Whats new in Java 8

Udemy Course Companion

Toby Weston
This book is for sale at http://leanpub.com/whatsnewjava8
This version was published on 2014-05-22
This is a Leanpub book. Leanpub empowers authors and publishers with the Lean Publishing
process. Lean Publishing is the act of publishing an in-progress ebook using lightweight tools and
many iterations to get reader feedback, pivot until you have the right book and build traction once
you do.
2014 Toby Weston
Tweet This Book!
Please help Toby Weston by spreading the word about this book on Twitter!
The suggested hashtag for this book is #whatsnewjava.
Find out what other people are saying about the book by clicking on this link to search for this
hashtag on Twitter:
https://twitter.com/search?q=#whatsnewjava
Also By Toby Weston
Essential Acceptance Testing
Contents
About the Companion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i
Introduction to the Udemy Course . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Delivery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
What you can expect? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Registration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Whats new in Java 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Lambdas Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
s in Functional Programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Functions vs classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Basic Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
In-depth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Functional Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Type Inference Improvements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Target-typing for lambdas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Method References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Scoping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Effectively Final . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Exception Handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Lambdas vs Closures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Invocation & Bytecode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Bytecode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
About the Companion
This book is intended as a companion to my Whats new in Java 8 course available on Udemy.com.
However, you dont have to watch the online course to make the most of this book. Its an
independent read and contains all the examples from the course so you dont have watch the course.
The book will help you get to grips with the new features of Java 8; lambdas and their impact as
well as the new APIs and behind-the-scenes changes.
https://www.udemy.com/whats-new-in-java-8
https://www.udemy.com/whats-new-in-java-8
Introduction to the Udemy Course
Welcome to my course on Whats new in Java 8 and this companion eBook.
The course is split into two parts which you can buy separately from Udemy.com.
Part 1 is all about lambdas; their background, how Javas had to change to support them, their syntax,
technicalities and usage. In Part 2, we have a good look at the rest of the new features and the API
changes.
https://www.udemy.com/whats-new-in-java-8
2
Delivery
In terms of how I deliver the Udemy course. For each section, theres a mixture of slides and demos.
I try and keep the slides entertaining so you wont see slide after slide of boring bullet points. What
you will see should lead nicely into into the live demos.
The demos are where well go over most of the syntax. I want this to feel like a pairing session, so
youll see me code and well talk over what Im doing.
3
What you can expect?
After completing Part 1, youll know pretty much all you need to know about lambdas. I talk about
the lambdas in-depth so youll be well equipped to make your code more concise as you move
towards Java 8.
Part 2 covers pretty much all of the remaining. Youll get a good feel for whats available and for the
big features, well explore things in as much detail as possible.
Well question when these things are really useful and try and show practical examples along the
way.
4
Registration
If youd like to get updates to the book and exclusive special offers directly from me, no one else,
please register your course at http://bit.ly/whatsnewjava8 or if you prefer QR codes, you scan this
one.
http://bit.ly/whatsnewjava8
5
Whats new in Java 8
Its been two years, seven months, and eighteen days since the last release of Java. Oracle have put
this time to good use and really crammed in the features.
The headliners are of-course lambdas and and a retrofit to support functional programming
ideas. With languages like Scala taking center stage and the current trends towards functional
programming, Java had to do something to keep up.
Although Java is not and never will be a pure functional programming language, Java 8 does allow
you to use functional idioms more easily. With disciple and experience, you can now get a lot of the
benefits of functional programming without the third-party libraries.
Features
Its a big shift for Java and an exciting time to be a Java developer, so lets take a whistle-stop tour
of whats in the Java 8 release.
Lambda support finally comes to Java
and along with it, a bunch of the core APIs have been updated to take advantage of them,
including the collection APIs and a new functional package to help build functional constructs.
Entirely new APIs have also been developed that use lambdas, things like the stream API
which bring functional style processing of data.
Functions like map and flatmap from the stream API allow a declarative way to process lists
for example and move away from external iteration to internal iteration. This in turn allows
the library vendors to worry about the details and they can optimise processing however they
like. For example, Java now comes with a parallel way to process streams without bothering
the developer with the details.
It feels like nearly all of the core APIs have been touched by the release. New helper methods
have popped up for strings, collections, comparators, numbers and maths.
Some of the additions may have a big influence on how we code in the future. For example,
the Optional class will be familiar to some and allows a better way to avoid dealing with
nulls.
There are various concurrency library improvements. Things like an improved concurrent
hash map, completable futures, thread safe accumulators, an improved read write lock (called
a StampedLock), an implementation of a work stealing thread pool and much more besides.
Apart from lambda support, other language features have also been added. Things like support
for adding static methods to interfaces and default methods (otherwise known as virtual
extension or defender methods).
To support lambdas a bunch of stuff had to go on behind the scenes. Type inference had to be
improved and new constructs like functional interfaces and method references have all been
worked in.
6
A new date and time API has been introduced.
The IO and NIO packages have welcome additions to allow you to work with IO streams using
the new streams API.
Reflection and annotations have been improved.
If thats not enough, an entirely new JavaScript engine ships with Java 8. Nashorn replaced
Rhino and is faster and has better support for ECMA-Script.
The JVM has had a bunch of improvements; it could be the fastest JVM to date as the
integration with JRocket is now complete.
The JVM has dropped the idea of perm gen, instead using native OS memory for class
metadata. This is a huge deal and should in theory mean better memory utilitsation.
The JRocket integration also brings Mission control to the JDK as standard. It compliments
jconsole and visualvm with similar functionality but adds very inexpensive profiling.
The list is almost endless, improvements to JavaFX, base64 encoding support and Im sure I must
have missed some out.
Lambdas Introduction
In this section well introduce the ideas of lambdas, well;
discuss some background to lambdas and functional programming in general.
talk about functions vs. classes in Java
look at the basic syntax for lambdas in Java
8
s in Functional Programming
Before we look at things in a more depth, I want to give you some general background to lambdas.
If you havent seen it before, the Greek letter (lambda) is often used as shorthand when talking
about lambdas.
In computer science, lambdas go back to the lambda-calculus. A mathematical notation for functions
introduced by Alonzo Church in the 1930s. It was a way to explore mathematics using functions and
was later re-discovered as a useful tool in computer science.
It formalised the notion of lambda terms and rules to transform those terms. These rules or
functions map directly into modern computer science ideas. All functions in the lambda-calculus
are anonymous which again has been taken literally in computer science.
Heres an example of a lambda-calculus expression.
A lambda-calculus expression
x.x+1
We define an anonymous function or lambda with an argument x. The body follows the dot and
adds one to that argument.
Then in the 1950s, John McCarthy invented LISP whilst at MIT. This was a programming language
designed to model mathematical problems and was heavily influenced by the lambda-calculus.
It used the word lambda as an operator to define an anonymous function.
Heres an example.
A LISP expression
(lambda (arg) (+ arg 1))
This LISP expression evaluates to a function, that when applied will take a single argument, bind it
to arg and then add 1 to it.
The two expressions produce the same thing, a function to increment a number. You can see the two
are very similar.
The lambda-calculus and LISP have had a huge influence on functional programming. The ideas
of applying functions and reasoning about problems using functions has moved directly into
programming languages. Hence the use of the term in our field. A lambda in the calculus is the
same thing as in modern programming languages and is used in the same way.
9
What is a Lambda
In simple terms then, a lambda is just an anonymous function. Thats it. Nothing special, its just a
compact way to define a function. Anonymous functions are useful when you want to pass around
functionality. For example, passing functionality as a argument to a method.
Many main stream languages already support lambdas including Scala, C#, Objective-C, Ruby,
C++(11), Python and many others.
10
Functions vs classes
Bear in mind that an anonymous function isnt the same as an anonymous class in Java. An
anonymous class in Java still needs to be instantiated to an object. It may not have a proper name
but its only as an object that it can be useful.
A function on the other hand has no instance associated with it. Functions are disassociated with
the data they act on whereas an object is intimately associated with the data it acts upon.
Lambdas in Java 8
You can use lambdas in Java 8 anywhere you would have previously used a single method interface
so it may just look like syntactic sugar but its not. Lets have a look at how they differ and compare
anonymous classes to lambdas; classes vs. functions.
A typical implementation of an anonymous class (a single method interface) in Java pre-8, might
look something like this. The anonymousClass method is calling the waitFor method passing in some
implementation of Condition, in this case its saying wait for some server to have shutdown.
Typical usage of an anonymous class
void anonymousClass() {
final Server server = new HttpServer();
waitFor(new Condition() {
@Override
public Boolean isSatisfied() {
return !server.isRunning();
}
});
}
The functionally equivalent lambda would look like this.
Equivalent functionality as a lambda
void closure() {
Server server = new HttpServer();
waitFor(() -> !server.isRunning());
}
Where in the interest of completeness, a naive polling waitFor method might look like this.
11
class WaitFor {
static void waitFor(Condition condition) throws InterruptedException {
while (!condition.isSatisfied())
Thread.sleep(250);
}
}
Some Theoretical Differences
Firstly, both implementations are in-fact closures, the later is also a lambda. Well look at this
distinction in more detail later in the lambdas vs. closures section. This means that both have to
capture their environment at runtime. In Java pre-8, this means copying the things the closure
needs into an instance of an class (an anonymous instances of Condition). In our example, the
server variable.
As its a copy, it has to be declared final to ensure that it can not be changed between when its
captured and when its used. These two points in time could be very different given that closures
are often used to defer execution until some later point (see lazy evaluation for example). Java 8
uses a neat trick whereby if it can reason that a variable is never updated, it might as well be final
so it treats it as effectively final and you dont need to declare it as final explicitly.
A lambda on the other hand, doesnt need to copy its environment or capture any terms. This means
it can be treated as a genuine function and not an instance of a class. Whats the difference? Plenty.
Functions vs. Classes
For a start, functions; genuine functions, dont need to be instantiated many times. Im not sure
if instantiation is even the right word to use when talking about allocating memory and loading
a chunk of machine code as a function. The point is, once its available, it can be re-used, its
idempotent in nature as it retains no state. Static class methods are the closest thing Java has to
functions.
For Java, this means that a lambda need not be instantiated every time its evaluated which is a big
deal. Unlike instantiating an anonymous class, the memory impact should be minimal.
In terms of some conceptual differences then;
Classes must be instantiated, whereas functions are not.
When classes are newed up, memory is allocated for the object.
Memory need only be allocated once for functions. They are stored in the permanent area
of the heap.
Objects act on their own data, functions act on unrelated data.
Static class methods in Java are roughly equivalent to functions.
http://en.wikipedia.org/wiki/Lazy_evaluation
http://en.wikipedia.org/wiki/Pure_function
12
Some Concrete Differences
Capture Semantics
Another difference is around capture semantics for this. In an anonymous class, this refers to the
instance of the anonymous class. For example, Foo$InnerClass and not Foo. Thats why you have
crazy syntax like Foo.this.x when you refer to the enclosing scope from the anonymous class.
In lambdas on the other hand, this refers to the enclosing scope (Foo directly in our example). In
fact, lambdas are entirely lexically scoped, meaning they dont inherit any names from a super
type or introduce a new level of scoping at all; you can directly access fields, methods and local
variables from the enclosing scope.
For example, this class shows that the lambda can reference the firstName variable directly.
public class Example {
private String firstName = "Jack";
public void example() {
Function<String, String> addSurname = surname -> {
// equivalent to this.firstName
return firstName + " " + surname;
};
}
}
The anonymous class equivalent would need to explicitly refer to firstName from the enclosing
scope.
public class Example {
private String firstName = "Charlie";
public void anotherExample() {
Function<String, String> addSurname = new Function<String, String>() {
@Override
public String apply(String surname) {
return Example.this.firstName + " " + surname;
}
};
}
}
Shadowing also becomes much more straight forward to reason about (when referencing shadowed
variables).
13
Summary
Functions in the academic sense are very different things from anonymous classes (which we often
treat like functions in Java pre-8). Its useful to keep the distinctions in your head to be able to justify
the use of Java 8 lambdas for something more than just their concise syntax. Of course, theres lots
of additional advantages in using lambdas (not least the retrofit of the JDK to heavily use them).
When we take a look at the new lambda syntax next, remember that although lambdas are used in
a very similar way to anonymous classes in Java, they are technically different. Lambdas in Java
need not be instantiated every time theyre evaluated unlike an instance of an anonymous class.
This should serve to remind you that lambdas in Java 8 are not just syntactic sugar.
14
Basic Syntax
Lets jump in and have a look at the basic lambda syntax.
A lambda is basically an anonymous block of functionality. Its a lot like using an anonymous class
instance. For example, if we want to sort an array in Java, we can use the Arrays.sort method
which takes an instance of a Comparator.
It would look something like this.
Arrays.sort(numbers, new Comparator<Integer>() {
@Override
public int compare(Integer first, Integer second) {
return first.compareTo(second);
}
});
The Comparator instance here is a an abstract piece of the functionality; it means nothing on its only
when its used by the sort method that it has purpose.
Using Javas new syntax, you can replace this with a lambda which looks like this
Arrays.sort(numbers, (first, second) -> first.compareTo(second));
Its a more succinct way of achieving the same thing. In fact, Java treats this as if it were an instance
of the Comparator class. If we were to extract a variable for the lambda (the second parameter), its
type would be Comparator<Integer> just like the anonymous instance above.
Comparator<Integer> ascending = (first, second) -> first.compareTo(second);
Arrays.sort(numbers, ascending);
Because Comparator has only a single abstract method on it; compareTo, the compiler can piece
together that when we have an anonymous block like this, we really mean an instance of
Comparator. It can do this thanks to a couple of the other new features that well talk about later;
functional interfaces and improvements to type inference.
Syntax Breakdown
You can always convert from using a single abstract method to a using lambda.
Lets say we have an an interface Example with a method apply, returning some type and taking
some argument
15
interface Example {
R apply(A arg);
}
We could instantiate an instance with something like this;
new Example() {
@Override
public R apply(A args) {
body
}
};
And to convert to a lambda, we basically trim the fat. We drop the instantiation and annotation,
drop the method details which leaves just the argument list and the body.
(args) {
body
}
we then introduce the new arrow symbol to indicate both that the whole thing is a lambda and that
what follows is the body.
(args) -> {
body
}
and thats our basic lambda syntax.
Lets take the sorting example from earlier through these steps. We start with the anonymous
instance;
Arrays.sort(numbers, new Comparator<Integer>() {
@Override
public int compare(Integer first, Integer second) {
return first.compareTo(second);
}
});
and trim the instantiation and method signature
16
Arrays.sort(numbers, (Integer first, Integer second) {
return first.compareTo(second);
});
introduce the lambda
Arrays.sort(numbers, (Integer first, Integer second) -> {
return first.compareTo(second);
});
and were done. Theres a couple of optimisations we can do though. You can drop the types if the
compiler knows enough to infer them.
Arrays.sort(numbers, (first, second) -> {
return first.compareTo(second);
});
and for simple expressions, you can drop the braces to produce a lambda expression
Arrays.sort(numbers, (first, second) -> first.compareTo(second));
In this case, the compiler can infer enough to know what you mean. The single statement returns
a value consistent with the interface, so it says, no need to tell me that youre going to return
something, I can see that for myself!.
For single argument interface methods, you can even drop the first brackets. For example the lambda
taking an argument x and returning x + 1;
(x) -> x + 1
can be written without the brackets
x -> x + 1
Summary
Lets recap with a summary of the syntax options.
17
Syntax Summary
1 (int x, int y) -> { return x + y; }
2 (x, y) -> { return x + y; }
3 (x, y) -> x + y;
4 x -> x * 2
5 () -> System.out.println(Hey there!);
6 System.out::println;
The first example ((int x, int y) -> { return x + y; }) is the most verbose way to create a
lambda. The arguments to the function along with their types are in parenthesise, followed by the
new arrow syntax and then the body; the code block to be executed.
You can often drop the types from the argument list, like (x, y) -> { return x + y; }. The
compiler will use type inference here to try and guess the types. It does this based on the context
that youre trying to use the lambda in.
