spark 机器学习应用 spark入门实战

 

spark 操作的几个步骤

1  数据关联  textFile 和 parallelize

2 转换操作（JavaRDD和JavaPairRDD他们可以通过mapToPair and flatMapToPair转换） 

3  action操作，获取数据结果 

 

一、wordcount的例子

 

//单词统计 
	public static void wordCount(JavaSparkContext ctx  ){
		String filePath = "e://log1.log";
		JavaRDD<String> lines = ctx.textFile(filePath, 1);

		JavaRDD<String> words = lines.flatMap(new FlatMapFunction<String, String>() {
			@Override
			public Iterable<String> call(String s) {
				return Arrays.asList(SPACE.split(s));
			}
		});

		JavaPairRDD<String, Integer> ones = words.mapToPair(new PairFunction<String, String, Integer>() {
			@Override
			public Tuple2<String, Integer> call(String s) {
				return new Tuple2<String, Integer>(s, 1);
			}
		});

		JavaPairRDD<String, Integer> counts = ones.reduceByKey(new Function2<Integer, Integer, Integer>() {
			@Override
			public Integer call(Integer i1, Integer i2) {
				return i1 + i2;
			}
		});

		
		List<Tuple2<String, Integer>> output = counts.collect();
		for (Tuple2<?, ?> tuple : output) {
			System.out.println(tuple._1() + " : " + tuple._2());
		}
	}

二、 各种Transformations 和action测试准备数据

 

  

public static void testMap(JavaSparkContext ctx) {

		List<Integer> data = Arrays.asList(1, 2, 3, 4, 5,1,2,3);
		JavaRDD<Integer> distData = ctx.parallelize(data);
		
		List<Integer> data2 = Arrays.asList(1, 2, 3, 4, 5,6);
		JavaRDD<Integer> distData2 = ctx.parallelize(data2);
		
		List<String> data3 = Arrays.asList("wang zhan,xiao ming,li xin,wang qiang,e,f".split(","));
		JavaRDD<String> data3RDD = ctx.parallelize(data3);
		
		JavaRDD<Integer> lineLengths ;
		JavaRDD<String> returnStr;
		JavaPairRDD<String, Integer> returnStr2;
		JavaPairRDD<String, Integer> returnStr3;
		
//		lineLengths = map( distData);
//		lineLengths = filter(distData );
//		lineLengths = sample(distData );
//		lineLengths = union(distData, distData2) ;
//		lineLengths = intersection(distData, distData2) ;
//		lineLengths = distinct(distData ) ;
		
//		returnStr = flatMap(data3RDD); //数据扁平打散 
//		returnStr3 = mapToPair(returnStr);// 数据变成键值对的形式 
//		reduceByKey(returnStr3); //对key进行分组计算 
		
		
//		List<Integer> list = lineLengths.collect();
//		WordCount.print(list);
		
//		List<String> list2 = returnStr.collect() ; 
//		WordCount.print(list2);
 
//		testPersist( data3RDD );
		
//		List<Tuple2<String, Integer>> listTuple = returnStr2.collect();
//		printTuple( listTuple );
		
//		returnStr3 = reduceByKey(returnStr2  );
//		List<Tuple2<String, Integer>> listTuple2 = returnStr3.collect();
//		printTuple( listTuple2 );
		//reduceByKey(returnStr2  );
		
		//action 
		//reduce( returnStr2);
		// count(returnStr2 );
	}

 

 

三、groupByKey 

//对key进行分组处理，但如果需要统计求和则最好不要这样处理 
	private static JavaPairRDD<String, Integer> groupByKey(JavaPairRDD<String, Integer> returnStr3) {
		JavaPairRDD<String, Iterable<Integer>>  rdd = returnStr3.groupByKey();
		
		return null;
	}

 四、数据去重

  

//数据去重 
	private static JavaRDD<Integer> distinct(JavaRDD<Integer> distData) {
		JavaRDD<Integer> rdd3 = distData.distinct( );
		print( rdd3);
		return null;
	}

