Functional programming with_jdk8-s_ritter

Functional ProgrammingWith JDK8

Simon RitterHead of Java Technology EvangelismOracle Corp.

Twitter: @speakjava

Copyright © 2014, Oracle and/or its affiliates. All rights reserved.


Safe Harbor Statement

The following is intended to outline our general product direction. It is intended for information purposes only, and may not be incorporated into any contract. It is not a commitment to deliver any material, code, or functionality, and should not be relied upon in making purchasing decisions. The development, release, and timing of any features or functionality described for Oracle’s products remains at the sole discretion of Oracle.

2


Lambda Expressions


1996 … 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

1.0 5.0 6 7 8

java.lang.Thread

java.util.concurrent(jsr166)

Fork/Join Framework(jsr166y)

Project LambdaConcurrency in Java

Phasers, etc(jsr166)


The Problem: External Iteration

List<Student> students = ...

double highestScore = 0.0;

for (Student s : students) {

if (s.getGradYear() == 2011) {

if (s.getScore() > highestScore)

highestScore = s.score;

}

}

• Our code controls iteration

• Inherently serial: iterate from beginning to end

• Not thread-safe

• Business logic is stateful

• Mutable accumulator variable


Internal Iteration With Inner Classes

• Iteration handled by the library

• Not inherently serial – traversal maybe done in parallel

• Traversal may be done lazily – so one pass, rather than three

• Thread safe – client logic is stateless

• High barrier to use

– Syntactically ugly

More Functional

double highestScore = students

.filter(new Predicate<Student>() {

public boolean op(Student s) {

return s.getGradYear() == 2011;

}

})

.map(new Mapper<Student,Double>() {

public Double extract(Student s) {

return s.getScore();

}

})

.max();


Internal Iteration With Lambdas

List<Student> students = ...


.filter(Student s -> s.getGradYear() == 2011)

.map(Student s -> s.getScore())

.max();

• More readable

• More abstract

• Less error-prone

NOTE: This is not JDK8 code


Lambda Expressions

• Lambda expressions represent anonymous functions

– Same structure as a method• typed argument list, return type, set of thrown exceptions, and a body

– Not associated with a class

• We now have parameterised behaviour, not just values

Some Details


.filter(Student s -> s.getGradYear() == 2011)

.map(Student s -> s.getScore())

.max();

What

How


Lambda Expression Types

• Single-method interfaces are used extensively in Java

– Definition: a functional interface is an interface with one abstract method

– Functional interfaces are identified structurally

– The type of a lambda expression will be a functional interface• Lambda expressions provide implementations of the abstract method

interface Comparator<T> { boolean compare(T x, T y); }

interface FileFilter { boolean accept(File x); }

interface Runnable { void run(); }

interface ActionListener { void actionPerformed(…); }

interface Callable<T> { T call(); }


Local Variable Capture

• Lambda expressions can refer to effectively final local variables from the surrounding scope

– Effectively final: A variable that meets the requirements for final variables (i.e., assigned once), even if not explicitly declared final

– Closures on values, not variables

void expire(File root, long before) {

root.listFiles(File p -> p.lastModified() <= before);

}


What Does ‘this’ Mean In A Lambda

• ‘this’ refers to the enclosing object, not the lambda itself

• Think of ‘this’ as a final predefined local

• Remember the Lambda is an anonymous function

– It is not associated with a class

– Therefore there can be no ‘this’ for the Lambda


Referencing Instance VariablesWhich are not final, or effectively final

class DataProcessor {

private int currentValue;

public void process() {

DataSet myData = myFactory.getDataSet();

dataSet.forEach(d -> d.use(currentValue++));

}

}


Referencing Instance VariablesThe compiler helps us out

class DataProcessor {

private int currentValue;

public void process() {

DataSet myData = myFactory.getDataSet();

dataSet.forEach(d -> d.use(this.currentValue++);

}

}

‘this’ (which is effectively final) inserted by the compiler


Type Inference

• The compiler can often infer parameter types in a lambda expression

Inferrence based on the target functional interface’s method signature

• Fully statically typed (no dynamic typing sneaking in)

– More typing with less typing

List<String> list = getList();Collections.sort(list, (String x, String y) -> x.length() - y.length());