Ig your code block returns something or is a single line expression, you can drop the braces and
return statement, for example (x, y) -> x + y;.
In the case of only a single argument, you can drop the parentheses x -> x * 2
If you have no arguments at all, the hamburger symbol is needed, () -> System.out.println(Hey
there!);.
In the interest of completeness, there is another variation; a kind of shortcut to a lambda called a
method reference. So this last one (System.out::println;) is actually a short cut to a lambda.
Were going to talk about those in more detail later, so for now, just be aware that they exist and
can be used anywhere you can use a lambda.
In-depth
In this section, well take a look at things in a little more detail and talk about some related topics,
things like
functional interfaces
method and constructor references
scope and effectively final variables
exception transparency and
as weve talked about howlambdas arent just syntactic sugar, well have a look at the bytecode
lambdas produce
the differences between lambdas and closures
19
Functional Interfaces
Java 8 treats lambdas as an instance of an interface type. It formalises this into something it calls
functional interfaces. A functional interface is just an interface with a single method. Java calls
the method a functional method but the name single abstract method or SAM is often used. All
the existing single method interfaces like Runnable and Callable in the JDK are now functional
interfaces and lambdas can be used anywhere a single abstract method interface is used. In fact, its
functional interfaces that allow for whats called target typing; they provide enough information
for the compiler to infer argument and return types.
@FunctionalInterface
Oracle have introduced a new annotation @FunctionalInterface to mark an interface as such. Its
basically to communicate intent but also allows the compiler to do some additional checks.
For example, this interface compiles,
public interface FunctionalInterfaceExample {
// compiles ok
}
but when you indicate that it should be a functional interface with the the new annotation,
@FunctionalInterface // <- error here
public interface FunctionalInterfaceExample {
}
the compiler will raise an error. It tells us Example is not a functional interface as no abstract
method was found. Often the IDE will also hint, IntelliJ will say something like no target method
was found. Its hinting that we left off the functional method. A single abstract method needs a
single, abstract method!
So what if we try and add a second method to the interface?
@FunctionalInterface
public interface FunctionalInterfaceExample {
void apply();
void illegal(); // <- error here
}
The compiler will error again, this time with a message along the lines of multiple, non-overriding
abstract methods were found. Functional interfaces can have only one method.
http://www.jetbrains.com/idea/
20
Extension
What about the case of an interfaces that extends another interfaces?
Lets create a new functional interface called A and another called B which extends A. B is
still functional. It inherits the parents apply method as youd expect.
@FunctionalInterface
interface A {
abstract void apply();
}
interface B extends A {
}
If you wanted to make this clearer, you can also override the functional method from the parent.
@FunctionalInterface
interface A {
abstract void apply();
}
interface B extends A {
@Override
abstract void apply();
}
We can verify it works as a functional interface if we use it as a lambda. So Ill implement a little
method here to showthat a lambda can be assigned to a type of A and a type of B. The implementation
just prints out A or B.
@FunctionalInterface
public interface A {
void apply();
}
public interface B extends A {
@Override
void apply();
}
public static void main(String... args) {
21
A a = () -> System.out.println("A");
B b = () -> System.out.println("B");
}
You cant add a new abstract method to the extending interface though, as the resulting type would
have two abstract methods and so the IDE will warn us and the compiler will error.
@FunctionalInterface
public interface A {
void apply();
}
public interface B extends A {
void illegal(); // <- can't do this
}
public static void main(String... args) {
A a = () -> System.out.println("A");
B b = () -> System.out.println("B"); // <- error
}
In both cases, you can override methods from Object without causing problems. You can also add
default methods (also new to Java 8). As youd probably expect, it doesnt make sense to try and
mark an abstract class as a functional interface.
Other Interface Improvements
Interfaces generally have had some new features added. Well look at those in detail in Part 2, but
just so that youre aware they include.
Default methods (virtual extension methods)
Static interface methods
and a bunch of new functional interfaces in the java.util.function package; things like
Function and Predicate
Summary
To recap then, in this section weve talked about how any interface with a single method is now a
functional interface.
Weve talked about how that single method is often called a functional method or SAM (for single
abstract method).
22
We looked at the new annotation and saw a couple of examples of how existing JDK interfaces like
Runnable and Callable have been retrofitted with the annotation.
We also introduced the idea of target typing which is how the compiler can use the signature of a
functional method to help work out what lambdas can be used where. We skimmed over this a little
as were going to talk about it later in the type inference section.
In the examples, we saw some examples of functional interfaces, how the compiler and IDE can help
us out when we make mistakes and got a feel for the kinds of errors we might encounter. Things
like adding more than one method to a functional interface. We also saw the exceptions to this rule,
namely when we override methods from Object or implement default methods.
We had a quick look at interface inheritance and how that affects things and I mentioned some of
the other interface improvements that well be covering later.
An important point to take away was the idea that any place a functional interface is used, you can
now use lambdas. Lambdas can be used in-lieu of anonymous implementations of the functional
interface. Using a lambda instead of the anonymous class may seem like syntactic sugar, but theyre
actually quiet different. See the Functions vs. classes section for more details.
23
Type Inference Improvements
There have been several type inference improvements in Java 8. Firstly to be able to support
lambdas, the way the compiler infers things has been improved to use target typing extensively.
Secondly, specific improvements were made over Java 7. These were managed under the Open JDK
Enhancement Proposal 101.
Before we get into those, lets recap on the basics.
Type inference refers to the ability for a programming language to automatically deduce the type of
an expression.
Statically typed languages know the types of things only at compile time. Dynamically typed
languages know the types only at runtime. A statically typed language can use type inference and
drop type information in source code and use the compiler to figure out whats missing.
So this means that type inference can be used by statically typed languages, like Scala, to look like
dynamic languages (like JavaScript). At least at the source code level.
Heres an example of a line of code in Scala.
val name = Henry
It just doesnt need to tell the compiler explicitly that the value is a string. It figures it out. You could
write it out explicitly like this,
24
val name : String = Henry
but theres no need.
As an aside, Scala can usually figure out when youre finished, so you dont often have to add a
terminating semi-colon.
Java Type Inference
Type inference is a fairly broad topic, Java doesnt support the type of inference Ive just been talking
about, at least for things like dropping the type annotations for variables. We have to keep that in.
String name = "Henry"; // <- we can't drop the String like Scala
So Java doesnt support type inference in the widest sense. It cant guess everything like some
languages. Type inference for Java then typically refers to the way the compiler can work out types
for generics. Java 7 improved this when it introduced the diamond operator (<>) but there are still
lots of limitations in what Java can figure out.
The Java compiler was built with type erasure; it actively removes the type information during
compilation. For historical reasons, when generics were introduced in Java 5, the developers couldnt
easily reverse the decision to use erasure. Java was left with the need to understand what types to
substitute for a given generic type but no information how to do it because it had all been erased.
Type inference was the solution.
Because of type erasure then, List<String> becomes List<Object> after compilation. The type
information was erased.
All generic values are really of type Object behind the scenes but by using type inference, the
compiler can check that all the usages are consistent with what it thinks the generic should be. At
runtime, everything is going to get passed around as instances of Object with appropriate casting.
Type inference just allows the compiler to check that the casts would be valid ahead of time.
So type inference is about guessing the types, Javas support for type inferences was due to be
improved in a couple of ways with Java 8.
1. Target-typing for lambdas
and using generalised target-typing to
1. Add support for parameter type inference in method calls
2. Add support for parameter type inference in chained calls
Lets have a look at the current problems and how Java 8 addresses them.
25
Target-typing for lambdas
The general improvements to type inference in Java 8 mean that lambdas can infer their type
parameters; so rather than use
(Integer x, Integer y) -> x + y;
you can drop the Integer type annotation and use the following instead.
(x, y) -> x + y;
This is because the functional interface describes the types, it gives the compiler all the information
it needs.
For example, if we take an example functional interface.
@FunctionalInterface
interface Calculation {
Integer apply(Integer x, Integer y);
}
When a lambda is used in-lieu of the interface, the first thing the compiler does is work out the
target type of the lambda. So if we create a method calculate that takes the interface and two
integers.
static Integer calculate(Calculation operation, Integer x, Integer y) {
return operation.apply(x, y);
}
and then create two lambdas; an addition and subtraction lambda
Calculation addition = (x, y) -> x + y;
Calculation subtraction = (x, y) -> x - y;
and use them like this
calculate(addition, 2, 2);
calculate(subtraction, 5, calculate(addition, 3, 2));
The compiler understands that the lambdas addition and subtraction have a target type of
Calculation (its the only shape that will fit the method signature of calculate). It can then
use the method signature to infer the types of the lambdas parameters. Theres only one method on
the interface, so theres no ambiguity, the argument types are obviously Integer.
Were going to look at lots of examples of target typing as we go on so Im not going to say much
more on this. Just be aware that the mechanism Java uses to achieve lots of the lambda goodness
relies on improvements to type inference and this idea of a target type.
26
Type parameters in method calls
The were some situations prior to Java 8 where the compiler couldnt infer types. One of these was
when calling methods with generics type parameters as arguments.
For example, the Collections class has a generified method for producing an empty list. It looks
like this.
public static final <T> List<T> emptyList() { ... }
In Java 7 this compiles
List<String> names = Collections.emptyList(); // compiles in Java 7
as the Java 7 compiler can work out that the generic needed for the emptyList method is of type
String. What it struggles with though is if the result of a generic method is passed as a parameter
to another method call.
So if we had a method to process a list that looks like this.
static void processNames(List<String> names) {
System.out.println("hello " + name);
}
and then call it with the empty list method.
processNames(Collections.emptyList()); // doesn't compile in Java 7
it wont compile because the generic type of the parameter has been erased to Object. It really looks
like this.
processNames(Collections.<Object>emptyList names);
This doesnt match the processList method.
processNames(Collections.<String>emptyList names);
So it wont compile until we give it an extra hint using an explicit type witness.
27
processNames(Collections.<String>emptyList()); // compiles in Java 7
Now the compiler knows enough about what generic type is being passed into the method.
The improvements in Java 8 include better support for this, so generally speaking where you would
have needed a type witness, you no longer do.
Our example of calling the processNames now compiles!
processNames(Collections.emptyList()); // compiles in Java 8
Type parameters in chained method calls
Another common problem with type inference is when methods are chained together. Lets suppose
we have a List class,
static class List<E> {
static <T> List<T> emptyList() {
return new List<T>();
}
List<E> add(E e) {
// add element
return this;
}
}
and we want to chain a call to add an element to the method creating an empty list. Type erasure
rears its head again; the type is erased and so cant be known by the next method in the chain. It
doesnt compile.
List<String> list = List.emptyList().add(":(");
This was due to be fixed in Java 8, but unfortunately it was dropped. So, at least for now, youll still
need to explicitly offer up a type to the compiler; youll still need a type witness.
List<String> list = List.<String>emptyList().add(":(");
28
Method References
I mentioned earlier that method references are kind of like shortcuts to lambdas. Theyre a compact
and convenient way to point to a method and allow that method to be used anywhere a lambda
would be used.
When you create a lambda, you create an anonymous function and supply the method body. When
you use a method reference as a lambda, its actually pointing to a named method that already exists;
it already has a body.
You can think of them as transforming a regular method into a functional interface.
The basic syntax looks like this.
Class::method
or, a more concrete example
String::valueOf
The part preceding the double colon is the target reference and after, the method name. So, in this
case, were targeting the String class and looking for a method called valueOf; were referring to
the static method on String.
public static String valueOf(Object obj) { ... }
The double colon is called the delimiter, when we use it, were not invoking the method, just
referencing it. So remember not to add brackets on the end.
String::valueOf(); // <-- error
You cant invoke method references directly, they can only be used in-lieu of a lambda. So anywhere
a lambda is used, you can use a method reference.
Example
This statement on its own wont compile.
29
public static void main(String... args) {
String::valueOf;
}
Thats because the method reference cant be transformed into a lambda as theres no context for
the compiler to infer what type of lambda to create.
We happen to know that this reference is equivalent to
(x) -> String.valueOf(x)
but the compiler doesnt know that yet. It can tell some things though. It knows, that as a lambda,
the return value should be of type String because all methods called valueOf in String return a
string. But it has no idea what to supply as a argument. We need to give it a little help and give it
some more context.
Well create a functional interface called Conversion that takes an integer and returns a String. This
is going to be the target type of our lambda.
@FunctionalInterface
interface Conversion {
String convert(Integer number);
}
Next, we need to create a scenario where we use this as a lambda. So we create a little method to
take in a functional interface and apply an interger it.
public static String convert(Integer number, Conversion function) {
return function.convert(number);
}
Now, heres the thing. Weve just given the compiler enough information to transform a method
reference into the equivalent lambda.
When we call convert method, we can do so with a lambda.
convert(100, (number) -> String.valueOf(number));
And we can literally replace the lambda with a reference to the valueOf method. The compiler now
knows we need a lambda that returns a String and takes an integer. It now knows that the valueOf
method fits and can substitute the integer argument.
30
convert(100, String::valueOf);
Another way to give the compiler the information it needs is just to assign the reference to a type.
Conversion b = (number) -> String.valueOf(number);
and as a method reference;
Conversion a = String::valueOf;
the shapes fit so it can be assigned.
Interestingly, we can assign the same lambda to any interface that requires the same shape. For
example, if we have another functional interface with the same shape,
Here Example returns a String and takes an Object so it has the same signature shape as valueOf.
interface Example {
String theNameIsUnimportant(Object object);
}
we can still assign the method reference (or lambda) to it.
Example a = String::valueOf;
Method Reference Types
There are four types of method reference;
constructor references
static method references
and two types of instance method references.
The last two are a little confusing. The first is a method reference of a particular object and the
second is a method reference of an arbitrary object but of a particular type. The difference is in how
you want to use the method and if you have the instance ahead of time or not.
Firstly then, lets have a look at constrictor references.
31
Constructor reference
The basic syntax looks like this,
String::new
A target type following by the double colon, followed by the new keyword. Its going to create a
lambda that will call the zero argument constructor of the String class.
Its equivalent to this lambda
() -> new String()
Remember that method references never have the parentheses; theyre not invoking methods, just
referencing one. This example is referring to the constructor of the String class but not instantiating
a String.
Lets have a look at how we might actually use a constructor reference.
If we create a list of objects we might want to populate that list say ten items. So we could create a
loop and add a new object ten times.
public void usage() {
List<Object> list = new ArrayList<>();
for (int i = 0; i < 10; i++) {
list.add(new Object());
}
}
but if we want to be able to reuse that initialising function, we could extract the code to a new
method called initialise and then use a factory to create the object.
public void usage() {
List<Object> list = new ArrayList<>();
initialise(list, ...);
}
private void initialise(List<Object> list, Factory<Object> factory) {
for (int i = 0; i < 10; i++) {
list.add(factory.create());
}
}
The Factory class is just a functional interface with a method called create that returns some object.
We can then add the object it created to the list. Because its a functional interface, we can use a
lambda to implement the factory to initialise the list;
32
public void usage() {
List<Object> list = new ArrayList<>();
initialise(list, () -> new Object());
}
Or we could swap in a constructor reference.
public void usage() {
List<Object> list = new ArrayList<>();
initialise(list, Object::new);
}
Theres a couple of other things we could do here. If we add some generics to the initialise method
we can reuse it when initialising lists of any type. For example, we can go back and change the type
of the list to be String and use a constructor reference to initialise it.
public void usage() {
List<String> list = new ArrayList<>();
initialise(list, String::new);
}
private <T> void initialise(List<T> list, Factory<T> factory) {
for (int i = 0; i < 10; i++) {
list.add(factory.create());
}
}
Weve seen how it works for zero argument constructors, but what about the case when classes have
multiple argument constructors?
When there are multiple constructors, you use the same syntax but the compiler figures out which
constructor would be the best match. It does this based on the target type and inferring functional
interfaces that it can use to create that type.
Lets take the example of a Person class, it looks like this and you can see the constructor takes a
bunch of arguments.
