一、示例代码
List<HashMap<String, Object>> mapList = buildMapArray();
List<Tuple2<String, Iterable<HashMap<String, Object>>>> collect = javaSparkContext.parallelize(mapList).mapToPair(new PairFunction<HashMap<String, Object>, String, HashMap<String, Object>>() {
@Override
public Tuple2<String, HashMap<String, Object>> call(HashMap<String, Object> student) throws Exception {
return new Tuple2<String, HashMap<String, Object>>("x", student);
}
//会产生shuffle
}).groupByKey().collect();
LOG.info("collect : " + collect);
二、执行结果:看read和write,这是Serializable的结果
三、使用KryoRegistrator需要注意事项
public void registerClasses(Kryo kryo) {
//kryo.register(SparkSerilizeTest.Student.class);
//kryo.register(SparkSerilizeTest.class);
kryo.register(java.util.HashMap.class);
//kryo.register(java.util.List.class);
}
sparkConf.set("spark.serializer", "org.apache.spark.serializer.KryoSerializer");
sparkConf.set("spark.kryo.registrator", "com.sdyc.ndmp.wechat.job.MyRegistrator");
四、执行结果:看read和write,这是KryoRegistrator结果
所以,序列化这块能提高不少效率