Lambdas can only operate on a functional interface, which is an interface with just one abstract method. Functional interfaces can have any number of default
or static
methods. (For this reason, they are sometimes referred to as Single Abstract Method Interfaces, or SAM Interfaces).
interface Foo1 {
void bar();
}
interface Foo2 {
int bar(boolean baz);
}
interface Foo3 {
String bar(Object baz, int mink);
}
interface Foo4 {
default String bar() { // default so not counted
return "baz";
}
void quux();
}
When declaring a functional interface the @FunctionalInterface
annotation can be added. This has no special effect, but a compiler error will be generated if this annotation is applied to an interface which is not functional, thus acting as a reminder that the interface should not be changed.
@FunctionalInterface
interface Foo5 {
void bar();
}
@FunctionalInterface
interface BlankFoo1 extends Foo3 { // inherits abstract method from Foo3
}
@FunctionalInterface
interface Foo6 {
void bar();
boolean equals(Object obj); // overrides one of Object's method so not counted
}
Conversely, this is not a functional interface, as it has more than one abstract method:
interface BadFoo {
void bar();
void quux(); // <-- Second method prevents lambda: which one should
// be considered as lambda?
}
This is also not a functional interface, as it does not have any methods:
interface BlankFoo2 { }
Take note of the following. Suppose you have
interface Parent { public int parentMethod(); }
and
interface Child extends Parent { public int ChildMethod(); }
Then Child
cannot be a functional interface since it has two
specified methods.
Java 8 also provides a number of generic templated functional interfaces in the package java.util.function
. For example, the built-in interface Predicate<T>
wraps a single method which inputs a value of type T
and returns a boolean
.
The basic structure of a Lambda expression is:
fi
will then hold a singleton instance of a class, similar to an anonymous class, which implements FunctionalInterface
and where the one method's definition is { System.out.println("Hello"); }
. In other words, the above is mostly equivalent to:
FunctionalInterface fi = new FunctionalInterface() {
@Override
public void theOneMethod() {
System.out.println("Hello");
}
};
The lambda is only "mostly equivalent" to the anonymous class because in a lambda, the meaning of expressions like this
, super
or toString()
reference the class within which the assignment takes place, not the newly created object.
You cannot specify the name of the method when using a lambda—but you shouldn't need to, because a functional interface must have only one abstract method, so Java overrides that one.
In cases where the type of the lambda is not certain, (e.g. overloaded methods) you can add a cast to the lambda to tell the compiler what its type should be, like so:
Object fooHolder = (Foo1) () -> System.out.println("Hello");
System.out.println(fooHolder instanceof Foo1); // returns true
If the functional interface's single method takes parameters, the local formal names of these should appear between the brackets of the lambda. There is no need to declare the type of the parameter or return as these are taken from the interface (although it is not an error to declare the parameter types if you want to). Thus, these two examples are equivalent:
Foo2 longFoo = new Foo2() {
@Override
public int bar(boolean baz) {
return baz ? 1 : 0;
}
};
Foo2 shortFoo = (x) -> { return x ? 1 : 0; };
The parentheses around the argument can be omitted if the function only has one argument:
Foo2 np = x -> { return x ? 1 : 0; }; // okay
Foo3 np2 = x, y -> x.toString() + y // not okay
If the code placed inside a lambda is a Java expression rather than a statement, it is treated as a method which returns the value of the expression. Thus, the following two are equivalent:
IntUnaryOperator addOneShort = (x) -> (x + 1);
IntUnaryOperator addOneLong = (x) -> { return (x + 1); };
Since lambdas are syntactic shorthand for anonymous classes, they follow the same rules for accessing local variables in the enclosing scope; the variables must be treated as final
and not modified inside the lambda.
IntUnaryOperator makeAdder(int amount) {
return (x) -> (x + amount); // Legal even though amount will go out of scope
// because amount is not modified
}
IntUnaryOperator makeAccumulator(int value) {
return (x) -> { value += x; return value; }; // Will not compile
}
If it is necessary to wrap a changing variable in this way, a regular object that keeps a copy of the variable should be used. Read more in Java Closures with lambda expressions.
Because a lambda is an implementation of an interface, nothing special needs to be done to make a method accept a lambda: any function which takes a functional interface can also accept a lambda.
public void passMeALambda(Foo1 f) {
f.bar();
}
passMeALambda(() -> System.out.println("Lambda called"));
A lambda expression, by itself, does not have a specific type. While it is true that the types and number of parameters, along with the type of a return value can convey some type information, such information will only constrain what types it can be assigned to. The lambda receives a type when it is assigned to a functional interface type in one of the following ways:
myPredicate = s -> s.isEmpty()
stream.filter(s -> s.isEmpty())
return s -> s.isEmpty()
(Predicate<String>) s -> s.isEmpty()
Until any such assignment to a functional type is made, the lambda does not have a definite type. To illustrate, consider the lambda expression o -> o.isEmpty()
. The same lambda expression can be assigned to many different functional types:
Predicate<String> javaStringPred = o -> o.isEmpty();
Function<String, Boolean> javaFunc = o -> o.isEmpty();
Predicate<List> javaListPred = o -> o.isEmpty();
Consumer<String> javaStringConsumer = o -> o.isEmpty(); // return value is ignored!
com.google.common.base.Predicate<String> guavaPredicate = o -> o.isEmpty();
Now that they are assigned, the examples shown are of completely different types even though the lambda expressions looked the same, and they cannot be assigned to each other.