Collections.sort(list, (x, y) -> x.length() - y.length());

static T void sort(List<T> l, Comparator<? super T> c);


Method References

• Method references let us reuse a method as a lambda expression

FileFilter x = File f -> f.canRead();

FileFilter x = File::canRead;


Method References

• Format: target_reference::method_name

• Three kinds of method reference

– Static method (e.g. Integer::parseInt)

– Instance method of an arbitary type (e.g. String::length)

– Instance method of an existing object

16

More Detail


Method References

17

Rules For Construction

Lambda

Method Ref

Lambda

Method Ref

Lambda

Method Ref

(args) -> ClassName.staticMethod(args)

(arg0, rest) -> arg0.instanceMethod(rest)

(args) -> expr.instanceMethod(args)

ClassName::staticMethod

ClassName::instanceMethod

expr::instanceMethod

instanceOf


Constructor References

• Same concept as a method reference

– For the constructor

Factory<List<String>> f = ArrayList<String>::new;

Factory<List<String>> f = () -> return new ArrayList<String>();

Replace with


Useful Methods That Can Use Lambdas

• Iterable.forEach(Consumer c)

List<String> myList = ...

myList.forEach(s -> System.out.println(s));

myList.forEach(System.out::println);



• Collection.removeIf(Predicate p)


myList.removeIf(s -> s.length() == 0)



• List.replaceAll(UnaryOperator o)


myList.replaceAll(s -> s.toUpperCase());

myList.replaceAll(String::toUpper);



• List.sort(Comparator c)

• Replaces Collections.sort(List l, Comparator c)


myList.sort((x, y) -> x.length() – y.length());


Library Evolution


Library Evolution Goal

• Requirement: aggregate operations on collections–New methods required on Collections to facilitate this

• This is problematic– Can’t add new methods to interfaces without modifying all implementations

– Can’t necessarily find or control all implementations

int heaviestBlueBlock = blocks

.filter(b -> b.getColor() == BLUE)

.map(Block::getWeight)

.reduce(0, Integer::max);


Solution: Default Methods

• Specified in the interface

• From the caller’s perspective, just an ordinary interface method

• Provides a default implementation

• Default only used when implementation classes do not provide a body for the extension method

• Implementation classes can provide a better version, or not

interface Collection<E> {

default Stream<E> stream() {

return StreamSupport.stream(spliterator());

}

}


Virtual Extension Methods

• Err, isn’t this implementing multiple inheritance for Java?

• Yes, but Java already has multiple inheritance of types

• This adds multiple inheritance of behavior too

• But not state, which is where most of the trouble is

• Can still be a source of complexity • Class implements two interfaces, both of which have default methods

• Same signature

• How does the compiler differentiate?

• Static methods also allowed in interfaces in Java SE 8

Stop right there!


Functional Interface Definition

• An interface

• Must have only one abstract method

– In JDK 7 this would mean only one method (like ActionListener)

• JDK 8 introduced default methods

– Adding multiple inheritance of types to Java

– These are, by definition, not abstract (they have an implementation)

• JDK 8 also now allows interfaces to have static methods

– Again, not abstract

• @FunctionalInterface can be used to have the compiler check

27


Is This A Functional Interface?

28

@FunctionalInterfacepublic interface Runnable {

public abstract void run();}

Yes. There is only one abstract method



29

@FunctionalInterfacepublic interface Predicate<T> {

default Predicate<T> and(Predicate<? super T> p) {…};default Predicate<T> negate() {…};default Predicate<T> or(Predicate<? super T> p) {…};static <T> Predicate<T> isEqual(Object target) {…};boolean test(T t);

}

Yes. There is still only one abstract method



30

@FunctionalInterfacepublic interface Comparator {

// default and static methods elidedint compare(T o1, T o2);boolean equals(Object obj);

}

The equals(Object) method is implicit from the Object class

Therefore only oneabstract method


Lambdas In Full Flow:Streams


Stream Overview

• Abstraction for specifying aggregate computations

– Not a data structure

– Can be infinite

• Simplifies the description of aggregate computations– Exposes opportunities for optimisation

