Java 8, Streams & Collectors, patterns, performances and parallelization

Java 8 Streams & Collectors
Patterns, performance, parallelization
@JosePaumard

@JosePaumard
Stream Collectors

@JosePaumard
#J8Stream
Code & slides
1st part: slides
-Stream
-Operations
-State
-Reduction
-Collectors
2nd part: live coding!
-Movies & actors
-Shakespeare

Questions ?
#J8Stream
@JosePaumard

@JosePaumard
#J8Stream
What is a stream?

@JosePaumard
#J8Stream
What is a stream?
Answer 1: a typed interface
public interface Stream<T> extends BaseStream<T, Stream<T>> {
// ...
}

@JosePaumard
#J8Stream
What is a stream?
Answer 1: a typed interface
In fact: a new concept in the JDK
public interface Stream<T> extends BaseStream<T, Stream<T>> {
// ...
}

@JosePaumard
#J8Stream
What is a stream?
What can I do with it?

@JosePaumard
#J8Stream
What is a stream?
What can I do with it?
Answer: efficiently process high volumes of data, but also small ones

@JosePaumard
#J8Stream
What is a stream?
What does efficiently mean?

@JosePaumard
#J8Stream
What is a stream?
Two things:

@JosePaumard
#J8Stream
What is a stream?
Two things:
1)in parallel, to leverage multicore CPU

@JosePaumard
#J8Stream
What is a stream?
Two things:
1)in parallel, to leverage multicore CPU
2)in pipeline, to avoid the computation of intermediate data structures

@JosePaumard
#J8Stream
What is a stream?
Why cant a collection be a stream?

@JosePaumard
#J8Stream
What is a stream?
In fact there are arguments for a Collection to be a Stream!

@JosePaumard
#J8Stream
What is a stream?
1)this is where my data is!

@JosePaumard
#J8Stream
What is a stream?
1)this is where my data is!
2)and I know the Collection API pretty well…

@JosePaumard
#J8Stream
What is a stream?
Two reasons:
1)we want to be free to create new concepts

@JosePaumard
#J8Stream
What is a stream?
Two reasons:
1)we want to be free to create new concepts
2)we dont want to add those concepts on well-known interfaces

@JosePaumard
#J8Stream
So: what is a Stream?
Answers:

@JosePaumard
#J8Stream
Answers:
1)an object on which I can define operations

@JosePaumard
#J8Stream
Answers:
2)that does not hold any data

@JosePaumard
#J8Stream
Answers:
3)that will not modify the data it processes

@JosePaumard
#J8Stream
Answers:
4)that will process data in « one pass »

@JosePaumard
#J8Stream
Answers:
4)that will process data in « one pass »
5)that is built on higly optimized algorithms, and that can work in parallel

@JosePaumard
#J8Stream
How can I build a Stream?
Many patterns!

@JosePaumard
#J8Stream
How can I build a Stream?
Many patterns!
Let us take one:
List<Person> persons = ... ;
Stream<Person> stream = persons.stream() ;

@JosePaumard
#J8Stream
Operations on a Stream
Let us see a first operation: forEach()
… that will display each person
stream.forEach(p -> System.out.println(p)) ;

@JosePaumard
#J8Stream
stream.forEach(System.out::println) ;

@JosePaumard
#J8Stream
stream.forEach(System.out::println) ;
o -> System.out.println(o) ≡ System.out::println

@JosePaumard
#J8Stream
Operation: forEach()
First operation: forEach()
forEach(): takes a Consumer<T> as a parameter
@FunctionalInterface public interface Consumer<T> { void accept(T t) ; }

@JosePaumard
#J8Stream
Operation: forEach()
First operation: forEach()
forEach(): takes a Consumer<T> as a parameter
What is a functional interface?
@FunctionalInterface
public interface Consumer<T> {
void accept(T t) ;
}

@JosePaumard
#J8Stream
Functional interface
Definition:
an interface with only one abstract method

@JosePaumard
#J8Stream
Definition:
(methods from Object do not count)

@JosePaumard
#J8Stream
Definition:
(default & static methods do not count)

@JosePaumard
#J8Stream
Definition:
It « can » be annotated by @FunctionalInterface

@JosePaumard
#J8Stream
Definition:
It « can » be annotated by @FunctionalInterface
If present: the Java 8 compiler will help me

@JosePaumard
#J8Stream
In fact…
Consumer is a bit more complex than that:
void accept(T t) ;
default Consumer<T> andThen(Consumer<? super T> after) {
Objects.requireNonNull(after);
return (T t) -> { accept(t); after.accept(t); };
}
}

@JosePaumard
#J8Stream
In fact…
What is a default method?
void accept(T t) ;
}
}

@JosePaumard
#J8Stream
Default methods
New concept in Java 8!

@JosePaumard
#J8Stream
Default methods
Added to allow the adding of methods in interfaces without breaking the old implementations

@JosePaumard
#J8Stream
Default methods
Added to allow the adding of methods in interfaces without breaking the old implementations
Ex: stream() on the Collection interface

@JosePaumard
#J8Stream
Collection.stream()
public interface Collection<E> {
default Stream<E> stream() {
return StreamSupport.stream(spliterator(), false);
}
}

@JosePaumard
#J8Stream
Collection.stream()
public interface Collection<E> {
default Stream<E> stream() {
return StreamSupport.stream(spliterator(), false);
}
default Spliterator<E> spliterator() {
return Spliterators.spliterator(this, 0);
}
}

@JosePaumard
#J8Stream
Back to the Consumer interface
void accept(T t) ;
}
}

@JosePaumard
#J8Stream
List<String> list = new ArrayList<>() ;
Consumer<String> c1 = s -> list.add(s) ;
Consumer<String> c2 = s -> System.out.println(s) ;

@JosePaumard
#J8Stream
Consumer<String> c1 = list::add ;
Consumer<String> c2 = System.out::println ;

@JosePaumard
#J8Stream
Consumer<String> c2 = System.out::println ;
Consumer<String> c3 = c1.andThen(c2) ; // and we could chain more

@JosePaumard
#J8Stream
Beware concurrency!
List<String> result = new ArrayList<>() ;
persons.stream()
.forEach(c1) ; // concurrency?

@JosePaumard
#J8Stream
Beware concurrency!
persons.stream().parallel()
.forEach(c1) ; // concurrency?

@JosePaumard
#J8Stream
Beware concurrency!
persons.stream().parallel()
.forEach(c1) ; // concurrency? Baaaad pattern!

@JosePaumard
#J8Stream
Problem: forEach() returns void
Consumer<String> c2 = System.out::println
persons.stream()
.forEach(c1.andThen(c2)) ;

@JosePaumard
#J8Stream
But I also have a peek() method
That returns another Stream
persons.stream()
.peek(System.out::println)
.filter(person -> person.getAge() > 20)
.peek(result::add) ; // Baaad pattern !