 

五、交集数据

 

//获取rdd数据的交集 数据
	private static JavaRDD<Integer> intersection(JavaRDD<Integer> distData,
			JavaRDD<Integer> distData2) {
		JavaRDD<Integer> rdd3 = distData.intersection(distData2 ); 
		print(rdd3 ); 
		
		return null;
	}

 

六数据持久化

 

/**
	 * 持久化数据 
	 * @param data3RDD
	 */
	private static void testPersist(JavaRDD<String> data3RDD) {
		System.out.println( "持久化数据到目录。。。");
		data3RDD.persist(StorageLevel.MEMORY_ONLY());
//		data3RDD.checkpoint();  
//		data3RDD.isCheckpointed() ;
	}

七、count统计

 

//计算统计  元素两两传入到reduce中然后计算统计
	private static void count(JavaPairRDD<String, Integer> returnStr2) {
		System.out.println("元素总数："+returnStr2.count()  ); //获取元素总数
		System.out.println("元素总数："+returnStr2.first() );//获取第一个元素
		System.out.println("元素总数："+returnStr2.take(2) );//获取RDD的前2个元素
		System.out.println("元素countByKey总数："+returnStr2.countByKey( )  ); //根据key进行统计数量 
		//returnStr2.saveAsTextFile("E://test.txt") ;//数据保存到文件中  	
		returnStr2.foreach(new VoidFunction<Tuple2<String,Integer>>() {
			@Override
			public void call(Tuple2<String, Integer> t) throws Exception {
				 System.out.println( t );
			}
		});
		
	}

  

八、reduce操作

  

//计算统计  元素两两传入到reduce中然后计算统计
	private static void reduce(JavaPairRDD<String, Integer> returnStr2) {
		
		Tuple2<String, Integer> t = returnStr2.reduce(
				new Function2<Tuple2<String,Integer>, Tuple2<String,Integer>,  Tuple2<String,Integer>>() {
					@Override
					public Tuple2<String, Integer> call( Tuple2<String, Integer> v1,  Tuple2<String, Integer> v2) throws Exception {
						 System.out.println( v1 +"  "+v2 );
						
						return new Tuple2<String, Integer>(v1._1+v2._1,v1._2+v2._2()  );
					}
					
			
		} );
		System.out.println("reduce结果 ："+t._1 +"  "+t._2);
	}

 

九、reduceBykey使用

  

//将原来的RDD每一个行数据 变成一个数组，然后所有的数组数据存到一个总得RDD数组中
	public static JavaPairRDD<String, Integer> reduceByKey(JavaPairRDD<String, Integer> rdd ) {
		
		JavaPairRDD<String, Integer> counts = rdd.reduceByKey(
				new Function2<Integer, Integer, Integer>() {//泛型分别是 ：两个计算参数  ，最后是返回值 
			
			@Override
			public Integer call(Integer i1, Integer i2) {//每次把key相同的数据，与上一次执行的结果，依次传进来计算
				System.out.println(i1+" == "+i2);
				return i1 + i2;
			}
		});
		printPair(counts );
		return counts;
	}

 

十 flatMap使用

  

//将原来的RDD每一个行数据 变成一个数组，然后所有的数组数据存到一个总得RDD数组中
	public static JavaRDD<String> flatMap(JavaRDD<String> rdd  ) {
		JavaRDD<String> lineLengths = rdd.flatMap( //返回值是输出的类型
				new FlatMapFunction<String, String>() { //第一个参数是输入，第二个参数是输出
			@Override
			public Iterable<String> call(String str) throws Exception {
				 
				return Arrays.asList(SPACE.split(str));
			}
		}) ; 
		
		printStr(lineLengths );
		return lineLengths;
	}

 十一、mapToPair使用

   

