Question

我尝试在spark中读取csv文件，我想分割以逗号分隔的行，以便我有一个带有二维数组的RDD。我是Spark的新手。

我试着这样做：

public class SimpleApp 
{   
    public static void main(String[] args) throws Exception 
    {       
        String master = "local[2]";
        String csvInput = "/home/userName/Downloads/countrylist.csv";
        String csvOutput = "/home/userName/Downloads/countrylist";

        JavaSparkContext sc = new JavaSparkContext(master, "loadwholecsv", System.getenv("SPARK_HOME"), System.getenv("JARS"));

        JavaRDD<String> csvData = sc.textFile(csvInput, 1);
        JavaRDD<String> words = csvData.map(new Function <List<String>>() { //line 43
              @Override
              public List<String> call(String s) {
                return Arrays.asList(s.split("\\s*,\\s*"));
              }
            });

        words.saveAsTextFile(csvOutput);
    }
}

这应该拆分行并返回一个ArrayList。但我不确定这一点。我收到这个错误：

SimpleApp.java:[43,58] wrong number of type arguments; required 2

Answer 1

因此该计划存在两个小问题。首先，你可能想要flatMap而不是map，因为你试图返回单词的RDD而不是单词列表的RDD，我们可以使用flatMap来平坦化结果。另一个是，我们的函数类还需要调用它的输入类型。我用以下内容替换了JavaRDD字样......

JavaRDD<String> words = rdd.flatMap(
  new FlatMapFunction<String, String>() { public Iterable<String> call(String s) {
      return Arrays.asList(s.split("\\s*,\\s*"));
    }});

Answer 2

这是你应该做的......

        //======Using flatMap(RDD of words)==============
       JavaRDD<String> csvData = spark.textFile(GlobalConstants.STR_INPUT_FILE_PATH, 1);
        JavaRDD<String> counts = csvData.flatMap(new FlatMapFunction<String, String>() {
          //line 43
              @Override
              public Iterable<String> call(String s) {
                return Arrays.asList(s.split("\\s*,\\s*"));
              }
            });     

    //======Using map(RDD of Lists of words)==============
        JavaRDD<String> csvData = spark.textFile(GlobalConstants.STR_INPUT_FILE_PATH, 1);
        JavaRDD<List<String>> counts = csvData.map(new Function <String, List<String>>() { //line 43
              @Override
              public List<String> call(String s) {
                return Arrays.asList(s.split("\\s*,\\s*"));
              }
            }); 

    //=====================================

    counts.saveAsTextFile(GlobalConstants.STR_OUTPUT_FILE_PATH);

Answer 3

这是Java https://opencredo.com/data-analytics-using-cassandra-and-spark/教程中的代码示例。

Scala代码：

  /* 1*/    val includedStatuses = Set("COMPLETED", "REPAID")
/* 2*/    val now = new Date();
/* 3*/    sc.cassandraTable("cc", "cc_transactions")
/* 4*/      .select("customerid", "amount", "card", "status", "id")
/* 5*/      .where("id < minTimeuuid(?)", now)
/* 6*/      .filter(includedStatuses contains _.getString("status"))
/* 7*/      .keyBy(row => (row.getString("customerid"), row.getString("card")))
/* 8*/      .map { case (key, value) => (key, value.getInt("amount")) }
/* 9*/      .reduceByKey(_ + _)
/*10*/      .map { case ((customerid, card), balance) => (customerid, card, balance, now) }
/*11*/      .saveToCassandra("cc", "cc_balance", SomeColumns("customerid", "card", "balance", "updated_at"))

Java代码：

SparkContextJavaFunctions functions = CassandraJavaUtil.javaFunctions(ProjectPropertie.context);
        JavaRDD<Balance> balances = functions.cassandraTable(ProjectPropertie.KEY_SPACE, Transaction.TABLE_NAME)
                .select("customerid", "amount", "card", "status", "id")
                .where("id < minTimeuuid(?)", date)
                .filter( row -> row.getString("status").equals("COMPLETED") )
                .keyBy(row -> new Tuple2<>(row.getString("customerid"), row.getString("card")))
                .mapToPair( row -> new Tuple2<>(row._1,row._2.getInt("amount")))
                .reduceByKey( (i1,i2) -> i1.intValue()+i2.intValue())
                .flatMap(new FlatMapFunction<Tuple2<Tuple2<String, String>, Integer>, Balance>() {

                    /**
                     * 
                     */
                    private static final long serialVersionUID = 1L;

                    @Override
                    public Iterator<Balance> call(Tuple2<Tuple2<String, String>, Integer> r) throws Exception {
                        List<Balance> list = new ArrayList<Balance>();
                        list.add(new Balance(r._1._1, r._1._2, r._2,reportDate));
                        return list.iterator();
                    }
                }).cache();

ProjectPropertie.context是SparkContext的地方这是获取SparkContext的方法（每个JVM仅应使用一个上下文）：

   SparkConf conf = new SparkConf(true).setAppName("App_name").setMaster("local[2]").set("spark.executor.memory", "1g")
                            .set("spark.cassandra.connection.host", "127.0.0.1,172.17.0.2")
                            .set("spark.cassandra.connection.port", "9042")
                            .set("spark.cassandra.auth.username", "cassandra")
                            .set("spark.cassandra.auth.password", "cassandra");
   SparkContext context = new SparkContext(conf);

对于数据源，我使用的是Cassandra，其中172.17.0.2是运行我的Cassandra节点的docker容器，而127.0.0.1是主机（在这种情况下是本地）

如何在SPARK中使用Java的map-function

3 个答案: