将 JavaPairRDD 转换为 JavaRDD

Question

I am fetching data from ElsticSearch using ElasticSearch-Hadoop Library.

JavaPairRDD<String, Map<String, Object>> esRDD = JavaEsSpark.esRDD(sc);

Now I have JavaPairRDD. I want to use Random Forest from MLLib on this RDD. So I am converting it to JavaPairRDD.toRDD(esRDD) this will give me RDD. Using RDD I am converting again to JavaRDD

JavaRDD<LabeledPoint>[] splits = (JavaRDD.fromRDD(JavaPairRDD.toRDD(esRDD),
            esRDD.classTag())).randomSplit(new double[] { 0.5, 0.5 });

JavaRDD<LabeledPoint> trainingData = splits[0];
JavaRDD<LabeledPoint> testData = splits[1];

I want to pass trainingData and TestData to Random Forest algorithm but it gives casting exception at compile time.

Type mismatch: cannot convert from JavaRDD[Tuple2[String,Map[String,Object]]][] to JavaRDD[LabeledPoint][]

Added square brackets as less than and greater than signs are not working

Could any one suggest me the proper way for Casting. I am new to Spark Datastrucutres.

Answer 1

What data do you have in the JavaPairRDD columns? A JavaPairRDD is a key/value mapping between the first and second column, unlike a normal RDD.

You possibly want to drop off the first column from the JavaPairRDD, returning just JavaRDD with just the value column.

To to this, simply run something like:

JavaRDD newRDD = esRDD.map(x => x._2);

or equivalent to create a new JavaRDD without the first column.