Using the java interface for java serialization

implementing your own serialization routines for your data types (e.g., using the java.io.Externalizable interface for Java Serialization, or using the reduce() method to define custom serialization for Python’s pickle library).

Working on a Per-Partition Basis

deffetchCallSigns(input):
"""Fetch call signs"""
returninput.mapPartitions(lambdacallSigns:processCallSigns(callSigns))

contactsContactList=fetchCallSigns(validSigns)

Working on a Per-Partition Basis	\|	107

// Use mapPartitions to reuse setup work.

JavaPairRDD<String,CallLog[]>contactsContactLists=
validCallSigns.mapPartitionsToPair(
newPairFlatMapFunction<Iterator<String>,String,CallLog[]>(){
publicIterable<Tuple2<String,CallLog[]>>call(Iterator<String>input){ // List for our results.

			We are called with	We return	Function signature on RDD[T]
			Iterator of the elements in		f: (Iterator[T]) →
that partition				return elements
			Integer of partition number,	Iterator of our
				return elements
that partition				return elements	tor[U]
108	\|	Chapter 6: Advanced Spark Programming