@JosePaumard
#J8Stream
Back to Stream
So far we saw:
-forEach(Consumer)
-peek(Consumer)

@JosePaumard
#J8Stream
Back to Stream
3rd method: filter(Predicate)

@JosePaumard
#J8Stream
Method filter()
List<Person> list = ... ;
Stream<Person> stream = list.stream() ;
Stream<Person> filtered =
stream.filter(person -> person.getAge() > 20) ;

@JosePaumard
#J8Stream
Method filter()
Predicate<Person> p = person -> person.getAge() > 20 ;

@JosePaumard
#J8Stream
Interface Predicate
Another functional interface
@FunctionalInterface public interface Predicate<T> { boolean test(T t) ; }

@JosePaumard
#J8Stream
Interface Predicate
In fact Predicate looks like this:
public interface Predicate<T> {
boolean test(T t) ;
default Predicate<T> and(Predicate<? super T> other) { ... }
default Predicate<T> or(Predicate<? super T> other) { ... }
default Predicate<T> negate() { ... }
}

@JosePaumard
#J8Stream
Interface Predicate
default Predicate<T> and(Predicate<? super T> other) {
return t -> test(t) && other.test(t) ;
}
default Predicate<T> or(Predicate<? super T> other) { return t -> test(t) || other.test(t) ; }
default Predicate<T> negate() {
return t -> !test(t) ;
}

@JosePaumard
#J8Stream
Interface Predicate: patterns
Predicate<Integer> p1 = i -> i > 20 ;
Predicate<Integer> p2 = i -> i < 30 ;
Predicate<Integer> p3 = i -> i == 0 ;

@JosePaumard
#J8Stream
Predicate<Integer> p = p1.and(p2).or(p3) ; // (p1 AND p2) OR p3

@JosePaumard
#J8Stream
Predicate<Integer> p = p1.and(p2).or(p3) ; // (p1 AND p2) OR p3
Predicate<Integer> p = p3.or(p1).and(p2) ; // (p3 OR p1) AND p2

@JosePaumard
#J8Stream
Interface Predicate
In fact Predicate looks like this (bis):
public interface Predicate<T> {
boolean test(T t) ;
// default methods
static <T> Predicate<T> isEqual(Object o) { ... }
}

@JosePaumard
#J8Stream
Interface Predicate
static <T> Predicate<T> isEqual(Object o) {
return t -> o.equals(t) ;
}

@JosePaumard
#J8Stream
Interface Predicate
In fact things are bit more complicated:
}
return (null == o)
? obj -> Objects.isNull(obj)
: t -> o.equals(t) ;
}

@JosePaumard
#J8Stream
Interface Predicate
In fact things are bit more complicated:
}
return (null == o)
? Objects::isNull
: o::equals ;
}

@JosePaumard
#J8Stream
Interface Predicate
We also have:
public interface BiPredicate<T, U> {
boolean test(T t, U u);
}

@JosePaumard
#J8Stream
Filtered Streams
In this code:
stream and filtered are different objects
The filter() method returns a new object

@JosePaumard
#J8Stream
Filtered Streams
In this code:
Question 1: what do we have in the filtered object?

@JosePaumard
#J8Stream
Filtered Streams
In this code:
Answer: the filtered data!

@JosePaumard
#J8Stream
Filtered Streams
In this code:
Answer: the filtered data… really?

@JosePaumard
#J8Stream
Filtered Streams
In this code:
Didnt we say: « a stream does not hold any data »?

@JosePaumard
#J8Stream
Filtered Streams
In this code:
Correct answer: nothing!

@JosePaumard
#J8Stream
Filtered Streams
In this code:
Question 2: what does this code do?

@JosePaumard
#J8Stream
Filtered Streams
In this code:
Answer: nothing…

@JosePaumard
#J8Stream
Filtered Streams
In this code:
Answer: nothing…
This call is a declaration, no data is processed!

@JosePaumard
#J8Stream
Filtered Streams
The filter() call is a lazy call
Generally speaking:
A call to a method that returns a Stream is a lazy call

@JosePaumard
#J8Stream
Filtered Streams
The filter() call is a lazy call
Another way of saying it:
An operation that returns a Stream is an intermediate operation

@JosePaumard
#J8Stream
Intermediate call
Stating that a Stream does not hold any data
creates the notion of intermediate operation
adds the notion of lazyness to the Stream API

@JosePaumard
#J8Stream
Intermediate call
Stating that a Stream does not hold any data
creates the notion of intermediate operation
adds the notion of lazyness to the Stream API
There must be some kind terminal operation somewhere…

@JosePaumard
#J8Stream
Back to Consumer (bis)
What happens in that code?
persons.stream()
.peek(resultat::add) ; // Baaad pattern !

@JosePaumard
#J8Stream
Answer is: nothing!
persons.stream()

@JosePaumard
#J8Stream
1) nothing displayed
2) result is empty
persons.stream()

@JosePaumard
#J8Stream
What do we have so far?
We saw:
-forEach(Consumer)
-peek(Consumer)
-filter(Predicate)

@JosePaumard
#J8Stream
We saw:
-forEach(Consumer)
-peek(Consumer)
-filter(Predicate)
The notion of lazyness / intermediate call

@JosePaumard
#J8Stream
Let us carry on
And see the mapping operation

@JosePaumard
#J8Stream
Mapping
Mapping operation
Stream<String> names =
stream.map(person -> person.getName()) ;

@JosePaumard
#J8Stream
Mapping
Mapping operation
The map() method returns a Stream
It is thus lazy / intermediate
Stream<String> names =
stream.map(person -> person.getName()) ;

@JosePaumard
#J8Stream
Interface Function
The map() method takes a Function:
public interface Function<T, R> {
R apply(T t) ;
}

@JosePaumard
#J8Stream
Interface Function
Which in fact is:
R apply(T t) ;
default <V> Function<V, R> compose(Function<V, T> before) ;
default <V> Function<T, V> andThen(Function<R, V> after) ;
}

@JosePaumard
#J8Stream
Interface Function
Which in fact is:
Beware the generics!
R apply(T t) ;
default <V> Function<V, R> compose(Function<V, T> before) ;
default <V> Function<T, V> andThen(Function<R, V> after) ;
}

@JosePaumard
#J8Stream
Interface Function
Which in fact is:
R apply(T t) ;
default <V> Function<V, R> compose(
Function<? super V, ? extends T> before) ;
default <V> Function<T, V> andThen(
Function<? super R, ? extends V> after) ;
}

@JosePaumard
#J8Stream
Interface Function
Which in fact is:
R apply(T t) ;
// default methods
static <T> Function<T, T> identity() {
return t -> t ;
}
}

@JosePaumard
#J8Stream
Other functions
There are special types of functions:
public interface UnaryOperator<T> extends Function<T, T> {
}

@JosePaumard
#J8Stream
Other functions
public interface BiFunction<T, U, R> {
R apply(T t, U u);
}

@JosePaumard
#J8Stream
Other functions
public interface BinaryOperator<T> extends BiFunction<T, T, T> {
}

@JosePaumard
#J8Stream
We saw:
-forEach(Consumer)
-peek(Consumer)
-filter(Predicate)
-map(Function)

@JosePaumard
#J8Stream
Let us carry on
Method flatMap()

@JosePaumard
#J8Stream
Method flatMap()
flatMap() = flattens a Stream
The signature is:
The passed function returns a Stream
<R> Stream<R> flatMap(Function<T, Stream<R>> mapper) ;

@JosePaumard
#J8Stream
Method flatMap()
Let us take an example:
List<Movie> movies = ... ; // a movie has a set of Actor
movies.stream()
.map(movie -> movie.actors()) // Stream<Set<Actors>>

@JosePaumard
#J8Stream
Method flatMap()
movies.stream()
.map(movie -> movie.actors().stream()) // Stream<Stream<Actors>>

@JosePaumard
#J8Stream
Method flatMap()
movies.stream()
.map(movie -> movie.actors().stream()) // Stream<Stream<Actors>>
Function<Movie, Stream<Actors>> mapper =
movie -> movie.actors().stream();

@JosePaumard
#J8Stream
Method flatMap()
List<Movie> movies = ... ; // a movie has a set of Actor movies.stream() .flatMap(movie -> movie.actors().stream()) // Stream<Actors>

@JosePaumard
#J8Stream
Method flatMap()
Signature:

@JosePaumard
#J8Stream
Method flatMap()
Signature:
Since it returns a Stream, it is a lazy / intermediate call

@JosePaumard
#J8Stream
We have 3 types of methods:
-forEach(): consumes
-peek(): consumes, and transmits
-filter(): filters
-map(): transforms
-flatMap(): transforms and flattens

@JosePaumard
#J8Stream
And what about reductions?

@JosePaumard
#J8Stream
Reduction
Reduction method:
List<Integer> ages = ... ;
Stream<Integer> stream = ages.stream() ;
Integer sum =
stream.reduce(0, (age1, age2) -> age1 + age2) ;

@JosePaumard
#J8Stream
Reduction
Reduction method:
Two things to stress out here
Integer sum =

@JosePaumard
#J8Stream
Reduction
Reduction method:
Two things to stress out here
Integer sum =
stream.reduce(0, Intege::sum) ;

@JosePaumard
#J8Stream
Reduction
Reduction method:
0 is the « default value » of the reduction operation
returned if the stream is « empty »
Integer sum =

@JosePaumard
#J8Stream
Reduction
Reduction method:
This « default value » has to be the identity element of the reduction operation
Integer sum =

@JosePaumard
#J8Stream
Reduction
Reduction method:
If it is not, then the computation will fail!
Integer sum =

@JosePaumard
#J8Stream
Reduction
Why?
-This 0 is returned in case of an empty Stream
-Red(0, p) is returned in case of a singleton Stream
In fact, 0 is used to compute the reduction of the Stream

@JosePaumard
#J8Stream
Reduction
What is going to happen if the reduction operation has no identity value?

@JosePaumard
#J8Stream
Reduction
What is going to happen if the reduction operation has no identity value?
But of course we do not use such reductions…

@JosePaumard
#J8Stream
Reduction
Reduction method:
The reduction operation should be associative
Integer sum =

@JosePaumard
#J8Stream
Reduction
Reduction method:
The reduction operation should be associative
If not, we will have trouble in parallel
Integer sum =

@JosePaumard
#J8Stream
Reduction
Associative, or not associative?
red1 = (i1, i2) -> i1 + i2 ;

@JosePaumard
#J8Stream
Reduction
red1 = (i1, i2) -> i1 + i2 ; 
red2 = (i1, i2) -> i1 > i2 ? i1 : i2 ;

@JosePaumard
#J8Stream
Reduction
red1 = (i1, i2) -> i1 + i2 ; 
red2 = (i1, i2) -> i1 > i2 ? i1 : i2 ; 
red3 = (i1, i2) -> i1*i2 ;

@JosePaumard
#J8Stream
Reduction
red1 = (i1, i2) -> i1 + i2 ; 
red2 = (i1, i2) -> i1 > i2 ? i1 : i2 ; 
red3 = (i1, i2) -> i1*i2 ; 
red4 = (i1, i2) -> i1*i1 + i2*i2 ;

@JosePaumard
#J8Stream
Reduction
red1 = (i1, i2) -> i1 + i2 ; 
red2 = (i1, i2) -> i1 > i2 ? i1 : i2 ; 
red3 = (i1, i2) -> i1*i2 ; 
red4 = (i1, i2) -> i1*i1 + i2*i2 ; 
red5 = (i1, i2) -> (i1 + i2) / 2 ;

@JosePaumard
#J8Stream
Reduction
red1 = (i1, i2) -> i1 + i2 ; 
red2 = (i1, i2) -> i1 > i2 ? i1 : i2 ; 
red3 = (i1, i2) -> i1*i2 ; 
red4 = (i1, i2) -> i1*i1 + i2*i2 ; 
red5 = (i1, i2) -> (i1 + i2) / 2 ; 
red6 = (i1, i2) -> i2 ;

@JosePaumard
#J8Stream
Reduction
red1 = (i1, i2) -> i1 + i2 ; 
red2 = (i1, i2) -> i1 > i2 ? i1 : i2 ; 
red3 = (i1, i2) -> i1*i2 ; 
red4 = (i1, i2) -> i1*i1 + i2*i2 ; 
red5 = (i1, i2) -> (i1 + i2) / 2 ; 
red6 = (i1, i2) -> i2 ; 

@JosePaumard
#J8Stream
Example of max
Let us reduce with a max
Integer max =
stream.reduce(0, (age1, age2) -> age1 > age2 ? age1 : age2) ;

@JosePaumard
#J8Stream
Example of max
Integer max =
stream.reduce(0, Integer::max) ;

@JosePaumard
#J8Stream
Example of max
Integer max =
List<Integer> ages = new ArrayList<>() ;
Stream<Integer> stream = ages.stream() ; // empty stream
> max = 0

@JosePaumard
#J8Stream
Example of max
Integer max =
List<Integer> ages = Arrays.asList(2) ;
Stream<Integer> stream = ages.stream() ; // singleton stream
> max = 2

@JosePaumard
#J8Stream
Example of max
Integer max =
List<Integer> ages = Arrays.asList(-1) ;
Stream<Integer> stream = ages.stream() ; // singleton stream
> max = 0

@JosePaumard
#J8Stream
Example of max
Integer max =
List<Integer> ages = Arrays.asList(-1, -2) ;
> max = 0

@JosePaumard
#J8Stream
Example of max
The reduction with a max has a problem : why?

