(int a) -> a * 2; // Calculate the double of a
a -> a * 2; // or simply without type
(a, b) -> a + b; // Sum of 2 parameters
If the lambda is more than one expression we can use { }
and return
(x, y) -> {
int sum = x + y;
int avg = sum / 2;
return avg;
}
A lambda expression cannot stand alone in Java, it need to be associated to a functional interface.
interface MyMath {
int getDoubleOf(int a);
}
MyMath d = a -> a * 2; // associated to the interface
d.getDoubleOf(4); // is 8
All examples with "list" use :
List<String> list = [Bohr, Darwin, Galilei, Tesla, Einstein, Newton]
sort sort(list, comparator)
list.sort((a, b) -> a.length() - b.length())
list.sort(Comparator.comparing(n -> n.length())); // same
list.sort(Comparator.comparing(String::length)); // same
//> [Bohr, Tesla, Darwin, Newton, Galilei, Einstein]
removeIf
list.removeIf(w -> w.length() < 6);
//> [Darwin, Galilei, Einstein, Newton]
merge
merge(key, value, remappingFunction)
Map<String, String> names = new HashMap<>();
names.put("Albert", "Ein?");
names.put("Marie", "Curie");
names.put("Max", "Plank");
// Value "Albert" exists
// {Marie=Curie, Max=Plank, Albert=Einstein}
names.merge("Albert", "stein", (old, val) -> old.substring(0, 3) + val);
// Value "Newname" don't exists
// {Marie=Curie, Newname=stein, Max=Plank, Albert=Einstein}
names.merge("Newname", "stein", (old, val) -> old.substring(0, 3) + val);
Allows to reference methods (and constructors) without executing them
// Lambda Form:
getPrimes(numbers, a -> StaticMethod.isPrime(a));
// Method Reference:
getPrimes(numbers, StaticMethod::isPrime);
Method Reference | Lambda Form |
---|---|
StaticMethod::isPrime |
n -> StaticMethod.isPrime(n) |
String::toUpperCase |
(String w) -> w.toUpperCase() |
String::compareTo |
(String s, String t) -> s.compareTo(t) |
System.out::println |
x -> System.out.println(x) |
Double::new |
n -> new Double(n) |
String[]::new |
(int n) -> new String[n] |
Similar to collections, but
- They don't store their own data
- The data comes from elsewhere (collection, file, db, web, ...)
- immutable (produce new streams)
- lazy (only computes what is necessary !)
// Will compute just 3 "filter"
Stream<String> longNames = list
.filter(n -> n.length() > 8)
.limit(3);
Create a new stream
Stream<Integer> stream = Stream.of(1, 2, 3, 5, 7, 11);
Stream<String> stream = Stream.of("Jazz", "Blues", "Rock");
Stream<String> stream = Stream.of(myArray); // or from an array
list.stream(); // or from a list
// Infinit stream [0; inf[
Stream<Integer> integers = Stream.iterate(0, n -> n + 1);
Collecting results
// Collect into an array (::new is the constructor reference)
String[] myArray = stream.toArray(String[]::new);
// Collect into a List or Set
List<String> myList = stream.collect(Collectors.toList());
Set<String> mySet = stream.collect(Collectors.toSet());
// Collect into a String
String str = list.collect(Collectors.joining(", "));
map map(mapper)
Applying a function to each element
// Apply "toLowerCase" for each element
res = stream.map(w -> w.toLowerCase());
res = stream.map(String::toLowerCase);
//> bohr darwin galilei tesla einstein newton
res = Stream.of(1,2,3,4,5).map(x -> x + 1);
//> 2 3 4 5 6
filter filter(predicate)
Retains elements that match the predicate
// Filter elements that begin with "E"
res = stream.filter(n -> n.substring(0, 1).equals("E"));
//> Einstein
res = Stream.of(1,2,3,4,5).filter(x -> x < 3);
//> 1 2
reduce
Reduce the elements to a single value
String reduced = stream
.reduce("", (acc, el) -> acc + "|" + el);
//> |Bohr|Darwin|Galilei|Tesla|Einstein|Newton
limit limit(maxSize)
The n first elements
res = stream.limit(3);
//> Bohr Darwin Galilei
skip Discarding the first n elements
res = strem.skip(2); // skip Bohr and Darwin
//> Galilei Tesla Einstein Newton
distinct Remove duplicated elemetns
res = Stream.of(1,0,0,1,0,1).distinct();
//> 1 0
sorted Sort elements (must be Comparable)
res = stream.sorted();
//> Bohr Darwin Einstein Galilei Newton Tesla
allMatch
// Check if there is a "e" in each elements
boolean res = words.allMatch(n -> n.contains("e"));
anyMatch: Check if there is a "e" in an element
noneMatch: Check if there is no "e" in elements
parallel Returns an equivalent stream that is parallel
findAny faster than findFirst on parallel streams
Wrappers (like Stream) are inefficients. It requires a lot of unboxing and boxing for each element. Better to use IntStream
, DoubleStream
, etc.