33
class Person {
public Person(String forename, String surname, LocalDate birthday,
Sex gender, String emailAddress, int age) {
// ...
}
}
Going back to our example from earlier and looking at the general purpose initialise method, we
could use a lambda like this
initialise(people, () -> new Person(forename, surname, birthday,
gender, email, age));
but to be able to use a constructor reference, wed need a lambda with variable arguments and that
would look like this
(a, b, c, d, e, f) -> new Person(a, b, c, d, e, f);
but this doesnt translate to a constructor reference directly. If we were to try and use
Person::new
it wont compile as it doesnt know anything about the parameters. If you try and compile it, the
error says youve created an invalid constructor reference that cannot be applied to the given types;
it found no arguments.
Instead, we have to introduce some indirection to give the compiler enough information to find an
appropriate constructor. We can create something that can be used as a functional interface and has
the right types to slot into the appropriate constructor;
Lets create a new functional interface called PersonFactory.
@FunctionalInterface
interface PersonFactory {
Person create(String forename, String surname, LocalDate birthday,
Sex gender, String emailAddress, int age);
}
Here, the arguments from PersonFactory match the available constructor on Person. Magically, this
means we can go back and use it with a constructor reference of Person.
34
public void example() {
List<Person> list = new ArrayList<>();
PersonFactory factory = Person::new;
// ...
}
Notice Im using the constructor reference from Person. The thing to note here is that a constructor
reference can be assigned to a target functional interface even though we dont yet know the
arguments.
It probably seems a strange that the type of the method reference is PersonFactory and not Person.
This extra target type information helps the compiler to know it has to go via PersonFactory to
create a Person. With this extra hint, the compiler is able to create a lambda based on the factory
interface that will later create a Person.
Writing it out long hand, the compiler would generate this.
public void example() {
PersonFactory factory = (a, b, c, d, e, f) -> new Person(a, b, c, d, e, f);
}
which could be used later like this;
public void example() {
PersonFactory factory = (a, b, c, d, e, f) -> new Person(a, b, c, d, e, f);
Person person = factory.create(forename, surname, birthday,
gender, email, age);
}
Fortunately, the compiler can do this for us once weve introduced the indirection.
It understands the target type to use is PersonFactory and it understands that its single abstract
method can be used in lieu of a constructor. Its kind of like a two step process, firstly, to work out
that the abstract method has the same argument list as a constructor and that it returns the right
type, then apply it with colon colon new syntax.
To finish off the example, we need to tweak our initialise method to add the type information (replace
the generics), add parameters to represent the persons details and actually invoke the factory.
35
private void initialise(List<Person> list, PersonFactory factory, String forename,
String surname, LocalDate birthday, Sex gender,
String emailAddress, int age) {
for (int i = 0; i < 10; i++) {
list.add(factory.create(forename, surname, birthday,
gender, emailAddress, age));
}
}
and then we can use it like this
public void example() {
List<Person> list = new ArrayList<>();
PersonFactory factory = Person::new;
initialise(people, factory, a, b, c, d, e, f);
}
or inline, like this
public void example() {
List<Person> list = new ArrayList<>();
initialise(people, Person::new, a, b, c, d, e, f);
}
36
Static method reference
A method reference can point directly to a static method. For example,
String::valueOf
This time, the left hand side refers to the type where a static method, in this case valueOf can be
found. Its equivalent to this lambda
x -> String.valueOf(x))
A more extended example would be where we sort a collection using a reference to a static method
on the class Comparators.
Collections.sort(Arrays.asList(5, 12, 4), Comparators::ascending);
// equivalent to
Collections.sort(Arrays.asList(5, 12, 4), (a, b) -> Comparators.ascending(a, b));
where, the static method ascending might be defined like this.
public static class Comparators {
public static Integer ascending(Integer first, Integer second) {
return first.compareTo(second);
}
}
37
Instance method reference of particular object (in this case, a closure)
Heres an example of an instance method reference of a specific instance.
x::toString
The x is a specific instance that we want to get at. Its lambda equivalent looks like this;
() -> x.toString()
The ability to reference the method of a specific instance also gives us a convenient way to convert
between different functional interface types. For example;
Callable<String> c = () -> "Hello";
Callables functional method is call, when its invoked the lambda will return Hello.
If we have another functional interface, Factory, we can convert the Callable using a method
reference.
Factory<String> f = c::call;
We could have just re-created the lambda but this trick is a useful way to get reuse out of predefined
lambdas. Assign them to variables and reuse them to avoid duplication.
Heres an example of it in use.
public void example() {
String x = "hello";
function(x::toString);
}
This is an example where the method reference is using a closure. It creates a lambda that will call
the toString method on the instance x.
The signature of function looks like this
public static String function(Supplier<String> supplier) {
return supplier.get();
}
The Supplier interface must provide a string value (the get call) and the only way it can do that is
if its been supplied to it on construction. Its equivalent to
38
public void example() {
String x = "";
function(() -> x.toString());
}
Notice here that the lambda has no arguments (it uses the hamburger symbol). This shows that
the value of x isnt available in the lambdas local scope and so can only be available from outside
its scope. Its a closure because must close over x.
If youre interested in seeing the long hand, anonymous class equivalent, itll look like this. Notice
again how x must be passed in.
public void example() {
String x = "";
function(new Supplier<String>() {
@Override
public String get() {
return x.toString(); // <- closes over 'x'
}
});
}
All three of these are equivalent. Compare this to the lambda variation of an instance method
reference where it doesnt have its argument explicitly passed in from an outside scope.
39
Instance method reference of a arbitrary object whos instance is supplied
later (lambda)
The last case is for a method reference that points to an arbitrary object referred to by its type.
Object::toString
So in this case, although it looks like the left hand side is pointing to a class, its actually pointing to
an instance. toString is an instance method on Object, not a static method. The reason why you
might not use the regular instance method syntax is because you may not yet have an instance to
refer to.
So before, when we call x colon colon toString, we know the value of x. There are some situations
where you dont have a value of x and in these cases, you can still pass around a reference to the
method but supply a value later using this syntax.
For example, the lambda equivalent doesnt have a bound value for x.
(x) -> x.toString()
There difference between the two types of instance method reference is basically academic.
Sometimes, youll need to pass something in, other times, the usage of the lambda will supply it
for you.
The example is similar to the regular method reference; it calls the toString method of a string only
this time, the string is supplied to the function thats making use of the lambda and not passed in
from an outside scope.
public void lambdaExample() {
function("value", String::toString);
}
The String part looks like its referring to a class but its actually referencing an instance. Its
confusing, I know but to see things more clearly, we need to see the function thats making use
of the lambda. It looks like this.
public static String function(String value, Function<String, String> function) {
return function.apply(value);
}
So, the string value is passed directly to the function, it would look like this as a fully qualified
lambda.
40
public void lambdaExample() {
function("value", x -> x.toString());
}
which Java can shortcut to look like String::toString; its saying supply the object instance at
runtime.
If you expand it fully to an anonymous interface, it looks like this. The x parameter is made available
and not closed over. Hence it being a lambda rather than a closure.
public void lambdaExample() {
function("value", new Function<String, String>() {
@Override
// takes the argument as a parameter, doesn't need to close over it
public String apply(String x) {
return x.toString();
}
});
}
41
Summary
Oracle describe the four kinds of method reference as follows.
Kind Example
Reference to a static method ContainingClass::staticMethodName
Reference to an instance method of a particular object ContainingObject::instanceMethodName
Reference to an instance method of an arbitrary object of a
particular type
ContainingType::methodName
Reference to a constructor ClassName::new
But the instance method descriptions are just plain confusing. What on earth is an instance method
of an arbitrary object of a particular type? Arent all objects of a particular type? Why is it important
that the object is arbitrary?
I prefer to think of the first as an instance method of a specific object known ahead of time and the
second as an instance method of an arbitrary object that will be supplied later. Interestingly, this
means the first is a closure and the second is a lambda. One is bound and the other unbound. The
distinction between a method reference that closes over something (a closure) and one that doesnt
(a lambda) may be a bit academic but at least its a more formal definition than Oracles unhelpful
description.
Kind Syntax Example
Reference to a static method Class::staticMethodName String::valueOf
Reference to an instance method of a specific
object
object::instanceMethodName x::toString
Reference to an instance method of a arbitrary
object supplied later
Class::instanceMethodName String::toString
Reference to a constructor ClassName::new String::new
or as equivalent lambdas
Kind Syntax As Lambda
Reference to a static method Class::staticMethodName (s) -> String.valueOf(s)
Reference to an instance method of a
specific object
object::instanceMethodName () -> "hello".toString()
Reference to an instance method of a
arbitrary object supplied later
Class::instanceMethodName (s) -> s.toString()
Reference to a constructor ClassName::new () -> new String()
Note that the syntax for a static method reference looks very similar to a reference to an instance
method of a class. The compiler determines which to use by going through each applicable static
method and each applicable instance method. If it were to find a match for both, the result would
http://docs.oracle.com/javase/tutorial/java/javaOO/methodreferences.html
42
be a compiler error.
You can think of the whole thing as a transformation from a method reference to a lambda. The
compiler provides the transformation function that takes a method reference and target typing and
can derive a lambda.
43
Scoping
The good news with lambdas is that they dont introduce any new scoping. Using variables within
a lambda will refer to variables residing in the enclosing environment.
This is whats called lexical scoping. It means that lambdas dont introduce a new level of scoping at
all; you can directly access fields, methods and variables from the enclosing scope. Its also the case
for the this and super keywords. So we dont have to worry about the crazy nested class syntax
for resolving scope.
Lets take a look at an example. We have an example class here, with a member variable i set to the
value of 5.
public static class Example {
int i = 5;
public Integer example() {
Supplier<Integer> function = () -> i * 2;
return function.get();
}
}
In the example method, a lambda uses a variable called i and multiplies it by two.
Because lambdas are lexically scoped, i simply refers to the enclosing classes variable. Its value at
run-time will be 5. Using this drives home the point; this within a lambda is the same as without.
public static class Example {
int i = 5;
public Integer example() {
Supplier<Integer> function = () -> this.i * 2;
return function.get();
}
}
In the anotherExample method below, a method parameter is used which is also called i. The usual
shadowing rules kick in here and i will refer to the method parameter and not the class member
variable. The method variable shadows the class variable. Its value will be whatever is passed into
the method.
44
public static class Example {
int i = 5;
public Integer anotherExample(int i) {
Supplier<Integer> function = () -> i * 2;
return function.get();
}
}
If you wanted to refer to the class variable i and not the parameter i from within the body, you
could make the variable explicit with this. For example, Supplier<Integer> function = () -> i
* 2;.
The following example has a locally scoped variable defined within the yetAnotherExample method.
Remember that lambdas use their enclosing scope as their own, so in this case, i within the lambda
refers to the methods variable; i will be 15 and not 5.
public static class Example {
int i = 5;
public Integer yetAnotherExample() {
int i = 15;
Supplier<Integer> function = () -> i * 2;
return function.get();
}
}
If you want to see this for yourself, you could use a method like the following to print out the values.
public static void main(String... args) {
Example scoping = new Example();
System.out.println("class scope = " + scoping.example());
System.out.println("method param scope = " + scoping.anotherExample(10));
System.out.println("method scope = " + scoping.yetAnotherExample());
}
The output would look like this.
45
class scope = 10
method param scope = 20
method scope = 30
So, the first method prints 10; 5 from the class variable multiplied by two. The second method prints
20 as the parameter value was 10 and was multiplied by two and the final method prints 30 as the
local method variable was set to 15 and again multiplied by two.
Lexical scoping means deferring to the enclosing environment. Each example had a different
enclosing environment or scope. You saw a variable defined as a class member, a method parameter
and locally from within a method. In all cases, the lambda behaved consistently and referenced the
variable from these enclosing scopes.
Lambda scoping should be intuitive if youre already familiar with basic Java scoping, theres really
nothing new here.
46
Effectively Final
In Java 7, any variable passed into an anonymous class instance would need to be made final.
This is because the compiler actually copies all the context or environment it needs into the instance
of the anonymous class. If those values were to change under it, unexpected side affects could
happen. So Java insists that the variable be final to ensure it doesnt change and the inner class can
operate on them safely. By safely, I mean without race conditions or visibility problems between
threads.
Lets have a look at an example. To start with well use Java 7 and create a method called filter
that takes a list of people and a predicate. Well create a temporary list to contain any matches we
find then enumerate each element testing to see if the predicate holds true for each person. If the
test is positive, well add them to the temporary list before returning all matches.
// java 7
private List<Person> filter(List<Person> people, Predicate<Person> predicate) {
ArrayList<Person> matches = new ArrayList<>();
for (Person person : people)
if (predicate.test(person))
matches.add(person);
return matches;
}
Then well create a method that uses this to find all the people in a list that are eligible for retirement.
We set a retirement age variable and then call the filter method with an arbitrary list of people and
a new anonymous instance of a Predicate interface.
Well implement this to return true if a persons age is greater than or equal to the retirement age
variable.
public void findRetirees() {
int retirementAge = 55;
List<Person> retirees = filter(allPeople, new Predicate<Person>() {
@Override
public boolean test(Person person) {
return person.getAge() >= retirementAge; // <-- compilation error
}
});
}
If you try and compile this, youll get a compiler failure when accessing the variable. This is because
the variable isnt final. Wed need to add final to make it compile.
47
final int retirementAge = 55;
..
Passing the environment into an anonymous inner class like this is an example of a closure. The
environment is what a closure closes over; it has to capture the variables it needs to do its job.
The Java compiler achieves this using the copy trick rather than try and manage multiple changes
to the same variable. In the context of closures, this is called variable capture.
Java 8 introduces the idea of effectively final which means that if the compiler can work out that a
particular variable is never changed, it can be used where ever a final variable would have be used.
It interprets it as effectively final.
In our example, if we switch to Java 8 and drop the final keyword. Things still compile. No need to
make the variable final. Java knows that the variable doesnt change so it makes it effectively final.
int retirementAge = 55;
Of course, it still compiles if you were to make it final.
But how about if we try and modify the variable after weve initialised it?
int retirementAge = 55;
// ...
retirementAge = 65;
The compiler spots the change and can no longer treat the variable as effectively final. We get the
original compilation error asking us to make it final. Conversely, if adding the final keyword to a
variable declaration doesnt cause a compiler error, then the variable is effectively final.
Ive been demonstrating the point here with an anonymous class examples because the idea of
effectively final isnt something specific to lambdas. It is of course applicable to lambdas though.
You can convert this anonymous class above into a lambda and nothing changes. Theres still no
need to make the variable final.
Circumventing Final
You can still get round the safety net by passing in final objects or arrays and then change their
internals in your lambda.
For example, taking our list of people, lets say we want to sum all their ages. We could create a
method to loop and sum like this;
48
private static int sumAllAges(List<Person> people) {
int sum = 0;
for (Person person : people) {
sum += person.getAge();
}
return sum;
}
where the sum count is maintained as the list is enumerated. As an alternative, we could try and
abstract the looping behaviour and pass in a function to be applied to each element. Like this.
public final static Integer forEach(List
and to achieve the summing behaviour, all wed need to do is create a function that can sum. You
could do this using an anonymous class like this;
private static void badExample() {
Function<Integer, Integer> sum = new Function<Integer, Integer>() {
private Integer sum = 0;
@Override
public Integer apply(Integer amount) {
sum += amount;
return sum;
}
};
}
Where the functions method takes an integer and returns an integer. In the implementation, the sum
variable is a class member variable and is mutated each time the function is applied. This kind of
mutation is generally bad form when it comes to functional programming.
Nether the less, we can pass this into our forEach method like this;
forEach(allPeople, sum);
and wed get the sum of all peoples ages. This works because were using the same instance of the
function so the sum variable is reused and mutated during each iteration.
The bad news is that we cant convert this into a lambda directly; theres no equivalent to a member
variable with lambdas, so theres nowhere to put the sum variable other than outside of the lambda.
49
double sum = 0;
forEach(allPeople, x -> {
return sum += x;
});
but this highlights that the variable isnt effectively final (its changed in the lambdas body) and so
it must be made final.