@JosePaumard
#J8Stream
Example of max
Answer: the max reduction has no identity element, thus no default value

@JosePaumard
#J8Stream
Example of max
Answer: the max reduction has no identity element, thus no default value
Solution: introduce the notion of Optional

@JosePaumard
#J8Stream
Method max()
Optional is the return type of the max() method
Optional<Integer> max =
stream.max(Comparator.naturalOrder()) ;

@JosePaumard
#J8Stream
Optional
An Optional is a wrapping type, that can be empty
How can I use it?
Optional<String> opt = ... ;
if (opt.isPresent()) {
String s = opt.get() ;
} else {
...
}

@JosePaumard
#J8Stream
Optional
How can I use it?
String s = opt.orElse("") ; // this is a default value that
// is valid for our application
Optional<String> opt = ... ;
if (opt.isPresent()) {
String s = opt.get() ;
} else {
...
}

@JosePaumard
#J8Stream
Optional
How can I use it?
String s = opt.orElseThrow(MyException::new) ; // lazy initialization

@JosePaumard
#J8Stream
Optional can be built
One can explicitly build optionals
Optional<String> opt = Optional.<String>empty() ;
Optional<String> opt = Optional.of("one") ; // not null
Optional<String> opt = Optional.ofNullable(s) ; // may be null

@JosePaumard
#J8Stream
Optional: more patterns
An Optional can be seen as a special Stream with zero or one element
void ifPresent(Consumer<T> consumer) ;

@JosePaumard
#J8Stream
Optional<T> filter(Predicate<T> mapper) ;

@JosePaumard
#J8Stream
Optional map(Function<T, U> mapper) ;

@JosePaumard
#J8Stream
Optional map(Function<T, U> mapper) ;
Optional flatMap(Function<T, Optional> mapper) ;

@JosePaumard
#J8Stream
Optional: new patterns sighted!
public class NewMath {
public static Optional<Double> inv(Double d) {
return d == 0.0d ? Optional.empty() :
Optional.of(1/d) ;
}
public static Optional<Double> sqrt(Double d) {
return d < 0.0d ? Optional.empty() :
Optional.of(Math.sqrt(d)) ;
}
}

@JosePaumard
#J8Stream
List<Double> doubles = Arrays.asList(-1d, 0d, 1d) ;
List<Double> result = new ArrayList<>() ;
doubles.forEach(
d1 -> NewMath.inv(d1) // Optional<Double>
.flatMap(d2 -> NewMath.sqrt(d2)) // Optional<Double>
.ifPresent(result::add)
) ;
doubles : [-1.0, 0.0, 1.0]
result : [1.0]

@JosePaumard
#J8Stream
Function<Double, Optional<Double>> f =
d -> NewMath.inv(d) // Optional<Double>
.flatMap(d -> NewMath.sqrt(d))// Optional<Double>

@JosePaumard
#J8Stream
Function<Double, Optional<Double>> f =
.flatMap(NewMath::sqrt); // Optional<Double>

@JosePaumard
#J8Stream
.flatMap(NewMath::sqrt) // Optional<Double>
.map(Stream::of) // Optional<Stream<Double>>

@JosePaumard
#J8Stream
.orElse(Stream.empty()) // Stream<Double>

@JosePaumard
#J8Stream
Function<Double, Stream<Double>> f =
.orElse(Stream.empty()) ; // Stream<Double>

@JosePaumard
#J8Stream
doubles.stream()
.flatMap(
) // Stream<Double>

@JosePaumard
#J8Stream
doubles.stream()
.flatMap(
) // Stream<Double>
.collect(Collectors.toList()) ;

@JosePaumard
#J8Stream
Back on the reduce() method
Two return types:
Integer sum =
Optional<Integer> opt =
stream.reduce((age1, age2) -> age1 + age2) ;

@JosePaumard
#J8Stream
Back on the reduce() method
Two return types:
Integer sum =
stream.reduce(0, Integer::sum) ;
Optional<Integer> opt =
stream.reduce(Integer::sum) ;

@JosePaumard
#J8Stream
More on the reduction
A reduction never returns a Stream
-max(), min() [optionals]
-count()
Boolean reductions:
-allMatch(), noneMatch, anyMatch()
Return optionals:
-findFirst(), findAny() (parallel!)

@JosePaumard
#J8Stream
Note on the reduction
A reduction never returns a Stream
So it is not a lazy step / intermediate operation
A reduction always triggers the computation
They are terminal operations

@JosePaumard
#J8Stream
Terminal operation
Example
... =
persons.map(person -> person.getAge()) // Stream<Integer>
.filter(age -> age > 20) // Stream<Integer>
.min(Comparator.naturalOrder()) ;

@JosePaumard
#J8Stream
Terminal operation
Example
... =
persons.map(Person::getAge) // Stream<Integer>
.min(Comparator.naturalOrder()) ;

@JosePaumard
#J8Stream
Terminal operation
Let us write a full map / filter / reduce
Optional<Integer> age =
persons.map(person -> person.getAge()) // Stream<Integer>
.min(Comparator.naturalOrder()) ; // terminal operation

@JosePaumard
#J8Stream
Terminal operation
boolean b =
persons.map(person -> person.getLastName())
.allMatch(length < 20) ; // terminal operation

@JosePaumard
#J8Stream
Terminal operation
boolean b =
persons.map(Person::getAge)

@JosePaumard
#J8Stream
Terminal operation
Much more efficient to compute this in one pass over the data: « short cutting methods »
boolean b =
persons.map(Person::getAge)

@JosePaumard
#J8Stream
What is a Stream?
An object on which we can define operations on data
No a priori limit on the size of the data
Typical operations: map / filter / reduce

@JosePaumard
#J8Stream
What is a Stream?
Pipeline approach:
1)we define the operations
2)we trigger the computations by calling a terminal operation

@JosePaumard
#J8Stream
What is a Stream?
And bytheway:
We can only have one terminal operation for a Stream
In case, another Stream must be built

@JosePaumard
#J8Stream
How one can build a Stream?
Let us have a look at the stream() default method
List<Person> persons = new ArrayList<>() ;
// Collection interface default Stream<E> stream() { return StreamSupport.stream(spliterator(), false); }

@JosePaumard
#J8Stream
Let us have a look at the stream() default method
// StreamSupport class
public static <T> Stream<T> stream(
Spliterator<T> spliterator, boolean parallel) {
Objects.requireNonNull(spliterator) ;
return new ReferencePipeline.Head<>(
spliterator,
StreamOpFlag.fromCharacteristics(spliterator),
parallel) ;
}

@JosePaumard
#J8Stream
The spliterator() is defined in ArrayList
// ArrayList class @Override public Spliterator<E> spliterator() { return new ArrayListSpliterator<>(this, 0, -1, 0); }

@JosePaumard
#J8Stream
The spliterator() is defined in ArrayList
The Spliterator defines how to access the data
ArrayList: it reads the array
// ArrayList class
@Override
public Spliterator<E> spliterator() {
return new ArrayListSpliterator<>(this, 0, -1, 0);
}

@JosePaumard
#J8Stream
Spliterator
Let us see the method we need to implement
Consumes the next element, if it exists
// Spliterator interface boolean tryAdvance(Consumer<? super T> action);

@JosePaumard
#J8Stream
Spliterator
Consumes the next element, if it exists
A Stream does not assume that all the elements are available at build time
// Spliterator interface
boolean tryAdvance(Consumer<? super T> action);

@JosePaumard
#J8Stream
Spliterator
Used in parallel computations: tries to split the data in two (fork / join)
// Spliterator interface Spliterator<T> trySplit();

@JosePaumard
#J8Stream
Spliterator
Returns an estimation of the size of this stream
long estimateSize();

@JosePaumard
#J8Stream
Spliterator
And there are also default methods
// Spliterator interface default void forEachRemaining(Consumer<? super T> action) { do { } while (tryAdvance(action)); }
default long getExactSizeIfKnown() {
return (characteristics() & SIZED) == 0 ? -1L : estimateSize();
}