– Fusing, laziness and parallelism

At The High Level


Stream Overview

• A stream pipeline consists of three types of things

– A source

– Zero or more intermediate operations

– A terminal operation• Producing a result or a side-effect

Pipeline

int total = transactions.stream().filter(t -> t.getBuyer().getCity().equals(“London”)).mapToInt(Transaction::getPrice).sum();

Source

Intermediate operation

Terminal operation


Stream Sources

• From collections and arrays

– Collection.stream()

– Collection.parallelStream()

– Arrays.stream(T array) or Stream.of()

• Static factories

– IntStream.range()

– Files.walk()

• Roll your own

– java.util.Spliterator

Many Ways To Create


Stream Terminal Operations

• The pipeline is only evaluated when the terminal operation is called

– All operations can execute sequentially or in parallel

– Intermediate operations can be merged• Avoiding multiple redundant passes on data

• Short-circuit operations (e.g. findFirst)

• Lazy evaluation

– Stream characteristics help identify optimisations• DISTINT stream passed to distinct() is a no-op


Maps and FlatMapsMap Values in a Stream

Map

FlatMap

Input Stream

Input Stream

1-to-1 mapping

1-to-many mapping

Output Stream

Output Stream


The Optional Class


Optional<T>Reducing NullPointerException Occurrences

String direction = gpsData.getPosition().getLatitude().getDirection();

String direction = “UNKNOWN”;

if (gpsData != null) {Position p = gpsData.getPosition();

if (p != null) {Latitude latitude = p.getLatitude();

if (latitude != null)direction = latitude.getDirection();

}}

String direction = gpsData.getPosition().getLatitude().getDirection();


Optional Class

• Terminal operations like min(), max(), etc do not return a direct result

• Suppose the input Stream is empty?

• Optional<T>– Container for an object reference (null, or real object)

– Think of it like a Stream of 0 or 1 elements

– use get(), ifPresent() and orElse() to access the stored reference

– Can use in more complex ways: filter(), map(), etc

– gpsMaybe.filter(r -> r.lastReading() < 2).ifPresent(GPSData::display);

Helping To Eliminate the NullPointerException


Optional ifPresent()Do something when set

if (x != null) {print(x);

}

opt.ifPresent(x -> print(x));opt.ifPresent(this::print);


Optional filter()Reject certain values of the Optional

if (x != null && x.contains("a")) {print(x);

}

opt.filter(x -> x.contains("a")).ifPresent(this::print);


Optional map()Transform value if present

if (x != null) {String t = x.trim();if (t.length() > 1)

print(t);}

opt.map(String::trim).filter(t -> t.length() > 1).ifPresent(this::print);


Optional flatMap()Going deeper

public String findSimilar(String s)

Optional<String> tryFindSimilar(String s)

Optional<Optional<String>> bad = opt.map(this::tryFindSimilar);Optional<String> similar = opt.flatMap(this::tryFindSimilar);


Update Our GPS Code

class GPSData {public Optional<Position> getPosition() { ... }

}

class Position {public Optional<Latitude> getLatitude() { ... }

}

class Latitude {public String getString() { ... }

}


Update Our GPS Code

String direction = Optional.ofNullable(gpsData).flatMap(GPSData::getPosition).flatMap(Position::getLatitude).map(Latitude::getDirection).orElse(“None”);


Stream Examples


Example 1Convert words in list to upper case

List<String> output = wordList.stream().map(String::toUpperCase).collect(Collectors.toList());


Example 1Convert words in list to upper case (in parallel)

List<String> output = wordList.parallelStream().map(String::toUpperCase).collect(Collectors.toList());


Example 2

• BufferedReader has new method

– Stream<String> lines()

Count lines in a file

long count = bufferedReader.lines().count();


Example 3Join lines 3-4 into a single string

String output = bufferedReader.lines().skip(2).limit(2).collect(Collectors.joining());


Example 4Collect all words in a file into a list

List<String> output = reader.lines().flatMap(line -> Stream.of(line.split(REGEXP))).filter(word -> word.length() > 0).collect(Collectors.toList());


Example 5List of unique words in lowercase, sorted by length

List<String> output = reader.lines().flatMap(line -> Stream.of(line.split(REGEXP))).filter(word -> word.length() > 0).map(String::toLowerCase).distinct().sorted((x, y) -> x.length() - y.length()).collect(Collectors.toList());