But if we make it final
final double sum = 0;
forEach(allPeople, x -> {
return sum += x;
});
we can no longer modify it in the body! Its a chicken and egg situation.
The trick around this is to use a object or an array; its reference can remain final but its internals
can be modified
int[] sum = {0};
forEach(allPeople, x -> sum[0] += x);
The array reference is indeed final here, but we can modify the array contents without reassigning
the reference. However, this is generally bad form as it opens up to all the safety issues we talked
about earlier. I wanted to mention it for illustration purposes but I dont recommend you do this
kind of thing often. Its generally better not to create functions with side affects and you can avoid
the issues completely if you use a more functional approach.
The idiomatic way to do this kind of summing is to use whats called a fold or in the Java vernacular
reduce. Well be looking at this in more detail when we look at streams and the java.util.function
Package.
50
Exception Handling
Theres no new syntax for exception handling in lambdas. Exceptions thrown in a lambda are
propagated to the caller, just as youd expect with a regular method call. Theres nothing special
about calling lambdas or handling their exceptions.
However, there are some subtleties that you need to be aware of. Firstly, as a caller of a lambda, you
are potentially unaware of what exceptions might be thrown, if any and secondly, as the author of
a lambda, youre potentially unaware what context your lambda will be run in.
When you create a lambda, you typically give up responsibility of how that lambda will be executed
to the method that you pass it to. For all you know, your lambda may be run in parallel or at some
point in the future and so any exception you throw may not get handled as you might expect. You
cant rely on exception handling as a way to control your programs flow.
To demonstrate this, lets write some code to call two things, one after the other. Well use Runnable
as a convenient lambda type.
public static void runInSequence(Runnable first, Runnable second) {
first.run();
second.run();
}
If the first call to run were to throw an exception, the method would terminate and the second
method would never be called. The caller is left to deal the exception. If we use this method to
transfer money between two bank accounts, we might write two lambdas. One for the debit action
and one for the credit.
public void transfer(BankAccount a, BankAccount b, Integer amount) {
Runnable debit = () -> a.debit(amount);
Runnable credit = () -> b.credit(amount);
}
we could then call our runInSequence method like this
public void transfer(BankAccount a, BankAccount b, Integer amount) {
Runnable debit = () -> a.debit(amount);
Runnable credit = () -> b.credit(amount);
runInSequence(debit, credit);
}
any exceptions could be caught and dealt with by using a try/catch like this.
51
public void transfer(BankAccount a, BankAccount b, Integer amount) {
Runnable debit = () -> a.debit(amount);
Runnable credit = () -> b.credit(amount);
try {
runInSequence(debit, credit);
} catch (Exception e) {
// check account balances and rollback
}
}
Heres the thing. As an author of the lambdas, I potentially have no idea how runInSequence is
implemented. It may well be implemented to run asynchronously like this.
public static void runInSequence(Runnable first, Runnable second) {
new Thread(() -> {
first.run();
second.run();
}).start();
}
In which case would mean that any exception in the first call would terminate the thread, the
exception would disappear to the default exception handler and our original client code wouldnt
get the chance to deal with the exception.
Using a Callback
Incidentally, one way round the specific problem with raising an exception on a different thread
than the caller can be addressed with a callback function. Firstly, youd defend against exceptions
in the runInSequence method.
public static void runInSequence(Runnable first, Runnable second) {
new Thread(() -> {
try {
first.run();
second.run();
} catch (Exception e) {
// ...
}
}).start();
}
Then introduce an exception handler which can be called in the event of an exception
52
public static void runInSequence(Runnable first, Runnable second,
Consumer<Throwable> exceptionHandler) {
new Thread(() -> {
try {
first.run();
second.run();
} catch (Exception e) {
exceptionHandler.accept(e);
}
}).start();
}
Consumer is a functional interface (new in Java 8) that in this case takes the exception as an argument
to its accept method.
When we wire this up to the client, we can pass in a callback lambda to handle any exception.
public void nonBlockingTransfer(BankAccount a, BankAccount b, Integer amount) {
Runnable debit = () -> a.debit(amount);
Runnable credit = () -> b.credit(amount);
runInSequence(debit, credit, (exception) -> {
/* check account balances and rollback */
});
}
This is a good example of deferred execution and so has its own foibles. The exception handler
method may (or may not) get executed at some later point in time. The nonBlockingTransfer
method will have finished and the bank accounts themselves may be in some other state by the
time it fires. You cant rely on the exception handler being called when its convenient for you;
weve opened up a whole can of concurrency worms.
Dealing with Exceptions When Writing Lambdas
Lets look at dealing with exceptions from the perspective of a lambda author, someone writing
lambdas. After this, well look at dealing with exceptions when calling lambdas.
Lets look at it as if we wanted to implement the transfer method using lambdas but this time wanted
to reuse an existing library that supplies the runInSequence method.
Before we start, lets take a look at the BankAccount class. Youll notice that this time, the debit and
credit methods both throw a checked exception; InsufficientFundsException.
53
class BankAccount {
public void debit(int amount) throws InsufficientFundsException {
// ...
}
public void credit(int amount) throws InsufficientFundsException {
// ...
}
}
class InsufficientFundsException extends Exception { }
Lets recreate the transfer method. Well try to create the debit and credit lambdas and pass these
into the runInSequence method. Remember that the runInSequence method was written by some
library author and we cant see or modify its implementation.
public void transfer(BankAccount a, BankAccount b, Integer amount) {
Runnable debit = () -> a.debit(amount); <- compiler error
Runnable credit = () -> b.credit(amount); <- compiler error
runInSequence(debit, credit);
}
The debit and credit both throw a checked exception, so this time, you can see a compiler error. It
makes no difference if we add this to the method signature; the exception would happen inside the
lambda. Remember I said exceptions in lambdas are propagated to the caller? In our case, this will be
the runInSequence method and not the point we define the lambda. The two arent communicating
between themselves that there could be an exception raised.
// still doesn't compile
public void transfer(BankAccount a, BankAccount b, Integer amount)
throws InsufficientFundsException {
Runnable debit = () -> a.debit(amount);
Runnable credit = () -> b.credit(amount);
runInSequence(debit, credit);
}
So if we cant force a checked exception to be transparent between the lambda and the caller, one
option is to wrap the checked exception as a runtime exception like this.
54
public void transfer(BankAccount a, BankAccount b, Integer amount) {
Runnable debit = () -> {
try {
a.debit(amount);
} catch (InsufficientFundsException e) {
throw new RuntimeException(e);
}
};
Runnable credit = () -> {
try {
b.credit(amount);
} catch (InsufficientFundsException e) {
throw new RuntimeException(e);
}
};
runInSequence(debit, credit);
}
That gets us out of the compilation error but its not the full story yet. Its very verbose and we still
have to catch and deal with, whats now a runtime exception, around the call to runInSequence.
public void transfer(BankAccount a, BankAccount b, Integer amount) {
Runnable debit = () -> { ... };
};
Runnable credit = () -> { ... };
try {
runInSequence(debit, credit);
} catch (RuntimeException e) {
// check balances and rollback
}
}
Theres still one or two niggles though; were throwing and catching a RuntimeException which
is perhaps a little loose. We dont really know what other exceptions, if any, might be thrown in
the runInSequence method. Perhaps its better to be more explicit. Lets create a new sub-type of
RuntimeException and use that instead.
55
class InsufficientFundsRuntimeException extends RuntimeException {
public InsufficientFundsRuntimeException(InsufficientFundsException cause) {
super(cause);
}
}
Ill just have to go back and throw the new exception from within the lambdas.
When we go back to modify the original code, we can restrict the final catch to deal with only
exceptions we knowabout; namely the InsufficientFundsRuntimeException. I can nowimplement
some kind of balance check and rollback functionality, confident that I know all the scenarios that
cause it.
public void transfer(BankAccount a, BankAccount b, Integer amount) {
Runnable debit = () -> {
try {
a.debit(amount);
} catch (InsufficientFundsException e) {
throw new InsufficientFundsRuntimeException(e);
}
};
Runnable credit = () -> {
try {
b.credit(amount);
} catch (InsufficientFundsException e) {
throw new InsufficientFundsRuntimeException(e);
}
};
try {
runInSequence(debit, credit);
} catch (InsufficientFundsRuntimeException e) {
// check balances and rollback
}
}
The trouble with all this, is that the code has more exception handling boilerplate than actual
business logic. Lambdas are supposed to make things less verbose but this is just full of noise. We can
make things better if we generalise the wrapping of checked exceptions to runtime equivalents. We
could create a functional interface that captures an exception type on the signature using generics.
Lets name it Callable and its single method; call. Dont confuse this with the class of the same
name in the JDK; were creating a new class to illustrate dealing with exceptions.
56
@FunctionalInterface
interface Callable<E extends Exception> {
void call() throws E;
}
Well change the old implementation of transfer and create lambdas to match the shape of the
new functional interface. Ill leave off the type for a moment.
public void transfer(BankAccount a, BankAccount b, Integer amount) {
??? debit = () -> a.debit(amount);
??? credit = () -> b.credit(amount);
}
Remember from the type inference lecture that Java would be able to see this as a lambda of type
Callable as it has no parameters and as does Callable, it has the same return type (none) and
throws an exception of the same type as the interface. We just need to give the compiler a hint, so
we can assign this to an instance of a Callable.
public void transfer(BankAccount a, BankAccount b, Integer amount) {
Callable<InsufficientFundsException> debit = () -> a.debit(amount);
Callable<InsufficientFundsException> credit = () -> b.credit(amount);
}
Creating the lambdas like this doesnt cause a compilation error as the functional interface declares
that it could be thrown. It doesnt need to warn us at the point we create the lambda, as the signature
of the functional method will cause the compiler to error if required when we actually try and call
it. Just like a regular method.
If we try and pass them into the runInSequence method, well get a compiler error though.
public void transfer(BankAccount a, BankAccount b, Integer amount) {
Callable<InsufficientFundsException> debit = () -> a.debit(amount);
Callable<InsufficientFundsException> credit = () -> b.credit(amount);
runInSequence(debit, credit); <- doesn't compile
}
The lambdas are of the wrong type. We still need a lambda of type Runnable. Well have to write a
method that can convert from a Callable to a Runnable. At the same time, well wrap the checked
exception to a runtime one. Something like this.
57
public static Runnable unchecked(Callable<InsufficientFundsException> function) {
return () -> {
try {
function.call();
} catch (InsufficientFundsException e) {
throw new InsufficientFundsRuntimeException(e);
}
};
}
All thats left to do is wire it in for our lambdas.
public void transfer(BankAccount a, BankAccount b, Integer amount) {
Runnable debit = unchecked(() -> a.debit(amount));
Runnable credit = unchecked(() -> b.credit(amount));
runInSequence(debit, credit);
}
Once we put the exception handling back in were back to a more concise method body and have
dealt with the potential exceptions in the same way as before.
public void transfer(BankAccount a, BankAccount b, Integer amount) {
Runnable debit = unchecked(() -> a.debit(amount));
Runnable credit = unchecked(() -> b.credit(amount));
try {
runInSequence(debit, credit);
} catch (InsufficientFundsRuntimeException e) {
// check balances and rollback
}
}
The downside is this isnt a totally generalised solution; wed still have to create variations of the
unchecked method for different functions. Weve also just hidden the verbose syntax away. The
verbosity is still there its just been moved. Yes, weve got some reuse out of it but if exception
handling were transparent or we didnt have checked exceptions, we wouldnt need to brush the
issue under the carpet quiet so much.
Its worth pointing out that wed probably end up doing something similar if we were in Java 7 and
using anonymous classes instead of lambdas. A lot of this stuff can still be done pre-Java 8 and youll
end up creating helper methods to push the verbosity to one side.
Its certainly the case that lambdas offer more concise representations for small anonymous pieces
of functionality but because of Javas checked exception model, dealing with exceptions in lambdas
will often cause all the same verbosity problems we had before.
58
As a Caller (Dealing with Exceptions when Calling Lambdas)
Weve seen things from the perspective of writing lambdas, now lets have a look at things when
calling lambdas.
Lets imagine that now were writing the library that offers the runInSequence method. We have
more control this time and arent limited to using Runnable as a lambda type. Because we dont want
to force our clients to jump through hoops dealing with exceptions in their lambdas (or wrap them
as runtime exceptions), well provide a functional interface that declares that a checked exception
might be thrown.
Well call it FinancialTransfer with a transfer method.
@FunctionalInterface
interface FinancialTransfer {
void transfer() throws InsufficientFundsException;
}
Were saying that whenever a banking transaction occurs, theres the possibility that insufficient
funds are available. Then when we implement our runInSequence method, we accept lambdas of
this type.
public static void runInSequence(FinancialTransfer first,
FinancialTransfer second) throws InsufficientFundsException {
first.transfer();
second.transfer();
}
This means that when clients use the method, theyre not forced to deal with exceptions within their
lambdas. For example, writing a method like this.
// example client usage
public void transfer(BankAccount a, BankAccount b, Integer amount) {
FinancialTransfer debit = () -> a.debit(amount);
FinancialTransfer credit = () -> b.credit(amount);
}
This time there is no compiler error when creating the lambdas. Theres no need to wrap the
exceptions from BankAccount methods as runtime exceptions; the functional interface has already
declared the exception. However, runInSequence would now throw a checked exception, so its
explicit that the client has to deal with the possibility and youll see a compiler error.
59
public void transfer(BankAccount a, BankAccount b, Integer amount) {
FinancialTransfer debit = () -> a.debit(amount);
FinancialTransfer credit = () -> b.credit(amount);
runInSequence(debit, credit); <- compiler error
}
So we need to wrap the call in a try/catch to make the compiler happy.
public void transfer(BankAccount a, BankAccount b, Integer amount) {
FinancialTransfer debit = () -> a.debit(amount);
FinancialTransfer credit = () -> b.credit(amount);
try {
runInSequence(debit, credit); <- compiler error
} catch (InsufficientFundsException e) {
// whatever
}
}
The end result is something like we saw previously but without the need for the unchecked method.
As a library developer, weve made it easier for clients to integrate with our code.
But what about if we try something more exotic? Lets make the runInSequence method asyn-
chronous again. Theres no need to throw the exception from the method signature as it wouldnt
propogate to the caller if it were thrown from a different thread. So this version of the runInSequence
method doesnt include the throws clause and the transfer method is no longer forced to deal with
it. However, the calls to .transfer will still throw an exception.
public static void runInSequence(Runnable first, Runnable second) {
new Thread(() -> {
first.transfer(); <- compiler error
second.transfer(); <- compiler error
}).start();
}
public void transfer(BankAccount a, BankAccount b, Integer amount) {
FinancialTransfer debit = () -> a.debit(amount);
FinancialTransfer credit = () -> b.credit(amount);
runInSequence(debit, credit); <- compiler error
}
With the compiler errors still in the runInSequence method, we need another way to handle the
exception. One technique is to pass in a function that will be called in the event of an exception. We
can use this lambda to bridge the code running asynchronously back to the caller.
60
To start with, well add the catch block back in and pass in a functional interface to use as
the exception handler. Ill use the Consumer interface here, its new in Java 8 and part of the
java.util.function package. We then call the interface method in the catch block, passing in the
cause.
public void runInSequence(FinancialTransfer first, FinancialTransfer second,
Consumer<InsufficientFundsException> exceptionHandler) {
new Thread(() -> {
try {
first.transfer();
second.transfer();
} catch (InsufficientFundsException e) {
exceptionHandler.accept(e);
}
}).start();
}
To call it, we need to update the transfer method to pass in a lambda for the callback. The parameter,
exception below, will be whatever is passed into the accept method in runInSequence. It will be
an instance of InsufficientFundsException and the client can deal with it however they chose.
public void transfer(BankAccount a, BankAccount b, Integer amount) {
FinancialTransfer debit = () -> a.debit(amount);
FinancialTransfer credit = () -> b.credit(amount);
Consumer<InsufficientFundsException> handler = (exception) -> {
/* check account balances and rollback */
};
runInSequenceNonBlocking(debit, credit, handler);
}
There we are. Weve provided the client to our library with an alternative exception handling
mechanism rather than forcing them to catch exceptions.