@JosePaumard
#J8Stream
Spliterator
And there is a last method:
Implémentation for ArrayList
// interface Spliterator
int characteristics();
// pour ArrayList
public int characteristics() {
return Spliterator.ORDERED |
Spliterator.SIZED |
Spliterator.SUBSIZED;
}

@JosePaumard
#J8Stream
Spliterator
And there is a last method:
Implémentation for HashSet
// interface Spliterator
int characteristics();
// pour HashSet
public int characteristics() {
return (fence < 0 || est == map.size ? Spliterator.SIZED : 0) |
Spliterator.DISTINCT;
}

@JosePaumard
#J8Stream
Characteristics of a Stream
Bits defined in the characteristics word
Characteristic
ORDERED
The order matters
DISTINCT
No duplication
SORTED
Sorted
SIZED
The size is known
NONNULL
No null values
IMMUTABLE
Immutable
CONCURRENT
Parallelism is possible
SUBSIZED
The size is known

@JosePaumard
#J8Stream
Some operations will switch some bits
Method
Set to 0
Set to 1
filter()
SIZED
-
map()
DISTINCT, SORTED
-
flatMap()
DISTINCT, SORTED, SIZED
-
sorted()
-
SORTED, ORDERED
distinct()
-
DISTINCT
limit()
SIZED
-
peek()
-
-
unordered()
ORDERED
-

@JosePaumard
#J8Stream
Characteristics are taken into account when computing a result
// HashSet
HashSet<String> strings = ... ;
strings.stream()
.distinct() // no processing is triggered
.sorted()

@JosePaumard
#J8Stream
Characteristics are taken into account when computing a result
// SortedSet
TreeSet<String> strings = ... ;
strings.stream()
.distinct()
.sorted() // no processing is triggered

@JosePaumard
#J8Stream
Stream implementation
Complex!
Two parts:
1)algorithms are in ReferencePipeline
2)data access: Spliterator, can be overriden

@JosePaumard
#J8Stream
A few words on Comparator
We wrote:
// Comparator interface
Comparator cmp = Comparator.naturalOrder() ;

@JosePaumard
#J8Stream
We wrote:
// Comparator interface Comparator cmp = Comparator.naturalOrder() ;
@SuppressWarnings("unchecked")
public static <T extends Comparable<? super T>>
Comparator<T> naturalOrder() {
return (Comparator<T>) Comparators.NaturalOrderComparator.INSTANCE;
}

@JosePaumard
#J8Stream
// Comparators class
enum NaturalOrderComparator
implements Comparator<Comparable<Object>> {
INSTANCE;
}

@JosePaumard
#J8Stream
INSTANCE;
}
public class MySingleton { INSTANCE; private MySingleton() {} public static MySingleton getInstance() { // some buggy double-checked locking code return INSTANCE; } }

@JosePaumard
#J8Stream
INSTANCE;
public int compare(Comparable<Object> c1, Comparable<Object> c2) {
return c1.compareTo(c2);
}
public Comparator<Comparable<Object>> reversed() {
return Comparator.reverseOrder();
}
}

@JosePaumard
#J8Stream
So we can write:
Comparator<Person> cmp =
Comparator.comparing(Person::getLastName)
.thenComparing(Person::getFirstName)
.thenComparing(Person::getAge) ;

@JosePaumard
#J8Stream
Method comparing()
public static
<T, U> Comparator<T>
comparing(Function<T, U> keyExtractor) {
Objects.requireNonNull(keyExtractor);
return
(c1, c2) ->
keyExtractor.apply(c1).compareTo(keyExtractor.apply(c2));
}

@JosePaumard
#J8Stream
Method comparing()
public static
<T, U extends Comparable> Comparator<T>
comparing(Function<T, U> keyExtractor) {
return (Comparator<T>)
(c1, c2) ->
}

@JosePaumard
#J8Stream
Method comparing()
public static
<T, U extends Comparable<? super U>> Comparator<T>
comparing(Function<? super T, ? extends U> keyExtractor) {
return (Comparator<T> & Serializable)
(c1, c2) ->
}

@JosePaumard
#J8Stream
Method thenComparing()
default
 Comparator<T>
thenComparing(Function<T, U> keyExtractor) {
return thenComparing(comparing(keyExtractor));
}

@JosePaumard
#J8Stream
default
> Comparator<T>
thenComparing(Function<? super T, ? extends U> keyExtractor) {
return thenComparing(comparing(keyExtractor));
}

@JosePaumard
#J8Stream
default Comparator<T> thenComparing(Comparator<? super T> other) {
Objects.requireNonNull(other);
return (Comparator<T> & Serializable) (c1, c2) -> {
int res = compare(c1, c2);
return (res != 0) ? res : other.compare(c1, c2);
};
}

@JosePaumard
#J8Stream
API Stream
-intermediate operations
-terminal operations
-implementation built on 2 classes

@JosePaumard
#J8Stream
Stateless / stateful operations
This code:
… selects the 1000 first people of the list
ArrayList<Person> persons = ... ;
Stream<Persons> stream = persons.limit(1_000) ;

@JosePaumard
#J8Stream
This code:
… selects the 1000 first people of the list
This operations needs a counter
How will it work in parallel?
ArrayList<Person> persons = ... ; Stream<Persons> stream = persons.limit(1_000) ;

@JosePaumard
#J8Stream
This code:
List<String> names =
persons.map(Person::getLastName)

@JosePaumard
#J8Stream
This code:
« the order of the names is the same as the order of the people »
ArrayList<Person> persons = ... ; List<String> names = persons.map(Person::getLastName) .collect(Collectors.toList()) ;

@JosePaumard
#J8Stream
This code:
How will it behave in parallel?
List<String> names =
persons.map(Person::getLastName)

@JosePaumard
#J8Stream
This code:
We should relax this constraint!
ArrayList<Person> persons = ... ; List<String> names = persons.map(Person::getLastName) .unordered() .collect(Collectors.toList()) ;

@JosePaumard
#J8Stream
Wrap-up on Stream
One can define operations on a Stream:
-intermediate & terminal
-stateless & stateful
A Stream can be processed in parallel
A Stream has a state, used to optimize computations

@JosePaumard
#J8Stream
Stream & performance
Two elements:
-lazy processing, on one pass over the data
-parallel processing

@JosePaumard
#J8Stream
Two elements:
-lazy processing, on one pass over the data
-parallel processing
Stream<T> versus IntStream, LongStream, DoubleStream

@JosePaumard
#J8Stream
Let see this example again:
persons.stream()
.map(Person::getAge)
.filter(age -> age > 20)
.sum() ;

@JosePaumard
#J8Stream
persons.stream() // Stream<Person>
.map(Person::getAge)
.sum() ;

@JosePaumard
#J8Stream
.map(Person::getAge) // Stream<Integer> boxing 
.sum() ;

@JosePaumard
#J8Stream
.filter(age -> age > 20) // Stream<Integer> re-boxing re-
.sum() ;

