<span style="color:#c678dd">return</span> <span style="color:#c678dd">new</span> WordWithCount<span style="color:#999999">(</span>a<span style="color:#999999">.</span>word<span style="color:#999999">,</span> a<span style="color:#999999">.</span>count<span style="color:#669900">+</span>b<span style="color:#999999">.</span>count<span style="color:#999999">)</span><span style="color:#999999">;</span>
			<span style="color:#999999">}</span>
		<span style="color:#999999">}</span><span style="color:#999999">)</span><span style="color:#999999">;</span>


		<span style="color:#5c6370">//单线程输出结果</span>
		windowCounts<span style="color:#999999">.</span><span style="color:#61aeee">print</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">.</span><span style="color:#61aeee">setParallelism</span><span style="color:#999999">(</span><span style="color:#98c379">1</span><span style="color:#999999">)</span><span style="color:#999999">;</span>

		<span style="color:#5c6370">// 执行</span>
		env<span style="color:#999999">.</span><span style="color:#61aeee">execute</span><span style="color:#999999">(</span><span style="color:#669900">"Flink Streaming Java API Skeleton"</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
	<span style="color:#999999">}</span>
</code></span>

构建

<span style="color:#000000"><code class="language-shell">mvn clean package -U
</code></span>

命令执行完毕后，在target目录下的socketwordcountdemo-1.0-SNAPSHOT.jar文件就是构建成功的jar包；

在Flink验证

Flink的安装和启动请参考《Flink1.7从安装到体验》；
登录到Flink所在机器，执行以下命令：

<span style="color:#000000"><code class="language-shell">nc -l 9999
</code></span>

我这边Flink所在机器的IP地址是192.168.1.103，因此用浏览器访问的Flink的web地址为：http://192.168.1.103:8081；
选择刚刚生成的jar文件作为一个新的任务，如下图：
点击下图红框中的"upload"，将文件提交：
目前还只是将jar文件上传了而已，接下来就是手工设置执行类并启动任务，操作如下图，红框2中填写的前面编写的StreamingJob类的完整名称：
提交后的页面效果如下图所示，可见一个job已经在运行中了：
回到Flink所在机器的控制台，在之前输入了nc -l 9999的窗口输入一些英文句子，然后按下回车键，例如：

<span style="color:#000000"><code class="language-shell"><span style="color:#999999">[</span>root@vostro flink-1.7.0<span style="color:#999999">]</span><span style="color:#5c6370"># ./bin/start-cluster.sh</span>
Starting cluster.
Starting standalonesession daemon on host vostro.
Starting taskexecutor daemon on host vostro.
<span style="color:#999999">[</span>root@vostro flink-1.7.0<span style="color:#999999">]</span><span style="color:#5c6370"># nc -l 9999</span>
Apache Flink is a framework and distributed processing engine <span style="color:#c678dd">for</span> stateful computations over unbounded and bounded data streams. Flink has been designed to run <span style="color:#c678dd">in</span> all common cluster environments, perform computations at in-memory speed and at any scale.

</code></span>