Weve internalised the exception handling into the library code. Its a good example of deferred
execution; should there be an exception, the client doesnt necessarily know when his exception
handler would get invoked. For example, as were running in another thread, the bank accounts
themselves may have be altered by the time it executes. Again it highlights that using exceptions
to control your programs flow is a flawed approach. You cant rely on the exception handler being
called when its convenient for you.
61
Lambdas vs Closures
The terms closure and lambda are often used interchangeably but they are actually distinct. In this
section well take a look at the differences so you can be clear about which is which.
Below is a table showing the release dates for each major version of Java. Java 5.0 came along in
2004 and included the first major language changes including things like generics support.
Around 2008 to 2010 there was a lot of work going on to introduce closures to Java. It was due to
go in to Java 7 but didnt quite make it in time. Instead it evolved into lambda support in Java 8.
Unfortunately, around that time, people used the term closures and lambdas interchangeably
and so its been a little confusing for the Java community since. In fact, theres still a project page
on the OpenJDK site for closures and one for lambdas.
From the OpenJDK projects perspective, they really should have been using lambda consistently
from the start. In fact, the OpenJDK got it so wrong, they ignored the fact that Java has had closure
support since 1.1!
Im being slightly pedantic here as although there are technical differences between closures and
lambdas, the goals of the two projects were to achieve the same thing, even if they used the
terminology inconsistently.
http://openjdk.java.net/projects/closures/
http://openjdk.java.net/projects/lambda/
62
So what is the difference between lambdas and closures? Basically, a closure is a type of lambda but
a lambda isnt necessarily a closure.
Basic Differences
Just like a lambda, a closure is effectively an anonymous block of functionality, but there are some
important distinctions. A closure depends on external values (not just its arguments) whereas a
lambda depends only on its arguments. A closure is said to close over the environment it requires.
For example, the following
(server) -> server.isRunning();
is a lambda, but this
() -> server.isRunning();
is a closure.
They both return a boolean indicating if some server is up but one uses its argument and the other
must get the variable from somewhere else. Both are lambdas; in the general sense, they are both
anonymous blocks of functionality and in the Java language sense, they both use the new lambda
syntax.
The first example refers to a server variable passed into the lambda as an argument whereas the
second example (the closure) gets the server variable from somewhere else; ie the environment. To
get the instance of the variable, the lambda has to close over the environment or capture the value
of server. Weve seen this in action when we talked about effectively final before.
Lets expand the example to see things more clearly. Firstly, well create a method in a static class to
perform a naive poll and wait. Itll check a functional interface on each poll to see if some condition
has been met.
class WaitFor {
static <T> void waitFor(T input, Predicate<T> predicate)
throws InterruptedException {
while (!predicate.test(input))
Thread.sleep(250);
}
}
We use Predicate as our functional interface and test it, pausing for a short while if the condition
is not satisfied. We can call this method with a simple lambda that checks if some HTTP server is
running.
63
void lambda() throws InterruptedException {
waitFor(new HttpServer(), (server) -> !server.isRunning());
}
The server parameter is supplied by our waitFor method and will be the instance of HttpServer
weve just defined. Its a lambda as the compiler doesnt need to capture the server variable as we
supply it manually at runtime.
..
Incidentally, we might have been able to use a method reference
waitFor(new HttpServer(), HttpServer::isRunning); // nowhere do you put the !
but we cant as we cant negate method references. You get a compile error.
Now, if we re-implement this as a closure, it would look like this. Firstly, we have to add another
waitFor method.
static void waitFor(Condition condition) throws InterruptedException {
while (!condition.isSatisfied())
Thread.sleep(250);
}
This time, with a simpler signature. We pass in a functional interface that requires no parameters.
The Condition interface has a simple isSatisfied method with no argument which implies well
have to supply any values an implementation might need. Its already hinting that usages of it may
result in closures.
Using it, wed write something like this.
void closure() throws InterruptedException {
Server server = new HttpServer();
waitFor(() -> !server.isRunning());
}
The server instance is not passed as a parameter to the lambda here but accessed from the enclosing
scope. Weve defined the variable and the lambda uses it directly. This variable has to be captured,
or copied by the compiler. The lambda closes over the server variable.
This expression to close over comes from the idea that a lambda expression with open bindings
(or free variables) have been closed by (or bound in) the lexical environment or scope. The result is
64
a closed expression. There are no unbound variables. To be more precise, closures close over values
not variables.
Weve seen a closure being used to provide an anonymous block of functionality and the difference
between an equivalent lambda but, there are still more useful distinctions we can make.
Other Differences
An anonymous function, is a function literal without a name, whilst a closure is an instance of a
function. By definition, a lambda has no instance variables; its not an instance. Its variables are
supplied as arguments. A closure however, has instances variables which are provided when the
instance is created.
With this in mind, a lambda will generally be more efficient that a closure as it only needs to
evaluated once. Once you have the function, you can re-use it. As a closure closes over something
not in its local environment, it has to be evaluated every time its called. An instance has to be
newed up each time its used.
All the issues we looked at in the functions vs classes section are relevant here too. There may be
memory considerations to using closures over lambdas.
Summary
Weve talked about a lot here so lets summarise the differences briefly.
Lambdas are just anonymous functions, similar to static methods in Java. Just like static methods,
they cant reference variables outside their scope except for their arguments. A special type of
lambda, called a closure, can capture variables outside their scope (or closes over them) so they
can use external variables or their arguments. So the simple rule is if a lambda uses a variable from
outside its scope, its also a closure.
Closures can be seen as instances of functions. Which is kind of an odd concept for Java developers.
A great example is the conventional anonymous class that we would pass around if we didnt have
the new lambda syntax. These can close over variables and so are themselves closures. So weve
had closure support in Java since 1.1. Take a look at this example. The server variable has to be
closed over by the compiler to be used in the anonymous instance of the Condition interface. This
is both an anonymous class instance and a closure.
65
@since Java 1.1!
void anonymousClassClosure() {
Server server = new HttpServer();
waitFor(new Condition() {
@Override
public Boolean isSatisfied() {
return !server.isRunning();
}
});
}
Lambdas arent always closures, but closures are always lambdas.
66
Invocation & Bytecode
In this section well explore how the compiler output differs when you compile anonymous classes
to when you compile lambdas. First well remind ourselves about java bytecode and how to read
it. Then well look at both anonymous classes and lambdas when they capture variables and when
they dont. Well compare pre-Java 8 closures with lambdas and explore how lambdas are not just
syntactic sugar but produce very different bytecode from the traditional approaches.
Bytecode Recap
To start with, lets recap on what we know about bytecode.
To get from source code to machine runnable code. The Java compiler produces bytecode. This is
either interpreted by the JVM or re-compiled by the Just-in-time compiler.
When its interpreted, the bytecode is turned into machine code on the fly and executed. This
happens each time the bytecode is encountered but he JVM.
When its Just-in-time compiled, the JVM compiles it directly into machine code the first time its
encountered and then goes on to execute it.
Both happen at run-time but Just-in-time compilation offer lots of optimisations.
So, Java bytecode is the intermediate representation between source code and machine code.
..
As a quick side bar: Javas JIT compiler has enjoyed a great reputation over the years. But going
back full circle to our introduction, it was John McCarthy that first wrote about JIT compilation
way back in 1960. So its interesting to think that its not just lambda support that was influenced
by LISP. (Aycock 2003, 2. JIT Compilation Techniques, 2.1 Genesis, p. 98).
http://user.it.uu.se/~kostis/Teaching/KT2-04/jit_survey.pdf
The bytecode is the instruction set of the JVM. As its name suggests, bytecode consists of single-
byte instructions (called opcodes) along with associated bytes for parameters. There are therefore a
possible 256 opcodes available although only about 200 are actually used.
The JVM uses a stack based computation model, if we want to increment a number, we have to do
it using the stack. All instructions or opcodes work against the stack.
So for example, 5 + 1 becomes 5 1 + where 5 is pushed to the stack,
http://en.wikipedia.org/wiki/Model_of_computation
67
+---------+
push 5 -----> | 5 |
+---------+
| |
+---------+
| |
+---------+
1 is pushed then
+---------+
push 1 -----> | 1 |
+---------+
| 5 |
+---------+
| |
+---------+
the + operator is applied. Plus would pop the top two frames, add the numbers together and push
the result back onto the stack. The result would look like this.
+---------+
add -----> | 6 |
+---------+
| |
+---------+
| |
+---------+
Each opcode works against the stack like this so we can translate our example into a sequence of
bytecodes;
+---------+
push 5 -----> | 5 |
+---------+
| |
+---------+
| |
+---------+
push 5 becomes iconst_5.
68
+---------+
iconst_5 -----> | 5 |
+---------+
| |
+---------+
| |
+---------+
push 1 becomes iconst_1
+---------+
iconst_1 -----> | 1 |
+---------+
| 5 |
+---------+
| |
+---------+
and add becomes iadd.
+---------+
iadd -----> | 6 |
+---------+
| |
+---------+
| |
+---------+
iconst_x and iadd are examples of opcodes. Opcodes often have prefixes and/or suffices to indicate
the types they work on, i in these examples refers to integer.
We can group the opcodes into the following categories.
Group Examples
Stack manipulation aload_n, istore, swap, dup2
Control flow instructions goto, ifeq, iflt
Object interactions new, invokespecial, areturn
Arithmetic, logic and type conversion iadd, fcmpl, i2b
Instructions concerned with stack manipulation, like weve seen before. Examples being aload,
istore etc. To control program flow with things like if and while, we use opcodes like goto and
if equal. Creating objects and accessing methods use codes like new and invokespecial. Well be
69
particularly interested in this group when we look at the different opcodes used to invoke lambdas.
The last group is about arithmetic, logic and type conversion and includes codes like iadd, float
compare long (fcmpl) and integer to byte (i2b).
Descriptors
Opcodes will often use parameters, these look a little cryptic in the bytecode as they usually
referenced via lookup tables. Internally, Java uses whats called descriptors to describe these
parameters.
They describe types and signatures using a specific grammar youll see throughout the bytecode.
Youll often see the same grammar used in compiler or debug output, so its useful to recap it here.
Heres an example of a method signature descriptor.
Example$1."<init>":(Lcom/foo/Example;Lcom/foo/Server;)V
Its describing the constructor of a class called $1, which we happen to know is the JVMs name
for the first anonymous class instance within another class. In this case Example. So weve got
a constructor of an anonymous class that takes two parameters, an instance of the outer class
com.foo.Example and an instance of com.foo.Server.
Being a constructor, the method doesnt return anything. The V symbol represents void.
Have a look at breakdown of the descriptor syntax below. If you see an uppercase Z in a descriptor,
its referring to a boolean, an uppercase B a byte and so on.
70
A couple of ones to mention;
classes are described with an uppercase L followed by the fully qualified class name, followed
by a semi-colon. The class name is separated with slashes rather than the dots.
and arrays are described using an opening square bracket followed by a type from the list. No
closing bracket.
71
Converting a Method Signature
Lets take the following method signature and turn it into a method descriptor.
long f (int n, String s, int[] array);
The method returns a long, so we describe the fact that it is a method with brackets and that it
returns a long with a uppercase J.
()J
The first argument is of type int, so we use an uppercase I.
(I)J
The next argument is an object, so we use L to describe its an object, fully qualify the name and
close it with a semi-colon.
(ILString;)J
The last argument is an integer array so we drop in the array syntax followed by int type.
(ILString;[I)J
and were done. A JVM method descriptor.
72
Code Examples
Lets have a look at the bytecode produced for some examples.
Were going to look at the bytecode for four distinct blocks of functionality based on the example
we looked at in lambdas vs closures section.
Well explore
1. A simple anonymous class
2. An anonymous class closing over some variable (an old style closure)
3. A lambda with no arguments
4. A lambda with arguments
5. A lambda closing over some variable (a new style closure)
The example bytecode was generated using the javap command line tool. Only partial bytecode
listings are shown in this section, for full source and bytecode listsings, see Appendix A. Also, fully
qualified class names have been shortened to better fit on the page.
Example 1
The first example is a simple anonymous class instance passed into our waitFor method.
public class Example1 {
// anonymous class
void example() throws InterruptedException {
waitFor(new Condition() {
@Override
public Boolean isSatisfied() {
return true;
}
});
}
}
If we look at the bytecode, the thing to notice is that an instance of the anonymous class is newed
up at line 6. The #2 refers to a lookup, the result of which is shown in the comment. So it uses the
new opcode with whatever is at #2 in the constant pool, this happens to be the anonymous class
Example$1.
73
1 void example() throws java.lang.InterruptedException;
2 descriptor: ()V
3 flags:
4 Code:
5 stack=3, locals=1, args_size=1
6 0: new #2 // class Example1$1
7 3: dup
8 4: aload_0
9 5: invokespecial #3 // Method Example1$1."<init>":(LExample1;)V
10 8: invokestatic #4 // Method WaitFor.waitFor:(LCondition;)V
11 11: return
12 LineNumberTable:
13 line 10: 0
14 line 16: 11
15 LocalVariableTable:
16 Start Length Slot Name Signature
17 0 12 0 this LExample1;
18 Exceptions:
19 throws java.lang.InterruptedException
Once created, the constructor is called using invokespecial on line 9. This opcode is used to call
constructor methods, private methods and accessible methods of a super class. You might notice
the method descriptor includes a reference to Example1. All anonymous class instances have this
implicit reference to the parent class.
The next step uses invokestatic to call our waitFor method passing in the anonymous class on line
10. invokestatic is used to call static methods and is very fast as it can direct dial a method rather
than figure out which to call as would be the case in an object hierarchy.
Example 2
Example 2 is another anonymous class but this time it closes over the server variable. Its an old
style closure.
74
public class Example2 {
// anonymous class (closure)
void example() throws InterruptedException {
Server server = new HttpServer();
waitFor(new Condition() {
@Override
public Boolean isSatisfied() {
return !server.isRunning();
}
});
}
}
The bytecode is similar to the previous except that an instance of the Server class is newed up (at
line 3.) and its constructor called at line 5. The instance of the anonymous class $1 is still constructed
with invokespecial (at line 11.) but this time it takes the instance of Server as an argument as well
as the instance of the calling class.
To close over the server variable, its passed directly into the anonymous class.
1 void example() throws java.lang.InterruptedException;
2 Code:
3 0: new #2 // class Server$HttpServer
4 3: dup
5 4: invokespecial #3 // Method Server$HttpServer."<init>":()V
6 7: astore_1
7 8: new #4 // class Example2$1
8 11: dup
9 12: aload_0
10 13: aload_1
11 14: invokespecial #5 // Method Example2$1."<init>":(LExample2;LServer;)V
12 17: invokestatic #6 // Method WaitFor.waitFor:(LCondition;)V
13 20: return
Example 3
Example 3 uses a Java 8 lambda with our waitFor method. The lambda doesnt do anything other
than return true. Its equivalent to example 1.
75
public class Example3 {
// simple lambda
void example() throws InterruptedException {
waitFor(() -> true);
}
}
The bytecode is super simple this time. It uses the invokedynamic opcode to create the lambda at
line 3. which is then passed to the invokestatic opcode on the next line.
1 void example() throws java.lang.InterruptedException;
2 Code:
3 0: invokedynamic #2, 0 // InvokeDynamic #0:isSatisfied:()LCondition;
4 5: invokestatic #3 // Method WaitFor.waitFor:(LCondition;)V
5 8: return
The descriptor for the invokedynamic call is targeting the isSatisfied method on the Condition
interface (line 3.).
What were not seeing here is the mechanics of invokedynamic. invokedynamic is a new opcode
to Java 7, it was intended to provide better support for dynamic languages on the JVM. It does this
by not linking the types to methods until run-time. The other invoke opcodes all resolve types at
compile time.
For lambdas, this means that placeholder method invocations can be put into the bytecode like weve
just seen and working out the implementation can be done on the JVM at runtime.