@JosePaumard
#J8Stream
.sum() ; // no such sum() method on Stream<T>

@JosePaumard
#J8Stream
Let see this example again
.mapToInt(age -> age.getValue()) // IntStream 
.sum() ;

@JosePaumard
#J8Stream
.mapToInt(Integer::getValue) // IntStream 
.sum() ;

@JosePaumard
#J8Stream
.mapToInt(Person::getAge) // IntStream 
.filter(age -> age > 20) // IntStream 
// .mapToInt(Integer::getValue)
.sum() ;

@JosePaumard
#J8Stream
int sum =
.sum() ;

@JosePaumard
#J8Stream
??? =
.max() ;

@JosePaumard
#J8Stream
OptionalInt opt =
.max() ;

@JosePaumard
#J8Stream
??? =
.average() ;

@JosePaumard
#J8Stream
OptionalInt opt =
.average() ;

@JosePaumard
#J8Stream
sur l’exemple
In one pass we have: count, sum, min, max, average
IntSummaryStatistics stats =
persons.stream()
.mapToInt(Person::getAge)
.summaryStatistics() ;

@JosePaumard
#J8Stream
Parallel streams
How to build a parallel Stream?
Built on top of fork / join
Stream<Person> stream1 = persons.parallelStream() ;
Stream<Person> stream2 = persons.stream().parallel() ;

@JosePaumard
#J8Stream
Parallel streams
By default, a parallel streams uses the default « fork join common pool », defined at the JVM level
The same pool is shared by all the streams

@JosePaumard
#J8Stream
Parallelism
By default, a parallel streams uses the default « fork join common pool », defined at the JVM level
By default, the parallelism level is the # of cores
We can set it with a sytem property:
System.setProperty(
"java.util.concurrent.ForkJoinPool.common.parallelism", 2) ;

@JosePaumard
#J8Stream
Parallelism
We can also decide the FJ pool for a given Stream
ForkJoinPool fjp = new ForkJoinPool(2) ;
fjp.submit(
() -> //
persons.stream().parallel() // this is an implementation
.mapToInt(p -> p.getAge()) // of Callable<Integer>
.filter(age -> age > 20) //
.average() //
).get() ;

@JosePaumard
#J8Stream
Reduction
A reduction can be seen as a SQL agregation (sum, min, max, avg, etc…)

@JosePaumard
#J8Stream
Reduction
A reduction can be seen as a SQL agregation (sum, min, max, avg, etc…)
But it can be seen in a more general sense

@JosePaumard
#J8Stream
Reduction
Example: sum()

@JosePaumard
#J8Stream
Reduction
Example: sum()
The result is an integer, its initial value being 0

@JosePaumard
#J8Stream
Reduction
Example: sum()
Adding an element to the result is… adding

@JosePaumard
#J8Stream
Reduction
Example: sum()
Adding an element to the result is… adding
Combining two partial results (think parallel) is also « adding »

@JosePaumard
#J8Stream
Reduction
So we have 3 operations:

@JosePaumard
#J8Stream
Reduction
-the definition of a container, which holds the result

@JosePaumard
#J8Stream
Reduction
-adding an element to that container

@JosePaumard
#J8Stream
Reduction
-combining two partialy filled containers

@JosePaumard
#J8Stream
Reduction
-combining two partialy filled containers
This is a much more broader vision than just an agregation

@JosePaumard
#J8Stream
Reduction
So a reduction is based on 3 operations:
-a constructor: Supplier
-an accumulator: Function
-a combiner: Function

@JosePaumard
#J8Stream
Reduction
() -> new StringBuffer() ;

@JosePaumard
#J8Stream
Reduction
(StringBuffer sb, String s) -> sb.append(s) ;

@JosePaumard
#J8Stream
Reduction
(StringBuffer sb, String s) -> sb.append(s) ;
(StringBuffer sb1, StringBuffer sb2) -> sb1.append(sb2) ;

@JosePaumard
#J8Stream
Reduction
StringBuffer::new ;
StringBuffer::append ;
StringBuffer::append ;

@JosePaumard
#J8Stream
Reduction: implementation
Example 1: reducing in a StringBuffer
List<Person> persons = ... ; StringBuffer result = persons.stream() .filter(person -> person.getAge() > 20) .map(Person::getLastName) .collect( StringBuffer::new, // constructor StringBuffer::append, // accumulator StringBuffer::append // combiner ) ;

@JosePaumard
#J8Stream
Reduction: collectors
Example 1: using a Collectors
List<Person> persons = ... ; String result = persons.stream() .filter(person -> person.getAge() > 20) .map(Person::getLastName) .collect( Collectors.joining() ) ;

@JosePaumard
#J8Stream
Collecting in a String
List<Person> persons = ... ; String result = persons.stream() .filter(person -> person.getAge() > 20) .map(Person::getLastName) .collect( Collectors.joining(", ") ) ;

@JosePaumard
#J8Stream
Collecting in a String
List<Person> persons = ... ; String result = persons.stream() .filter(person -> person.getAge() > 20) .map(Person::getLastName) .collect( Collectors.joining(", ", "{", "}") ) ;

@JosePaumard
#J8Stream
Reduction
Example 2:
() -> new ArrayList() ;
(list, element) -> list.add(element) ;
(list1, list2) -> list1.addAll(list2) ;

@JosePaumard
#J8Stream
Reduction
Example 2:
ArrayList::new ;
Collection::add ;
Collection::allAll ;

@JosePaumard
#J8Stream
Example 2: reduction in a List
List<Person> persons = ... ; ArrayList<String> names = persons.stream() .filter(person -> person.getAge() > 20) .map(Person::getLastName) .collect( ArrayList::new, // constructor Collection::add, // accumulator Collection::addAll // combiner ) ;

@JosePaumard
#J8Stream
Example 2: reduction in a List
List<Person> persons = ... ; List<String> result = persons.stream() .filter(person -> person.getAge() > 20) .map(Person::getLastName) .collect( Collectors.toList() ) ;

@JosePaumard
#J8Stream
Example 3: reduction in a Set
List<Person> persons = ... ; HashSet<String> names = persons.stream() .filter(person -> person.getAge() > 20) .map(Person::getLastName) .collect( HashSet::new, // constructor Collection::add, // accumulator Collection::addAll // combiner ) ;

@JosePaumard
#J8Stream
Example 3: reduction in a Set
List<Person> persons = ... ; Set<String> result = persons.stream() .filter(person -> person.getAge() > 20) .map(Person::getLastName) .collect( Collectors.toSet() ) ;

@JosePaumard
#J8Stream
Example 4: reduction in a given collection
List<Person> persons = ... ; TreeSet<String> result = persons.stream() .filter(person -> person.getAge() > 20) .map(Person::getLastName) .collect( Collectors.toCollection(TreeSet::new) ) ;

@JosePaumard
#J8Stream
Collectors
Collectors class: 33 static methods
An amazing toolbox!