Creation
IntStream stream = IntStream.of(1, 2, 3, 5, 7);
stream = IntStream.of(myArray); // from an array
stream = IntStream.range(5, 80); // range from 5 to 80
Random gen = new Random();
IntStream rand = gen(1, 9); // stream of randoms
Use mapToX (mapToObj, mapToDouble, etc.) if the function yields Object, double, etc. values.
Collectors.groupingBy
// Groupe by length
Map<Integer, List<String>> groups = stream
.collect(Collectors.groupingBy(w -> w.length()));
//> 4=[Bohr], 5=[Tesla], 6=[Darwin, Newton], ...
Collectors.toSet
// Same as before but with Set
... Collectors.groupingBy(
w -> w.substring(0, 1), Collectors.toSet()) ...
Collectors.counting Count the number of values in a group
Collectors.summing__
summingInt
, summingLong
, summingDouble
to sum group values
Collectors.averaging__
averagingInt
, averagingLong
, ...
// Average length of each element of a group
Collectors.averagingInt(String::length)
PS: Don't forget Optional (like Map<T, Optional<T>>
) with some Collection methods (like Collectors.maxBy
).
Creation
Stream<String> parStream = list.parallelStream();
Stream<String> parStream = Stream.of(myArray).parallel();
unordered
Can speed up the limit
or distinct
stream.parallelStream().unordered().distinct();
PS: Work with the streams library. Eg. use filter(x -> x.length() < 9)
instead of a forEach
with an if
.
In Java, it is common to use null to denote absence of result.
Problems when no checks: NullPointerException
.
// Optional<String> contains a string or nothing
Optional<String> res = stream
.filter(w -> w.length() > 10)
.findFirst();
// length of the value or "" if nothing
int length = res.orElse("").length();
// run the lambda if there is a value
res.ifPresent(v -> results.add(v));
Return an Optional
Optional<Double> squareRoot(double x) {
if (x >= 0) { return Optional.of(Math.sqrt(x)); }
else { return Optional.empty(); }
}
Note on inferance limitations
interface Pair<A, B> {
A first();
B second();
}
A steam of type Stream<Pair<String, Long>>
:
stream.sorted(Comparator.comparing(Pair::first)) // ok
stream.sorted(Comparator.comparing(Pair::first).thenComparing(Pair::second)) // dont work
Java cannot infer type for the .comparing(Pair::first)
part and fallback to Object, on which Pair::first
cannot be applied.
The required type for the whole expression cannot be propagated through the method call (.thenComparing
) and used to infer type of the first part.
Type must be given explicitly.
stream.sorted(
Comparator.<Pair<String, Long>, String>comparing(Pair::first)
.thenComparing(Pair::second)
) // ok
This cheat sheet was based on the lecture of Cay Horstmann http://horstmann.com/heig-vd/spring2015/poo/