If we look at a more verbose bytecode that includes the constant pool we can dereference the lookups.
For example, if we look up number 2, we can see it references #0 and #26.
1 Constant pool:
2 #1 = Methodref #6.#21 // Object."<init>":()V
3 #2 = InvokeDynamic #0:#26 // #0:isSatisfied:()LCondition;
4 ...
5 BootstrapMethods:
6 0: #23 invokestatic LambdaMetafactory.metafactory:
7 (LMethodHandles$Lookup;LString;
8 LMethodType;LMethodType;
9 LMethodHandle;LMethodType;)LCallSite;
10 Method arguments:
11 #24 ()LBoolean;
12 #25 invokestatic Example3.lambda$example$25:()LBoolean;
13 #24 ()LBoolean;
76
The constant 0 is in a special lookup table for bootstrapping methods (line 6.). It refers to a static
method call to the JDK LambdaMetafactory to create the lambda. This is where the heavy lifting
goes on. All the target type inference to adapt types and any partial argument evaluation goes on
here.
The actual lambda is shown as a method handle called lambda$example$25 (line 12.) with no
arguments, returning a boolean. Its invoked using invokestatic which shows that its accessed
as a genuine function; theres no object associated with it. Theres also no implicit reference to a
containing class unlike the anonymous examples before.
Its passed into the LambdaMetafactory and we know its a method handle by looking it up in the
constant pool. The number of the lambda is compiler assigned and just increments from zero for
each lambda required.
1 Constant pool:
2 // invokestatic Example3.lambda$example$25:()LBoolean;
3 #25 = MethodHandle #6:#35
77
Example 4
Example 4 is another lambda but this time it takes an instance of Server as an argument. Its
equivalent in functionality to example 2 but it doesnt close over the variable; its not a closure.
public class Example4 {
// lambda with arguments
void example() throws InterruptedException {
waitFor(new HttpServer(), (server) -> server.isRunning());
}
}
Just like example 2, the bytecode has to create the instance of server but this time, the invokedynamic
opcode references the test method of type Predicate. If we were to follow the reference (#4) to the
boostrap methods table, we would see the actual lambda requires an argument of type HttpServer
and returns a Z which is a primitive boolean.
1 void example() throws java.lang.InterruptedException;
2 descriptor: ()V
3 flags:
4 Code:
5 stack=2, locals=1, args_size=1
6 0: new #2 // class Server$HttpServer
7 3: dup
8 4: invokespecial #3 // Method Server$HttpServer."<init>":()V
9 7: invokedynamic #4, 0 // InvokeDynamic #0:test:()LPredicate;
10 12: invokestatic #5 // Method WaitFor.waitFor:(LObject;LPredicate;)V
11 15: return
12 LineNumberTable:
13 line 13: 0
14 line 15: 15
15 LocalVariableTable:
16 Start Length Slot Name Signature
17 0 16 0 this LExample4;
18 Exceptions:
19 throws java.lang.InterruptedException
So the call to the lambda is still a static method call like before but this time takes the variable as a
parameter when its invoked.
78
Example 4 (with method reference)
Interestingly, if we use a method reference instead, the functionality is exactly the same but we get
different bytecode.
public class Example4_method_reference {
// lambda with method reference
void example() throws InterruptedException {
waitFor(new HttpServer(), HttpServer::isRunning);
}
}
Via the call to the LambdaMetafactory when the final execution occurs, the method reference results
in a call to invokevirtual rather than invokestatic. invokevirtual is used to call public, protected
an package protected methods so it implies an instance is required. The instance is supplied to the
metafactory method and no lambda (or static function) is needed at all; there are no lambda$ in this
bytecode.
1 void example() throws java.lang.InterruptedException;
2 descriptor: ()V
3 flags:
4 Code:
5 stack=2, locals=1, args_size=1
6 0: new #2 // class Server$HttpServer
7 3: dup
8 4: invokespecial #3 // Method Server$HttpServer."<init>":()V
9 7: invokedynamic #4, 0 // InvokeDynamic #0:test:()LPredicate;
10 12: invokestatic #5 // Method WaitFor.waitFor:(LObject;LPredicate;)V
11 15: return
12 LineNumberTable:
13 line 11: 0
14 line 12: 15
15 LocalVariableTable:
16 Start Length Slot Name Signature
17 0 16 0 this LExample4_method_reference;
18 Exceptions:
19 throws java.lang.InterruptedException
Example 5
Lastly, example 5 uses a lambda but closes over the server instance. Its equivalent to example 2
and is a new style closure.
79
public class Example5 {
// closure
void example() throws InterruptedException {
Server server = new HttpServer();
waitFor(() -> !server.isRunning());
}
}
It goes through the basics in the same way as the other lambdas but if we lookup the metafactory
method in the bootstrap methods table, youll notice that this time, the lambdas method handle has
an argument of type Server. Its invoked using invokestatic (line 9.) and the variable is passed
directly into the lambda at invocation time.
1 BootstrapMethods:
2 0: #34 invokestatic LambdaMetafactory.metafactory:
3 (LMethodHandles$Lookup;
4 LString;LMethodType;
5 LMethodType;
6 LMethodHandle;LMethodType;)LCallSite;
7 Method arguments:
8 #35 ()LBoolean; // <-- SAM method to be implemented by the lambda
9 #36 invokestatic Example5.lambda$example$35:(LServer;)LBoolean;
10 #35 ()LBoolean; // <-- type to be enforced at invocation time
So like the anonymous class in example 2, an argument is added by the compiler to capture the term
although this time, its a method argument rather than a constructor argument.
80
Summary
We saw how using an anonymous class will create a new instance and call its constructor with
invokespecial.
We saw anonymous classes that close over variables have an extra argument on their constructor to
capture that variable.
and we saw how Java 8 lambdas use the invokedynamic instruction to defer binding of the types and
that a special lambda$ method handle is used to actually represent the lambda. This method handle
has no arguments in this case and is invoked using invokestatic making it a genuine function.
The lambda was created by the LambdaMetafactory class which itself was the target of the
invokedynamic call.
When a lambda has arguments, we saw how the LambdaMetafactory describes the argument to be
passed into the lambda. invokestatic is used to execute the lambda like before. But we also had a
look at a method reference used in-lieu of a lambda. In this case, no lambda$ method handle was
created and invokevirtual was used to call the method directly.
Lastly, we looked at a lambda that closes over a variable. This one creates an argument on the
lambda$ method handle and again is called with invokestatic.
Appendix A
Appendix A 82
Bytecode
WaitFor
package jdk8.byte_code;
class WaitFor {
static void waitFor(Condition condition) throws InterruptedException {
while (!condition.isSatisfied())
Thread.sleep(250);
}
static <T> void waitFor(T input, Predicate<T> predicate)
throws InterruptedException {
while (!predicate.test(input))
Thread.sleep(250);
}
}
Appendix A 83
Example 1
package jdk8.byte_code;
import static jdk8.byte_code.WaitFor.waitFor;
@SuppressWarnings("all")
public class Example1 {
// anonymous class
void example() throws InterruptedException {
waitFor(new Condition() {
@Override
public Boolean isSatisfied() {
return true;
}
});
}
}
Classfile Example1.class
Last modified 08-May-2014; size 603 bytes
MD5 checksum 7365ca98fe204fc9198043cef5d241be
Compiled from "Example1.java"
public class jdk8.byte_code.Example1
SourceFile: "Example1.java"
InnerClasses:
#2; //class jdk8/byte_code/Example1$1
minor version: 0
major version: 52
flags: ACC_PUBLIC, ACC_SUPER
Constant pool:
#1 = Methodref #6.#20 // java/lang/Object."<init>":()V
#2 = Class #21 // jdk8/byte_code/Example1$1
#3 = Methodref #2.#22 // jdk8/byte_code/Example1$1."<init>":\
(Ljdk8/byte_code/Example1;)V
#4 = Methodref #23.#24 // jdk8/byte_code/WaitFor.waitFor:(Ljd\
k8/byte_code/Condition;)V
#5 = Class #25 // jdk8/byte_code/Example1
#6 = Class #26 // java/lang/Object
#7 = Utf8 InnerClasses
#8 = Utf8 <init>
Appendix A 84
#9 = Utf8 ()V
#10 = Utf8 Code
#11 = Utf8 LineNumberTable
#12 = Utf8 LocalVariableTable
#13 = Utf8 this
#14 = Utf8 Ljdk8/byte_code/Example1;
#15 = Utf8 example
#16 = Utf8 Exceptions
#17 = Class #27 // java/lang/InterruptedException
#18 = Utf8 SourceFile
#19 = Utf8 Example1.java
#20 = NameAndType #8:#9 // "<init>":()V
#21 = Utf8 jdk8/byte_code/Example1$1
#22 = NameAndType #8:#28 // "<init>":(Ljdk8/byte_code/Example1;\
)V
#23 = Class #29 // jdk8/byte_code/WaitFor
#24 = NameAndType #30:#31 // waitFor:(Ljdk8/byte_code/Condition;\
)V
#25 = Utf8 jdk8/byte_code/Example1
#26 = Utf8 java/lang/Object
#27 = Utf8 java/lang/InterruptedException
#28 = Utf8 (Ljdk8/byte_code/Example1;)V
#29 = Utf8 jdk8/byte_code/WaitFor
#30 = Utf8 waitFor
#31 = Utf8 (Ljdk8/byte_code/Condition;)V
{
public jdk8.byte_code.Example1();
descriptor: ()V
flags: ACC_PUBLIC
Code:
stack=1, locals=1, args_size=1
0: aload_0
1: invokespecial #1 // Method java/lang/Object."<init>"\
:()V
4: return
LineNumberTable:
line 6: 0
LocalVariableTable:
Start Length Slot Name Signature
0 5 0 this Ljdk8/byte_code/Example1;
void example() throws java.lang.InterruptedException;
Appendix A 85
descriptor: ()V
flags:
Code:
stack=3, locals=1, args_size=1
0: new #2 // class jdk8/byte_code/Example1$1
3: dup
4: aload_0
5: invokespecial #3 // Method jdk8/byte_code/Example1$1\
."<init>":(Ljdk8/byte_code/Example1;)V
8: invokestatic #4 // Method jdk8/byte_code/WaitFor.wa\
itFor:(Ljdk8/byte_code/Condition;)V
11: return
LineNumberTable:
line 10: 0
line 16: 11
LocalVariableTable:
Start Length Slot Name Signature
0 12 0 this Ljdk8/byte_code/Example1;
Exceptions:
throws java.lang.InterruptedException
}
{pagebreak}
Example 2
package jdk8.byte_code;
public interface Server {
Boolean isRunning();
public class HttpServer implements Server {
@Override
public Boolean isRunning() {
return false;
}
}
}
package jdk8.byte_code;
Appendix A 86
import static jdk8.byte_code.Server.*;
import static jdk8.byte_code.WaitFor.waitFor;
public class Example2 {
// anonymous class (closure)
void example() throws InterruptedException {
Server server = new HttpServer();
waitFor(new Condition() {
@Override
public Boolean isSatisfied() {
return !server.isRunning();
}
});
}
}
Classfile Example2.class
Last modified 08-May-2014; size 775 bytes
MD5 checksum 2becf3c32e2b08abc50465aca7398c4b
Compiled from "Example2.java"
public class jdk8.byte_code.Example2
SourceFile: "Example2.java"
InnerClasses:
#4; //class jdk8/byte_code/Example2$1
public static #27= #2 of #25; //HttpServer=class jdk8/byte_code/Server$Htt\
pServer of class jdk8/byte_code/Server
minor version: 0
major version: 52
flags: ACC_PUBLIC, ACC_SUPER
Constant pool:
#1 = Methodref #8.#24 // java/lang/Object."<init>":()V
#2 = Class #26 // jdk8/byte_code/Server$HttpServer
#3 = Methodref #2.#24 // jdk8/byte_code/Server$HttpServer."<\
init>":()V
#4 = Class #28 // jdk8/byte_code/Example2$1
#5 = Methodref #4.#29 // jdk8/byte_code/Example2$1."<init>":\
(Ljdk8/byte_code/Example2;Ljdk8/byte_code/Server;)V
#6 = Methodref #30.#31 // jdk8/byte_code/WaitFor.waitFor:(Ljd\
k8/byte_code/Condition;)V
#7 = Class #32 // jdk8/byte_code/Example2
Appendix A 87
#8 = Class #33 // java/lang/Object
#9 = Utf8 InnerClasses
#10 = Utf8 <init>
#11 = Utf8 ()V
#12 = Utf8 Code
#13 = Utf8 LineNumberTable
#14 = Utf8 LocalVariableTable
#15 = Utf8 this
#16 = Utf8 Ljdk8/byte_code/Example2;
#17 = Utf8 example
#18 = Utf8 server
#19 = Utf8 Ljdk8/byte_code/Server;
#20 = Utf8 Exceptions
#21 = Class #34 // java/lang/InterruptedException
#22 = Utf8 SourceFile
#23 = Utf8 Example2.java
#24 = NameAndType #10:#11 // "<init>":()V
#25 = Class #35 // jdk8/byte_code/Server
#26 = Utf8 jdk8/byte_code/Server$HttpServer
#27 = Utf8 HttpServer
#28 = Utf8 jdk8/byte_code/Example2$1
#29 = NameAndType #10:#36 // "<init>":(Ljdk8/byte_code/Example2;\
Ljdk8/byte_code/Server;)V
#30 = Class #37 // jdk8/byte_code/WaitFor
#31 = NameAndType #38:#39 // waitFor:(Ljdk8/byte_code/Condition;\
)V
#32 = Utf8 jdk8/byte_code/Example2
#33 = Utf8 java/lang/Object
#34 = Utf8 java/lang/InterruptedException
#35 = Utf8 jdk8/byte_code/Server
#36 = Utf8 (Ljdk8/byte_code/Example2;Ljdk8/byte_code/Server;)V
#37 = Utf8 jdk8/byte_code/WaitFor
#38 = Utf8 waitFor
#39 = Utf8 (Ljdk8/byte_code/Condition;)V
{
public jdk8.byte_code.Example2();
descriptor: ()V
flags: ACC_PUBLIC
Code:
stack=1, locals=1, args_size=1
0: aload_0
1: invokespecial #1 // Method java/lang/Object."<init>"\
Appendix A 88
:()V
4: return
LineNumberTable:
line 6: 0
LocalVariableTable:
Start Length Slot Name Signature
0 5 0 this Ljdk8/byte_code/Example2;
void example() throws java.lang.InterruptedException;
descriptor: ()V
flags:
Code:
stack=4, locals=2, args_size=1
0: new #2 // class jdk8/byte_code/Server$Http\
Server
3: dup
4: invokespecial #3 // Method jdk8/byte_code/Server$Htt\
pServer."<init>":()V
7: astore_1
8: new #4 // class jdk8/byte_code/Example2$1
11: dup
12: aload_0
13: aload_1
14: invokespecial #5 // Method jdk8/byte_code/Example2$1\
."<init>":(Ljdk8/byte_code/Example2;Ljdk8/byte_code/Server;)V
17: invokestatic #6 // Method jdk8/byte_code/WaitFor.wa\
itFor:(Ljdk8/byte_code/Condition;)V
20: return
LineNumberTable:
line 10: 0
line 11: 8
line 17: 20
LocalVariableTable:
Start Length Slot Name Signature
0 21 0 this Ljdk8/byte_code/Example2;
8 13 1 server Ljdk8/byte_code/Server;
Exceptions:
throws java.lang.InterruptedException
}
Appendix A 89
Example 3
package jdk8.byte_code;
import static jdk8.byte_code.WaitFor.waitFor;
public class Example3 {
// simple lambda
void example() throws InterruptedException {
waitFor(() -> true);
}
}
Classfile Example3.class
Last modified 08-May-2014; size 1155 bytes
MD5 checksum 22e120de85528efc921bb158588bbaa1
Compiled from "Example3.java"
public class jdk8.byte_code.Example3
SourceFile: "Example3.java"
InnerClasses:
public static final #50= #49 of #53; //Lookup=class java/lang/invoke/Metho\
dHandles$Lookup of class java/lang/invoke/MethodHandles
BootstrapMethods:
0: #23 invokestatic java/lang/invoke/LambdaMetafactory.metafactory:(Ljava/lan\
g/invoke/MethodHandles$Lookup;Ljava/lang/String;Ljava/lang/invoke/MethodType;Ljav\
a/lang/invoke/MethodType;Ljava/lang/invoke/MethodHandle;Ljava/lang/invoke/MethodT\
ype;)Ljava/lang/invoke/CallSite;
Method arguments:
#24 ()Ljava/lang/Boolean;
#25 invokestatic jdk8/byte_code/Example3.lambda$example$25:()Ljava/lang/B\
oolean;
#24 ()Ljava/lang/Boolean;
minor version: 0
major version: 52
flags: ACC_PUBLIC, ACC_SUPER
Constant pool:
#1 = Methodref #6.#21 // java/lang/Object."<init>":()V
#2 = InvokeDynamic #0:#26 // #0:isSatisfied:()Ljdk8/byte_code/Co\
ndition;
#3 = Methodref #27.#28 // jdk8/byte_code/WaitFor.waitFor:(Ljd\
k8/byte_code/Condition;)V
#4 = Methodref #29.#30 // java/lang/Boolean.valueOf:(Z)Ljava/\
Appendix A 90
lang/Boolean;
#5 = Class #31 // jdk8/byte_code/Example3
#6 = Class #32 // java/lang/Object
#7 = Utf8 <init>
#8 = Utf8 ()V
#9 = Utf8 Code
#10 = Utf8 LineNumberTable
#11 = Utf8 LocalVariableTable
#12 = Utf8 this
#13 = Utf8 Ljdk8/byte_code/Example3;
#14 = Utf8 example
#15 = Utf8 Exceptions
#16 = Class #33 // java/lang/InterruptedException
#17 = Utf8 lambda$example$25
#18 = Utf8 ()Ljava/lang/Boolean;
#19 = Utf8 SourceFile
#20 = Utf8 Example3.java
#21 = NameAndType #7:#8 // "<init>":()V
#22 = Utf8 BootstrapMethods
#23 = MethodHandle #6:#34 // invokestatic java/lang/invoke/Lambd\
aMetafactory.metafactory:(Ljava/lang/invoke/MethodHandles$Lookup;Ljava/lang/Strin\
g;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/Met\
hodHandle;Ljava/lang/invoke/MethodType;)Ljava/lang/invoke/CallSite;
#24 = MethodType #18 // ()Ljava/lang/Boolean;
#25 = MethodHandle #6:#35 // invokestatic jdk8/byte_code/Example\
3.lambda$example$25:()Ljava/lang/Boolean;
#26 = NameAndType #36:#37 // isSatisfied:()Ljdk8/byte_code/Condi\
tion;
#27 = Class #38 // jdk8/byte_code/WaitFor
#28 = NameAndType #39:#40 // waitFor:(Ljdk8/byte_code/Condition;\
)V
#29 = Class #41 // java/lang/Boolean
#30 = NameAndType #42:#43 // valueOf:(Z)Ljava/lang/Boolean;
#31 = Utf8 jdk8/byte_code/Example3
#32 = Utf8 java/lang/Object
#33 = Utf8 java/lang/InterruptedException
#34 = Methodref #44.#45 // java/lang/invoke/LambdaMetafactory.\
metafactory:(Ljava/lang/invoke/MethodHandles$Lookup;Ljava/lang/String;Ljava/lang/\
invoke/MethodType;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/MethodHandle;Lja\
va/lang/invoke/MethodType;)Ljava/lang/invoke/CallSite;
#35 = Methodref #5.#46 // jdk8/byte_code/Example3.lambda$exam\
ple$25:()Ljava/lang/Boolean;
Appendix A 91
#36 = Utf8 isSatisfied
#37 = Utf8 ()Ljdk8/byte_code/Condition;
#38 = Utf8 jdk8/byte_code/WaitFor
#39 = Utf8 waitFor
#40 = Utf8 (Ljdk8/byte_code/Condition;)V
#41 = Utf8 java/lang/Boolean
#42 = Utf8 valueOf
#43 = Utf8 (Z)Ljava/lang/Boolean;
#44 = Class #47 // java/lang/invoke/LambdaMetafactory
#45 = NameAndType #48:#52 // metafactory:(Ljava/lang/invoke/Meth\
odHandles$Lookup;Ljava/lang/String;Ljava/lang/invoke/MethodType;Ljava/lang/invoke\
/MethodType;Ljava/lang/invoke/MethodHandle;Ljava/lang/invoke/MethodType;)Ljava/la\
ng/invoke/CallSite;
#46 = NameAndType #17:#18 // lambda$example$25:()Ljava/lang/Bool\
ean;
#47 = Utf8 java/lang/invoke/LambdaMetafactory
#48 = Utf8 metafactory
#49 = Class #54 // java/lang/invoke/MethodHandles$Look\
up
#50 = Utf8 Lookup
#51 = Utf8 InnerClasses
#52 = Utf8 (Ljava/lang/invoke/MethodHandles$Lookup;Ljava/lang/Str\
ing;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/M\
ethodHandle;Ljava/lang/invoke/MethodType;)Ljava/lang/invoke/CallSite;
#53 = Class #55 // java/lang/invoke/MethodHandles
#54 = Utf8 java/lang/invoke/MethodHandles$Lookup
#55 = Utf8 java/lang/invoke/MethodHandles
{
public jdk8.byte_code.Example3();
descriptor: ()V
flags: ACC_PUBLIC
Code:
stack=1, locals=1, args_size=1
0: aload_0
1: invokespecial #1 // Method java/lang/Object."<init>"\
:()V
4: return
LineNumberTable:
line 6: 0
LocalVariableTable:
Start Length Slot Name Signature
0 5 0 this Ljdk8/byte_code/Example3;
Appendix A 92
void example() throws java.lang.InterruptedException;
descriptor: ()V
flags:
Code:
stack=1, locals=1, args_size=1
0: invokedynamic #2, 0 // InvokeDynamic #0:isSatisfied:()L\
jdk8/byte_code/Condition;
5: invokestatic #3 // Method jdk8/byte_code/WaitFor.wa\
itFor:(Ljdk8/byte_code/Condition;)V
8: return
LineNumberTable:
line 10: 0
line 11: 8
LocalVariableTable:
Start Length Slot Name Signature
0 9 0 this Ljdk8/byte_code/Example3;
Exceptions:
throws java.lang.InterruptedException
}
Appendix A 93

Example 4
package jdk8.byte_code;
import static jdk8.byte_code.Server.HttpServer;
import static jdk8.byte_code.WaitFor.waitFor;
public class Example4 {
// lambda with arguments
void example() throws InterruptedException {
waitFor(new HttpServer(), (server) -> server.isRunning());
}
}
Classfile Example4.class
Last modified 08-May-2014; size 1414 bytes
MD5 checksum 7177f97fdf30b0648a09ab98109a479c
Compiled from "Example4.java"
public class jdk8.byte_code.Example4
SourceFile: "Example4.java"
InnerClasses:
public static #21= #2 of #29; //HttpServer=class jdk8/byte_code/Server$Htt\
pServer of class jdk8/byte_code/Server
public static final #65= #64 of #67; //Lookup=class java/lang/invoke/Metho\
dHandles$Lookup of class java/lang/invoke/MethodHandles
BootstrapMethods:
0: #32 invokestatic java/lang/invoke/LambdaMetafactory.metafactory:(Ljava/lan\
g/invoke/MethodHandles$Lookup;Ljava/lang/String;Ljava/lang/invoke/MethodType;Ljav\
a/lang/invoke/MethodType;Ljava/lang/invoke/MethodHandle;Ljava/lang/invoke/MethodT\
ype;)Ljava/lang/invoke/CallSite;
Method arguments:
#33 (Ljava/lang/Object;)Z
#34 invokestatic jdk8/byte_code/Example4.lambda$example$33:(Ljdk8/byte_co\
de/Server$HttpServer;)Z
#35 (Ljdk8/byte_code/Server$HttpServer;)Z
minor version: 0
major version: 52
flags: ACC_PUBLIC, ACC_SUPER
Constant pool:
Appendix A 94
#1 = Methodref #9.#28 // java/lang/Object."<init>":()V
#2 = Class #30 // jdk8/byte_code/Server$HttpServer
#3 = Methodref #2.#28 // jdk8/byte_code/Server$HttpServer."<\
init>":()V
#4 = InvokeDynamic #0:#36 // #0:test:()Ljava/util/function/Predi\
cate;
#5 = Methodref #37.#38 // jdk8/byte_code/WaitFor.waitFor:(Lja\
va/lang/Object;Ljava/util/function/Predicate;)V
#6 = Methodref #2.#39 // jdk8/byte_code/Server$HttpServer.is\
Running:()Ljava/lang/Boolean;
#7 = Methodref #40.#41 // java/lang/Boolean.booleanValue:()Z
#8 = Class #42 // jdk8/byte_code/Example4
#9 = Class #43 // java/lang/Object
#10 = Utf8 <init>
#11 = Utf8 ()V
#12 = Utf8 Code
#13 = Utf8 LineNumberTable
#14 = Utf8 LocalVariableTable
#15 = Utf8 this
#16 = Utf8 Ljdk8/byte_code/Example4;
#17 = Utf8 example
#18 = Utf8 Exceptions
#19 = Class #44 // java/lang/InterruptedException
#20 = Utf8 lambda$example$33
#21 = Utf8 HttpServer
#22 = Utf8 InnerClasses
#23 = Utf8 (Ljdk8/byte_code/Server$HttpServer;)Z
#24 = Utf8 server
#25 = Utf8 Ljdk8/byte_code/Server$HttpServer;
#26 = Utf8 SourceFile
#27 = Utf8 Example4.java
#28 = NameAndType #10:#11 // "<init>":()V
#29 = Class #45 // jdk8/byte_code/Server
#30 = Utf8 jdk8/byte_code/Server$HttpServer
#31 = Utf8 BootstrapMethods
#32 = MethodHandle #6:#46 // invokestatic java/lang/invoke/Lambd\
aMetafactory.metafactory:(Ljava/lang/invoke/MethodHandles$Lookup;Ljava/lang/Strin\
g;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/Met\
hodHandle;Ljava/lang/invoke/MethodType;)Ljava/lang/invoke/CallSite;
#33 = MethodType #47 // (Ljava/lang/Object;)Z
#34 = MethodHandle #6:#48 // invokestatic jdk8/byte_code/Example\
4.lambda$example$33:(Ljdk8/byte_code/Server$HttpServer;)Z
Appendix A 95
#35 = MethodType #23 // (Ljdk8/byte_code/Server$HttpServer;\
)Z
#36 = NameAndType #49:#50 // test:()Ljava/util/function/Predicat\
e;
#37 = Class #51 // jdk8/byte_code/WaitFor
#38 = NameAndType #52:#53 // waitFor:(Ljava/lang/Object;Ljava/ut\
il/function/Predicate;)V
#39 = NameAndType #54:#55 // isRunning:()Ljava/lang/Boolean;
#40 = Class #56 // java/lang/Boolean
#41 = NameAndType #57:#58 // booleanValue:()Z
#42 = Utf8 jdk8/byte_code/Example4
#43 = Utf8 java/lang/Object
#44 = Utf8 java/lang/InterruptedException
#45 = Utf8 jdk8/byte_code/Server
#46 = Methodref #59.#60 // java/lang/invoke/LambdaMetafactory.\
metafactory:(Ljava/lang/invoke/MethodHandles$Lookup;Ljava/lang/String;Ljava/lang/\
invoke/MethodType;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/MethodHandle;Lja\
va/lang/invoke/MethodType;)Ljava/lang/invoke/CallSite;
#47 = Utf8 (Ljava/lang/Object;)Z
#48 = Methodref #8.#61 // jdk8/byte_code/Example4.lambda$exam\
ple$33:(Ljdk8/byte_code/Server$HttpServer;)Z
#49 = Utf8 test
#50 = Utf8 ()Ljava/util/function/Predicate;
#51 = Utf8 jdk8/byte_code/WaitFor
#52 = Utf8 waitFor
#53 = Utf8 (Ljava/lang/Object;Ljava/util/function/Predicate;)V
#54 = Utf8 isRunning
#55 = Utf8 ()Ljava/lang/Boolean;
#56 = Utf8 java/lang/Boolean
#57 = Utf8 booleanValue
#58 = Utf8 ()Z
#59 = Class #62 // java/lang/invoke/LambdaMetafactory
#60 = NameAndType #63:#66 // metafactory:(Ljava/lang/invoke/Meth\
odHandles$Lookup;Ljava/lang/String;Ljava/lang/invoke/MethodType;Ljava/lang/invoke\
/MethodType;Ljava/lang/invoke/MethodHandle;Ljava/lang/invoke/MethodType;)Ljava/la\
ng/invoke/CallSite;
#61 = NameAndType #20:#23 // lambda$example$33:(Ljdk8/byte_code/\
Server$HttpServer;)Z
#62 = Utf8 java/lang/invoke/LambdaMetafactory
#63 = Utf8 metafactory
#64 = Class #68 // java/lang/invoke/MethodHandles$Look\
up
Appendix A 96
#65 = Utf8 Lookup
#66 = Utf8 (Ljava/lang/invoke/MethodHandles$Lookup;Ljava/lang/Str\
ing;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/M\
ethodHandle;Ljava/lang/invoke/MethodType;)Ljava/lang/invoke/CallSite;
#67 = Class #69 // java/lang/invoke/MethodHandles
#68 = Utf8 java/lang/invoke/MethodHandles$Lookup
#69 = Utf8 java/lang/invoke/MethodHandles
{
public jdk8.byte_code.Example4();
descriptor: ()V
flags: ACC_PUBLIC
Code:
stack=1, locals=1, args_size=1
0: aload_0
1: invokespecial #1 // Method java/lang/Object."<init>"\
:()V
4: return
LineNumberTable:
line 9: 0
LocalVariableTable:
Start Length Slot Name Signature
0 5 0 this Ljdk8/byte_code/Example4;
void example() throws java.lang.InterruptedException;
descriptor: ()V
flags:
Code:
stack=2, locals=1, args_size=1
0: new #2 // class jdk8/byte_code/Server$Http\
Server
3: dup
4: invokespecial #3 // Method jdk8/byte_code/Server$Htt\
pServer."<init>":()V
7: invokedynamic #4, 0 // InvokeDynamic #0:test:()Ljava/ut\
il/function/Predicate;
12: invokestatic #5 // Method jdk8/byte_code/WaitFor.wa\
itFor:(Ljava/lang/Object;Ljava/util/function/Predicate;)V
15: return
LineNumberTable:
line 13: 0
line 15: 15
LocalVariableTable:
Appendix A 97
Start Length Slot Name Signature
0 16 0 this Ljdk8/byte_code/Example4;
Exceptions:
throws java.lang.InterruptedException
}
Appendix A 98
Example 4 (with Method Reference)
package jdk8.byte_code;