// 将普通的RDD转换为 map数据的RDD方便计算处理  ， a  a->1
	public static  JavaPairRDD<String, Integer>   mapToPair(JavaRDD<String> rdd ){
		JavaPairRDD<String, Integer> ones = rdd.mapToPair( new PairFunction<String, String, Integer>() {
			@Override
			public Tuple2<String, Integer> call(String t) throws Exception {
				return  new Tuple2<String, Integer>( t, 1);
			}
		}) ; 
		return ones ; 
	}

 十二、 合并Rdd

  

// 合并rdd 
	public static JavaRDD<Integer> union(JavaRDD<Integer> rdd , JavaRDD<Integer> rdd2) {
		JavaRDD<Integer> lineLengths = rdd.union(rdd2) ; 
		print(lineLengths);
		return lineLengths;
	}

 十三、抽样

   

// Return a sampled subset of this RDD. 返回一个RDD子集抽样 
	public static JavaRDD<Integer> sample(JavaRDD<Integer> rdd) {
		JavaRDD<Integer> lineLengths = rdd.sample(false , 0.4 ) ; 
		print(lineLengths);
		return lineLengths;
	}

 十四、map使用

  

// map 源中的每一个元素都进行一个函数操作，生成一个新的RDD ，即每个元素进行一次转换
	public static JavaRDD<Integer> map(JavaRDD<Integer> rdd) {
		JavaRDD<Integer> lineLengths = rdd.map(new Function<Integer, Integer>() {
			@Override
			public Integer call(Integer v1) throws Exception {
				
				return v1+1;
			}
		});
		
		print( lineLengths);
		return lineLengths;
	}

 十五、打印数据 

   

/**
	 * 如果rdd中数据过多，则调用take获取一部分数据打印
	 * 1 直接rdd.foreach 打印rdd的数据，数据打印在各个executor中
	 * 2 调用rdd.collect.foreach 打印数据，数据打印在driver上 
	 * @param lineLengths
	 */
	public static void print(JavaRDD<Integer> lineLengths ){
//		lineLengths.foreach( new VoidFunction<Integer>() {
//			@Override
//			public void call(Integer t) throws Exception {
//			}
//		});
		System.out.println("开始打印");
		lineLengths.collect().forEach( new Consumer<Integer>(){
			@Override
			public void accept(Integer t) {
				System.out.println( t); 
			}
		});
		System.out.println("结束打印");
	}
	
	public static void printStr(JavaRDD<String> lineLengths ){
		System.out.println("开始打印");
		lineLengths.collect().forEach( new Consumer<String>(){
			@Override
			public void accept(String t) {
				System.out.println( t); 
			}
		});
		System.out.println("结束打印");
	}
	/**
	 * 打印tuble
	 * @param lineLengths
	 */
	public static void printPair( JavaPairRDD<String, Integer> rdd ){
 
		System.out.println("开始打印");
		rdd.collect().forEach( new Consumer<Tuple2<String,Integer>>(){
			@Override
			public void accept(Tuple2<String, Integer> t) {
				System.out.println( t._1() +"  "+t._2() );
			}
			
		});
		System.out.println("结束打印");
	}
	public static void print(List  list) {
		if (list == null || list.size() == 0) {
			return;
		}
		for (int i = 0; i < list.size(); i++) {
			System.out.println(list.get(i));
		}
	}
	
	/**
	 * 打印map的RDD
	 * @param listTuple
	 */
	public static void printTuple(List<Tuple2<String, Integer>> listTuple) {
		
		if (listTuple == null || listTuple.size() == 0) {
			return;
		}

		for (int i = 0; i < listTuple.size(); i++) {
			System.out.println(listTuple.get(i));
		}
	}

 十六、过滤

  

//对每一个元素进行过滤，然后返回 ,false 的数据会被过滤掉
	public static JavaRDD<Integer> filter(JavaRDD<Integer> rdd) {
		JavaRDD<Integer> lineLengths = rdd.filter(new Function<Integer, Boolean>() {
			@Override
			public Boolean call(Integer v1) throws Exception {
				if (v1 != 1) {
					return true;
				}
				return false;
			}
		});
		
		print( lineLengths);
		return lineLengths;
	}