@JosePaumard
#J8Stream
Collectors: groupingBy
HashMap: age / list of the people
Map<Integer, List<Person>> result =
persons.stream()
.collect(
Collectors.groupingBy(Person::getAge)
) ;

@JosePaumard
#J8Stream
HashMap: age / # of people
Map<Integer, Long> result =
persons.stream()
.collect(
Collectors.groupingBy(
Person::getAge,
Collectors.counting() // « downstream collector »
)
) ;

@JosePaumard
#J8Stream
HashMap: age / list of the names
Map<Integer, List<String>> result =
persons.stream()
.collect(
Person::getAge,
Collectors.mapping( //
Person::getLastName // downstream collector
) //
)
) ;

@JosePaumard
#J8Stream
HashMap: age / names joined in a single String
Map<Integer, String> result =
persons.stream()
.collect(
Person::getAge,
Collectors.mapping( // 1st downstream collector
Person::getLastName
Collectors.joining(", ") // 2nd downstream collector
)
)
) ;

@JosePaumard
#J8Stream
HashMap: age / names sorted alphabetically
Map<Integer, TreeSet<String>> result =
persons.stream()
.collect(
Person::getAge,
Collectors.mapping(
Person::getLastName
Collectors.toCollection(TreeSet::new)
)
)
) ;

@JosePaumard
#J8Stream
HashMap: the same, sorted by age
TreeMap<Integer, TreeSet<String>> result =
persons.stream()
.collect(
Person::getAge,
TreeMap::new,
Collectors.mapping(
Person::getLastName
Collectors.toCollection(TreeSet::new)
)
)
) ;

@JosePaumard
#J8Stream
Wrap-up
Stream + Collectors =
New tools for data processing (map / filter / reduce)
1)lazy execution
2)parallelism
Live coding to come!

@JosePaumard
#J8Stream
Wrap-up
More to come:
-use cases
-live coding!

@JosePaumard
Coffee time !
#Stream8

« movies and actors »
https://github.com/JosePaumard/jdk8-lambda-tour
http://introcs.cs.princeton.edu/java/data/

@JosePaumard
#J8Stream
Movies and actors
File: « movies-mpaa.txt »
With 14k movies from 1916 to 2004
With:
-the title of the movie
-the release year
-the list of the actors (170k)

@JosePaumard
#J8Stream
Greatest release year
1st question: which year saw the most movies?

@JosePaumard
#J8Stream
1st step: we can build a hashmap
-the keys are the release year
-the values the # of movies in that year

@JosePaumard
#J8Stream
1st step: we can build a hashmap
-the keys are the release year
-the values the # of movies in that year
2nd step: get the key value pair with the greatest value

@JosePaumard
#J8Stream
Live coding

@JosePaumard
#J8Stream
Solution: map / reduce

@JosePaumard
#J8Stream
Most seen actor
2nd question: most seen actor?

@JosePaumard
#J8Stream
Most seen actor
Once again, we can build a hashmap:
-the keys are the actors
-the values are the # of movies

@JosePaumard
#J8Stream
Most seen actor
Let us build the set of all the actors
Set<Actor> actors =
movies.stream()
.flatMap(movie -> movie.actors().stream())
.collect(Collector.toSet()) ;

@JosePaumard
#J8Stream
Most seen actor
actors.stream() .collect( ) .entrySet().stream() .max(Map.Entry.comparingByValue()).get() ;

@JosePaumard
#J8Stream
Most seen actor
actors.stream()
.collect(
Collectors.toMap(
)
)
.entrySet().stream()
.max(Map.Entry.comparingByValue()).get() ;

@JosePaumard
#J8Stream
Most seen actor
actors.stream() .collect( Collectors.toMap( actor -> actor, ) ) .entrySet().stream() .max(Map.Entry.comparingByValue()).get() ;

@JosePaumard
#J8Stream
Most seen actor
actors.stream()
.collect(
Collectors.toMap(
Function.identity(),
)
)

@JosePaumard
#J8Stream
Most seen actor
actors.stream() .collect( Collectors.toMap( Function.identity(), actor -> movies.stream() .filter(movie -> movie.actors().contains(actor)) .count() ) ) .entrySet().stream() .max(Map.Entry.comparingByValue()).get() ;

@JosePaumard
#J8Stream
Most seen actor
actors.stream().parallel() // T = 40s 
.collect(
Collectors.toMap(
Function.identity(),
actor -> movies.stream()
.filter(movie -> movie.actors().contains(actor))
.count()
)
)

@JosePaumard
#J8Stream
Most seen actor
actors.stream().parallel() // 170k .collect( Collectors.toMap( Function.identity(), actor -> movies.stream() // 14k 2,4G!! .filter(movie -> movie.actors().contains(actor)) .count() ) ) .entrySet().stream() .max(Map.Entry.comparingByValue()).get() ;

@JosePaumard
#J8Stream
Most seen actor
Set<Actor> actors =
movies.stream()
.flatMap(movie -> movie.actors().stream()) // stream of actors
.collect(Collector.toSet()) ;

@JosePaumard
#J8Stream
Most seen actor
And in this stream, if an actor played in 2 movies, she appears twice
Set<Actor> actors = movies.stream() .flatMap(movie -> movie.actors().stream()) // stream of actors .collect(Collector.toSet()) ;

@JosePaumard
#J8Stream
Most seen actor
Live coding

@JosePaumard
#J8Stream
Most seen actor
Answer: Frank Welker

@JosePaumard
#J8Stream
Most seen actor
… and in fact he’s not really the most « seen » actor

@JosePaumard
#J8Stream
Most seen actor
2nd question: most seen heard actor?

@JosePaumard
#J8Stream
Most seen actor in a year
3rd question: most seen actor in a given year?

@JosePaumard
#J8Stream
So it is a hashmap in which:
-the values are other hashmaps

@JosePaumard
#J8Stream
-in which keys are the release years
-and the values the # of movies → AtomicLong

@JosePaumard
#J8Stream
movies.stream()
.flatMap(movie -> movie.actors().stream())
.collect(
Collectors.groupingBy( // build a Map<year, ...>
movie -> movie.releaserYear(),
// build a Map<Actor, # of movies>
)
)

@JosePaumard
#J8Stream
movies.stream() .flatMap(movie -> movie.actors().stream()) .collect( Collectors.groupingBy( // build a Map<year, ...> movie -> movie.releaserYear(), // build a Map<Actor, # of movies> -> custom DS Collector ) ) .entrySet().stream() .max(Map.Entry.comparingByValue()).get() ;

@JosePaumard
#J8Stream
-in which keys are the release years
-and the values the # of movies → AtomicLong
And we want the max by the # of movies

@JosePaumard
#J8Stream
We need 3 things to build a collector
-a constructor for the resulting container
-an accumulator to add an element to the container
-a combiner that merges two partially filled containers

@JosePaumard
#J8Stream
Live coding

@JosePaumard
#J8Stream
Answer: Phil Hawn, in 1999 played in 24 movies

« Shakespeare plays Scrabble »
http://introcs.cs.princeton.edu/java/data/

@JosePaumard
#J8Stream
Scrabble
Two files:
-OSPD = official personnal Scrabble dictionary
-The words used by Shakespeare