import static jdk8.byte_code.Server.HttpServer;
import static jdk8.byte_code.WaitFor.waitFor;
@SuppressWarnings("all")
public class Example4_method_reference {
// lambda with method reference
void example() throws InterruptedException {
waitFor(new HttpServer(), HttpServer::isRunning);
}
}
Classfile Example4_method_reference.class
Last modified 08-May-2014; size 1271 bytes
MD5 checksum f8aef942361f29ef599adfec7a594948
Compiled from "Example4_method_reference.java"
public class jdk8.byte_code.Example4_method_reference
SourceFile: "Example4_method_reference.java"
InnerClasses:
public static #23= #2 of #21; //HttpServer=class jdk8/byte_code/Server$Htt\
pServer of class jdk8/byte_code/Server
public static final #52= #51 of #56; //Lookup=class java/lang/invoke/Metho\
dHandles$Lookup of class java/lang/invoke/MethodHandles
BootstrapMethods:
0: #26 invokestatic java/lang/invoke/LambdaMetafactory.metafactory:(Ljava/lan\
g/invoke/MethodHandles$Lookup;Ljava/lang/String;Ljava/lang/invoke/MethodType;Ljav\
a/lang/invoke/MethodType;Ljava/lang/invoke/MethodHandle;Ljava/lang/invoke/MethodT\
ype;)Ljava/lang/invoke/CallSite;
Method arguments:
#27 (Ljava/lang/Object;)Z
#28 invokevirtual jdk8/byte_code/Server$HttpServer.isRunning:()Ljava/lang\
/Boolean;
#29 (Ljdk8/byte_code/Server$HttpServer;)Z
minor version: 0
major version: 52
flags: ACC_PUBLIC, ACC_SUPER
Constant pool:
#1 = Methodref #7.#20 // java/lang/Object."<init>":()V
Appendix A 99
#2 = Class #22 // jdk8/byte_code/Server$HttpServer
#3 = Methodref #2.#20 // jdk8/byte_code/Server$HttpServer."<\
init>":()V
#4 = InvokeDynamic #0:#30 // #0:test:()Ljava/util/function/Predi\
cate;
#5 = Methodref #31.#32 // jdk8/byte_code/WaitFor.waitFor:(Lja\
va/lang/Object;Ljava/util/function/Predicate;)V
#6 = Class #33 // jdk8/byte_code/Example4_method_refe\
rence
#7 = Class #34 // java/lang/Object
#8 = Utf8 <init>
#9 = Utf8 ()V
#10 = Utf8 Code
#11 = Utf8 LineNumberTable
#12 = Utf8 LocalVariableTable
#13 = Utf8 this
#14 = Utf8 Ljdk8/byte_code/Example4_method_reference;
#15 = Utf8 example
#16 = Utf8 Exceptions
#17 = Class #35 // java/lang/InterruptedException
#18 = Utf8 SourceFile
#19 = Utf8 Example4_method_reference.java
#20 = NameAndType #8:#9 // "<init>":()V
#21 = Class #36 // jdk8/byte_code/Server
#22 = Utf8 jdk8/byte_code/Server$HttpServer
#23 = Utf8 HttpServer
#24 = Utf8 InnerClasses
#25 = Utf8 BootstrapMethods
#26 = MethodHandle #6:#37 // invokestatic java/lang/invoke/Lambd\
aMetafactory.metafactory:(Ljava/lang/invoke/MethodHandles$Lookup;Ljava/lang/Strin\
g;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/Met\
hodHandle;Ljava/lang/invoke/MethodType;)Ljava/lang/invoke/CallSite;
#27 = MethodType #38 // (Ljava/lang/Object;)Z
#28 = MethodHandle #5:#39 // invokevirtual jdk8/byte_code/Server\
$HttpServer.isRunning:()Ljava/lang/Boolean;
#29 = MethodType #40 // (Ljdk8/byte_code/Server$HttpServer;\
)Z
#30 = NameAndType #41:#42 // test:()Ljava/util/function/Predicat\
e;
#31 = Class #43 // jdk8/byte_code/WaitFor
#32 = NameAndType #44:#45 // waitFor:(Ljava/lang/Object;Ljava/ut\
il/function/Predicate;)V
Appendix A 100
#33 = Utf8 jdk8/byte_code/Example4_method_reference
#34 = Utf8 java/lang/Object
#35 = Utf8 java/lang/InterruptedException
#36 = Utf8 jdk8/byte_code/Server
#37 = Methodref #46.#47 // java/lang/invoke/LambdaMetafactory.\
metafactory:(Ljava/lang/invoke/MethodHandles$Lookup;Ljava/lang/String;Ljava/lang/\
invoke/MethodType;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/MethodHandle;Lja\
va/lang/invoke/MethodType;)Ljava/lang/invoke/CallSite;
#38 = Utf8 (Ljava/lang/Object;)Z
#39 = Methodref #2.#48 // jdk8/byte_code/Server$HttpServer.is\
Running:()Ljava/lang/Boolean;
#40 = Utf8 (Ljdk8/byte_code/Server$HttpServer;)Z
#41 = Utf8 test
#42 = Utf8 ()Ljava/util/function/Predicate;
#43 = Utf8 jdk8/byte_code/WaitFor
#44 = Utf8 waitFor
#45 = Utf8 (Ljava/lang/Object;Ljava/util/function/Predicate;)V
#46 = Class #49 // java/lang/invoke/LambdaMetafactory
#47 = NameAndType #50:#53 // metafactory:(Ljava/lang/invoke/Meth\
odHandles$Lookup;Ljava/lang/String;Ljava/lang/invoke/MethodType;Ljava/lang/invoke\
/MethodType;Ljava/lang/invoke/MethodHandle;Ljava/lang/invoke/MethodType;)Ljava/la\
ng/invoke/CallSite;
#48 = NameAndType #54:#55 // isRunning:()Ljava/lang/Boolean;
#49 = Utf8 java/lang/invoke/LambdaMetafactory
#50 = Utf8 metafactory
#51 = Class #57 // java/lang/invoke/MethodHandles$Look\
up
#52 = Utf8 Lookup
#53 = Utf8 (Ljava/lang/invoke/MethodHandles$Lookup;Ljava/lang/Str\
ing;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/M\
ethodHandle;Ljava/lang/invoke/MethodType;)Ljava/lang/invoke/CallSite;
#54 = Utf8 isRunning
#55 = Utf8 ()Ljava/lang/Boolean;
#56 = Class #58 // java/lang/invoke/MethodHandles
#57 = Utf8 java/lang/invoke/MethodHandles$Lookup
#58 = Utf8 java/lang/invoke/MethodHandles
{
public jdk8.byte_code.Example4_method_reference();
descriptor: ()V
flags: ACC_PUBLIC
Code:
stack=1, locals=1, args_size=1
Appendix A 101
0: aload_0
1: invokespecial #1 // Method java/lang/Object."<init>"\
:()V
4: return
LineNumberTable:
line 7: 0
LocalVariableTable:
Start Length Slot Name Signature
0 5 0 this Ljdk8/byte_code/Example4_method_reference;
void example() throws java.lang.InterruptedException;
descriptor: ()V
flags:
Code:
stack=2, locals=1, args_size=1
0: new #2 // class jdk8/byte_code/Server$Http\
Server
3: dup
4: invokespecial #3 // Method jdk8/byte_code/Server$Htt\
pServer."<init>":()V
7: invokedynamic #4, 0 // InvokeDynamic #0:test:()Ljava/ut\
il/function/Predicate;
12: invokestatic #5 // Method jdk8/byte_code/WaitFor.wa\
itFor:(Ljava/lang/Object;Ljava/util/function/Predicate;)V
15: return
LineNumberTable:
line 11: 0
line 12: 15
LocalVariableTable:
Start Length Slot Name Signature
0 16 0 this Ljdk8/byte_code/Example4_method_reference;
Exceptions:
throws java.lang.InterruptedException
}
Appendix A 102
Example 5
package jdk8.byte_code;
import static jdk8.byte_code.Server.*;
import static jdk8.byte_code.WaitFor.waitFor;
public class Example5 {
// closure
void example() throws InterruptedException {
Server server = new HttpServer();
waitFor(() -> !server.isRunning());
}
}
Classfile Example5.class
Last modified 08-May-2014; size 1470 bytes
MD5 checksum 7dd0f577d4b4b903500264acf9649c30
Compiled from "Example5.java"
public class jdk8.byte_code.Example5
SourceFile: "Example5.java"
InnerClasses:
public static #31= #2 of #29; //HttpServer=class jdk8/byte_code/Server$Htt\
pServer of class jdk8/byte_code/Server
public static final #68= #67 of #70; //Lookup=class java/lang/invoke/Metho\
dHandles$Lookup of class java/lang/invoke/MethodHandles
BootstrapMethods:
0: #34 invokestatic java/lang/invoke/LambdaMetafactory.metafactory:(Ljava/lan\
g/invoke/MethodHandles$Lookup;Ljava/lang/String;Ljava/lang/invoke/MethodType;Ljav\
a/lang/invoke/MethodType;Ljava/lang/invoke/MethodHandle;Ljava/lang/invoke/MethodT\
ype;)Ljava/lang/invoke/CallSite;
Method arguments:
#35 ()Ljava/lang/Boolean; // <-- signature and return type of the SAM met\
hod to be implemented by the lambda
#36 invokestatic jdk8/byte_code/Example5.lambda$example$35:(Ljdk8/byte_co\
de/Server;)Ljava/lang/Boolean;
#35 ()Ljava/lang/Boolean; // <-- signature and return type to be enforced\
at invocation time
minor version: 0
major version: 52
flags: ACC_PUBLIC, ACC_SUPER
Appendix A 103
Constant pool:
#1 = Methodref #10.#28 // java/lang/Object."<init>":()V
#2 = Class #30 // jdk8/byte_code/Server$HttpServer
#3 = Methodref #2.#28 // jdk8/byte_code/Server$HttpServer."<\
init>":()V
#4 = InvokeDynamic #0:#37 // #0:isSatisfied:(Ljdk8/byte_code/Ser\
ver;)Ljdk8/byte_code/Condition;
#5 = Methodref #38.#39 // jdk8/byte_code/WaitFor.waitFor:(Ljd\
k8/byte_code/Condition;)V
#6 = InterfaceMethodref #29.#40 // jdk8/byte_code/Server.isRunning:()L\
java/lang/Boolean;
#7 = Methodref #41.#42 // java/lang/Boolean.booleanValue:()Z
#8 = Methodref #41.#43 // java/lang/Boolean.valueOf:(Z)Ljava/\
lang/Boolean;
#9 = Class #44 // jdk8/byte_code/Example5
#10 = Class #45 // java/lang/Object
#11 = Utf8 <init>
#12 = Utf8 ()V
#13 = Utf8 Code
#14 = Utf8 LineNumberTable
#15 = Utf8 LocalVariableTable
#16 = Utf8 this
#17 = Utf8 Ljdk8/byte_code/Example5;
#18 = Utf8 example
#19 = Utf8 server
#20 = Utf8 Ljdk8/byte_code/Server;
#21 = Utf8 Exceptions
#22 = Class #46 // java/lang/InterruptedException
#23 = Utf8 lambda$example$35
#24 = Utf8 (Ljdk8/byte_code/Server;)Ljava/lang/Boolean;
#25 = Utf8 StackMapTable
#26 = Utf8 SourceFile
#27 = Utf8 Example5.java
#28 = NameAndType #11:#12 // "<init>":()V
#29 = Class #47 // jdk8/byte_code/Server
#30 = Utf8 jdk8/byte_code/Server$HttpServer
#31 = Utf8 HttpServer
#32 = Utf8 InnerClasses
#33 = Utf8 BootstrapMethods
#34 = MethodHandle #6:#48 // invokestatic java/lang/invoke/Lambd\
aMetafactory.metafactory:(Ljava/lang/invoke/MethodHandles$Lookup;Ljava/lang/Strin\
g;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/Met\
Appendix A 104
hodHandle;Ljava/lang/invoke/MethodType;)Ljava/lang/invoke/CallSite;
#35 = MethodType #49 // ()Ljava/lang/Boolean;
#36 = MethodHandle #6:#50 // invokestatic jdk8/byte_code/Example\
5.lambda$example$35:(Ljdk8/byte_code/Server;)Ljava/lang/Boolean;
#37 = NameAndType #51:#52 // isSatisfied:(Ljdk8/byte_code/Server\
;)Ljdk8/byte_code/Condition;
#38 = Class #53 // jdk8/byte_code/WaitFor
#39 = NameAndType #54:#55 // waitFor:(Ljdk8/byte_code/Condition;\
)V
#40 = NameAndType #56:#49 // isRunning:()Ljava/lang/Boolean;
#41 = Class #57 // java/lang/Boolean
#42 = NameAndType #58:#59 // booleanValue:()Z
#43 = NameAndType #60:#61 // valueOf:(Z)Ljava/lang/Boolean;
#44 = Utf8 jdk8/byte_code/Example5
#45 = Utf8 java/lang/Object
#46 = Utf8 java/lang/InterruptedException
#47 = Utf8 jdk8/byte_code/Server
#48 = Methodref #62.#63 // java/lang/invoke/LambdaMetafactory.\
metafactory:(Ljava/lang/invoke/MethodHandles$Lookup;Ljava/lang/String;Ljava/lang/\
invoke/MethodType;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/MethodHandle;Lja\
va/lang/invoke/MethodType;)Ljava/lang/invoke/CallSite;
#49 = Utf8 ()Ljava/lang/Boolean;
#50 = Methodref #9.#64 // jdk8/byte_code/Example5.lambda$exam\
ple$35:(Ljdk8/byte_code/Server;)Ljava/lang/Boolean;
#51 = Utf8 isSatisfied
#52 = Utf8 (Ljdk8/byte_code/Server;)Ljdk8/byte_code/Condition;
#53 = Utf8 jdk8/byte_code/WaitFor
#54 = Utf8 waitFor
#55 = Utf8 (Ljdk8/byte_code/Condition;)V
#56 = Utf8 isRunning
#57 = Utf8 java/lang/Boolean
#58 = Utf8 booleanValue
#59 = Utf8 ()Z
#60 = Utf8 valueOf
#61 = Utf8 (Z)Ljava/lang/Boolean;
#62 = Class #65 // java/lang/invoke/LambdaMetafactory
#63 = NameAndType #66:#69 // metafactory:(Ljava/lang/invoke/Meth\
odHandles$Lookup;Ljava/lang/String;Ljava/lang/invoke/MethodType;Ljava/lang/invoke\
/MethodType;Ljava/lang/invoke/MethodHandle;Ljava/lang/invoke/MethodType;)Ljava/la\
ng/invoke/CallSite;
#64 = NameAndType #23:#24 // lambda$example$35:(Ljdk8/byte_code/\
Server;)Ljava/lang/Boolean;
Appendix A 105
#65 = Utf8 java/lang/invoke/LambdaMetafactory
#66 = Utf8 metafactory
#67 = Class #71 // java/lang/invoke/MethodHandles$Look\
up
#68 = Utf8 Lookup
#69 = Utf8 (Ljava/lang/invoke/MethodHandles$Lookup;Ljava/lang/Str\
ing;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/MethodType;Ljava/lang/invoke/M\
ethodHandle;Ljava/lang/invoke/MethodType;)Ljava/lang/invoke/CallSite;
#70 = Class #72 // java/lang/invoke/MethodHandles
#71 = Utf8 java/lang/invoke/MethodHandles$Lookup
#72 = Utf8 java/lang/invoke/MethodHandles
{
public jdk8.byte_code.Example5();
descriptor: ()V
flags: ACC_PUBLIC
Code:
stack=1, locals=1, args_size=1
0: aload_0
1: invokespecial #1 // Method java/lang/Object."<init>"\
:()V
4: return
LineNumberTable:
line 6: 0
LocalVariableTable:
Start Length Slot Name Signature
0 5 0 this Ljdk8/byte_code/Example5;
void example() throws java.lang.InterruptedException;
descriptor: ()V
flags:
Code:
stack=2, locals=2, args_size=1
0: new #2 // class jdk8/byte_code/Server$Http\
Server
3: dup
4: invokespecial #3 // Method jdk8/byte_code/Server$Htt\
pServer."<init>":()V
7: astore_1
8: aload_1
9: invokedynamic #4, 0 // InvokeDynamic #0:isSatisfied:(Ljdk8\
/byte_code/Server;)Ljdk8/byte_code/Condition;
Appendix A 106
14: invokestatic #5 // Method jdk8/byte_code/WaitFor.wa\
itFor:(Ljdk8/byte_code/Condition;)V
17: return
LineNumberTable:
line 10: 0
line 11: 8
line 12: 17
LocalVariableTable:
Start Length Slot Name Signature
0 18 0 this Ljdk8/byte_code/Example5;
8 10 1 server Ljdk8/byte_code/Server;
Exceptions:
throws java.lang.InterruptedException
}