Example 6Create a map: Keys are word length, values are lists of words of that length

Map<Integer, List<String>> result = reader.lines().flatMap(line -> Stream.of(line.split(REGEXP))).collect(groupingBy(String::length));


Example 7Create a map: Keys are word length, values are how many words of that length

Map<Integer, Long> result = reader.lines().flatMap(line -> Stream.of(line.split(REGEXP))).collect(groupingBy(String::length, counting()));

Downstream collectorto return count of elements


Example 8: Real WorldInfinite stream from thermal sensor

private int double currentTemperature;...

thermalReader.lines().mapToDouble(s ->

Double.parseDouble(s.substring(0, s.length() - 1))).map(t -> ((t – 32) * 5 / 9).filter(t -> t != currentTemperature).peek(t -> listener.ifPresent(l -> l.temperatureChanged(t))).forEach(t -> currentTemperature = t);


Example 8: Real WorldInfinite stream from thermal sensor

private int double currentTemperature;...

thermalReader.lines().mapToDouble(s ->

Double.parseDouble(s.substring(0, s.length() - ))).map(t -> ((t – 32) * 5 / 9).filter(t -> t != this.currentTemperature).peek(t -> listener.ifPresent(l -> l.temperatureChanged(t))).forEach(t -> this.currentTemperature = t);


Advanced Lambdas And Streams


Exception Handling In Lambdas


Exception Handling In Lambda Expressions

• Compiler error: uncaught exception in filter()

Can Be Problematic

public boolean isValid(Object o) throws IOException { ... }

public List<String> processList(List<String> list) throws IOException {

return list.stream().filter(w -> isValid(w)).collect(Collectors.toList());

}



60

Catch the Exception In The Lambda

public double processList(List list) {return list.stream().filter(w -> {

try {return isValid(w)

} catch (IOException ioe) {return false;

}}).collect(Collectors.toList());

}



• Which is great, except you can't use it with Streams

61

Define A New Functional Interface

@FunctionalInterfacepublic interface PredicateWithException {

public boolean test(String s) throws IOException;}


Debugging


Logging Mid-StreamThe peek() method

List<String> output = reader.lines().flatMap(line -> Stream.of(line.split(REGEXP))).filter(word -> word.length() > 5).peek(s -> System.out.println(“Word = “ + s)).collect(Collectors.toList());


Using peek() To Set A Breakpoint

List<String> output = reader.lines().flatMap(line -> Stream.of(line.split(REGEXP))).filter(word -> word.length() > 5).peek(s -> s).collect(Collectors.toList());

Set breakpoint on peekSome debuggers don’t like empty bodies


Streams and Concurrency


Serial And Parallel Streams

• Collection Stream sources

– stream()

– parallelStream()

• Stream can be made parallel or sequential at any point

– parallel()

– sequential()

• The last call wins

– Whole stream is either sequential or parallel

66


Parallel Streams

• Implemented underneath using the fork-join framework

• Will default to as many threads for the pool as the OS reports processors

– Which may not be what you want

System.setProperty("java.util.concurrent.ForkJoinPool.common.parallelism", "32767");

• Remember, parallel streams always need more work to process

– But they might finish it more quickly

67


When To Use Parallel Streams

• Data set size is important, as is the type of data structure

– ArrayList: GOOD

– HashSet, TreeSet: OK

– LinkedList: BAD

• Operations are also important

– Certain operations decompose to parallel tasks better than others

– filter() and map() are excellent

– sorted() and distinct() do not decompose well

68

Sadly, no simple answer


When To Use Parallel Streams

• N = size of the data set

• Q = Cost per element through the Stream pipeline

• N x Q = Total cost of pipeline operations

• The bigger N x Q is the better a parallel stream will perform

• It is easier to know N than Q, but Q can be estimated

• If in doubt, profile

69

Quantative Considerations


Streams: Pitfalls For The Unwary


Functional v. Imperative

• For functional programming you should not modify state

• Java supports closures over values, not closures over variables

• But state is really useful…

71


Counting Methods That Return Streams

72

Still Thinking Imperatively

Set<String> sourceKeySet = streamReturningMethodMap.keySet();