@JosePaumard
#J8Stream
Scrabble
Two files:
-OSPD = official personnal Scrabble dictionary
-The words used by Shakespeare
-And some documentation (Wikipedia)
private static final int [] scrabbleENScore = {
// a, b, c, d, e, f, g, h, i, j, k, l, m, n, o, p, q, r, s, t, u, v, w, x, y, z
1, 3, 3, 2, 1, 4, 2, 4, 1, 8, 5, 1, 3, 1, 1, 3, 10, 1, 1, 1, 1, 4, 4, 8, 4, 10} ;

@JosePaumard
#J8Stream
Scrabble: score of a word
1st question: how to compute the score of a word
H E L L O 4 1 1 1 1 // from the scrabbleENScore array

@JosePaumard
#J8Stream
H E L L O
4 1 1 1 1 // from the scrabbleENScore array
SUM
= 8

@JosePaumard
#J8Stream
H E L L O // this is a stream of the letters of the word HELLO 4 1 1 1 1 SUM = 8

@JosePaumard
#J8Stream
H E L L O // this is a stream of the letters of the word HELLO
4 1 1 1 1 // mapping : letter -> score of the letter
SUM
= 8

@JosePaumard
#J8Stream
H E L L O // this is a stream of the letters of the word HELLO 4 1 1 1 1 // mapping : letter -> score of the letter SUM // reduction : sum = 8

@JosePaumard
#J8Stream
Live coding

@JosePaumard
#J8Stream
= map / reduce

@JosePaumard
#J8Stream
Scrabble: score of Shakespeare
2nd question: compute the best word of Shakespeare

@JosePaumard
#J8Stream
1)Histogram: score of the words / list of the words

@JosePaumard
#J8Stream
2)Max by key value

@JosePaumard
#J8Stream
HashMap, Collectors.groupingBy()
2)Max by key value

@JosePaumard
#J8Stream
HashMap, Collectors.groupingBy()
2)Max by key value
Shouldnt be too hard

@JosePaumard
#J8Stream
Live coding

@JosePaumard
#J8Stream
3rd question: filter out the wrong words
We need to add a filtering step

@JosePaumard
#J8Stream
3rd question: filter out the wrong words
Solution: map / filter / reduce

@JosePaumard
#J8Stream
Question: can one really write this word?

@JosePaumard
#J8Stream
Question: can one really write this word?
private static final int [] scrabbleENDistribution = {
9, 2, 2, 1, 12, 2, 3, 2, 9, 1, 1, 4, 2, 6, 8, 2, 1, 6, 4, 6, 4, 2, 2, 1, 2, 1} ;

@JosePaumard
#J8Stream
Answer: no…
9, 2, 2, 1, 12, 2, 3, 2, 9, 1, 1, 4, 2, 6, 8, 2, 1, 6, 4, 6, 4, 2, 2, 1, 2, 1} ;

@JosePaumard
#J8Stream
Filtering V2
4th question: how to check if I have the available letters?
9, 2, 2, 1, 12, 2, 3, 2, 9, 1, 1, 4, 2, 6, 8, 2, 1, 6, 4, 6, 4, 2, 2, 1, 2, 1} ;

@JosePaumard
#J8Stream
Filtering V2
1
1
2
2
1
1
needed letters

@JosePaumard
#J8Stream
Filtering V2
1
1
2
2
1
1
1
4
9
1
6
3
needed letters
available letters

@JosePaumard
#J8Stream
Filtering V2
I need a allMatch() reducer!
1
1
2
2
1
1
1
4
9
1
6
3
needed letters
available letters

@JosePaumard
#J8Stream
Filtering V2
Live coding

@JosePaumard
#J8Stream
Filtering V2

@JosePaumard
#J8Stream
Scrabble: use of blanks
5th question: and what about using blanks?

@JosePaumard
#J8Stream
5th question: and what about using blanks?
9, 2, 2, 1, 12, 2, 3, 2, 9, 1, 1, 4, 2, 6, 8, 2, 1, 6, 4, 6, 4, 2, 2, 1, 2, 1} ;

@JosePaumard
#J8Stream
It has 2 impacts:
-on the filtering of the words
-on the computation of the score

@JosePaumard
#J8Stream
5th question: how to compute the score?
1
1
2
2
1
1

@JosePaumard
#J8Stream
1
1
2
2
1
1
1
4
9
1
6
3

@JosePaumard
#J8Stream
1
1
2
2
1
1
1
4
9
1
6
3
10
1
2
10
1
2
→ 26

@JosePaumard
#J8Stream
Live coding

@JosePaumard
#J8Stream

@JosePaumard
#J8Stream
6th question: how many blanks do we need?
1
1
2
2
1
1

@JosePaumard
#J8Stream
1
1
2
2
1
1
1
4
9
1
6
3

@JosePaumard
#J8Stream
1
1
2
2
1
1
1
4
9
1
6
3
0
-3
-7
+1
-5
-2

@JosePaumard
#J8Stream
1
1
2
2
1
1
1
4
9
1
6
3
0
0
0
+1
0
0
→ 1

@JosePaumard
#J8Stream
Use of a blank

@JosePaumard
#J8Stream
Scrabble: and what about the board?
7th question: what about the Double letter score?

@JosePaumard
#J8Stream
6 solutions to put the word:
-right square = 3 last letters
-left square= 3 first letters

@JosePaumard
#J8Stream
6 solutions to put the word:
-right square = 3 last letters
-left square= 3 first letters
If the word is long enough (7 letters)!

@JosePaumard
#J8Stream
So we need to take a max, on which set?
word.chars().skip(4) // 1st stream
word.chars().limit(Integer.max(0, word.length() - 4)) // 2nd stream

@JosePaumard
#J8Stream
IntStream.concat(
word.chars().skip(4),
word.chars().limit(Integer.max(0, word.length() - 4))
)
.map(scrabbleENScore)
.max()

@JosePaumard
#J8Stream
Problem: concat does not parallel well
IntStream.concat(
)
.max()

@JosePaumard
#J8Stream
Stream.of(
) // returns a Stream of Stream!

@JosePaumard
#J8Stream
Stream.of(
)
.flatMapToInt(Function.identity()) // chars() is an IntStream
.max() // we have an IntStream!

@JosePaumard
#J8Stream
What should we do with this Optional?
Stream.of(
)

@JosePaumard
#J8Stream
… that can be empty!
Stream.of(
)

@JosePaumard
#J8Stream
… that can be empty!
Stream.of(
)
.orElse(0) ;

@JosePaumard
#J8Stream
Live coding

@JosePaumard
#J8Stream
7th question: can we take the Double letter score?

@JosePaumard
#J8Stream
Conclusion
Streams + Collectors = no need to use the Iterator pattern any more

@JosePaumard
#J8Stream
Conclusion
Combined with lambdas = new way of writing data processing and applications in Java

@JosePaumard
#J8Stream
Conclusion
Combined with lambdas = new way of writing data processing in Java
-writing code
-architecture

@JosePaumard
#J8Stream
Conclusion
Some books:

Thank you!
@JosePaumard
#Stream8

Java 8, Streams & Collectors, patterns, performances and parallelization

Recommended

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Java 8, Streams & Collectors, patterns, performances and parallelization (20)

More from José Paumard (19)

Recently uploaded (20)

Java 8, Streams & Collectors, patterns, performances and parallelization