LongAdder sourceCount = new LongAdder();

sourceKeySet.stream().forEach(c ->

sourceCount.add(streamReturningMethodMap.get(c).size()));


Counting Methods That Return Streams

73

Functional Way

sourceKeySet.stream().mapToInt(c -> streamReturningMethodMap.get(c).size()).sum();


Printing And Counting Functional Interfaces

74

Still Thinking Imperatively

LongAdder newMethodCount = new LongAdder();

functionalParameterMethodMap.get(c).stream().forEach(m -> {

output.println(m);

if (isNewMethod(c, m)) newMethodCount.increment();

});

return newMethodCount.intValue();



75

More Functional, But Not Pure Functional

int count = functionalParameterMethodMap.get(c).stream().mapToInt(m -> {

int newMethod = 0;output.println(m);

if (isNewMethod(c, m)) newMethod = 1;

return newMethod}).sum();

There is still state being modified in the Lambda



76

Even More Functional, But Still Not Pure Functional

int count = functionalParameterMethodMap.get(nameOfClass).stream().peek(method -> output.println(method)).mapToInt(m -> isNewMethod(nameOfClass, m) ? 1 : 0) .sum();

Strictly speaking printing is a side effect, which is not purely functional


The Art Of Reduction(Or The Need to Think Differently)


A Simple Problem

• Find the length of the longest line in a file

• Hint: BufferedReader has a new method, lines(), that returns a Stream

78

BufferedReader reader = ...

reader.lines().mapToInt(String::length).max().getAsInt();


Another Simple Problem

• Find the length of the longest line in a file

79


Naïve Stream Solution

• That works, so job done, right?

• Not really. Big files will take a long time and a lot of resources

• Must be a better approach

80

String longest = reader.lines().sort((x, y) -> y.length() - x.length()).findFirst().get();


External Iteration Solution

• Simple, but inherently serial

• Not thread safe due to mutable state

81

String longest = "";

while ((String s = reader.readLine()) != null)if (s.length() > longest.length())

longest = s;


Recursive Approach: The Method

82

String findLongestString(String s, int index, List<String> l) {if (index >= l.size())return s;

if (index == l.size() - 1) {if (s.length() > l.get(index).length())return s;

return l.get(index);}

String s2 = findLongestString(l.get(start), index + 1, l);

if (s.length() > s2.length())return s;

return s2;}


Recursive Approach: Solving The Problem

• No explicit loop, no mutable state, we’re all good now, right?

• Unfortunately not - larger data sets will generate an OOM exception

83

List<String> lines = new ArrayList<>();

while ((String s = reader.readLine()) != null)lines.add(s);

String longest = findLongestString("", 0, lines);


A Better Stream Solution

• The Stream API uses the well known filter-map-reduce pattern

• For this problem we do not need to filter or map, just reduce

Optional<T> reduce(BinaryOperator<T> accumulator)

• The key is to find the right accumulator

– The accumulator takes a partial result and the next element, and returns a new partial result

– In essence it does the same as our recursive solution

– Without all the stack frames

• BinaryOperator is a subclass of BiFunction, but all types are the same

• R apply(T t, U u) or T apply(T x, T y)

84



• Use the recursive approach as an accululator for a reduction

85

String longestLine = reader.lines().reduce((x, y) -> {

if (x.length() > y.length())return x;

return y;}).get();



• Use the recursive approach as an accululator for a reduction

86

String longestLine = reader.lines().reduce((x, y) -> {

if (x.length() > y.length())return x;

return y;}).get();

x in effect maintains state for us, by always holding the longest string found so far


The Simplest Stream Solution

• Use a specialised form of max()

• One that takes a Comparator as a parameter

• comparingInt() is a static method on Comparator– Comparator<T> comparingInt(ToIntFunction<? extends T> keyExtractor)

87

reader.lines().max(comparingInt(String::length)).get();


Conclusions

• Java needs lambda statements

– Significant improvements in existing libraries are required

• Require a mechanism for interface evolution

– Solution: virtual extension methods

• Bulk operations on Collections– Much simpler with Lambdas

• Java SE 8 evolves the language, libraries, and VM together

Simon RitterOracle Corporartion

Twitter: @speakjava


Software

Functional programming with_jdk8-s_ritter