Beginner

Introductory, no previous programming experience

• C++ Primer * (Stanley Lippman, Josée Lajoie, and Barbara E. Moo) (updated for C++11) Coming at 1k pages, this is a very thorough introduction into C++ that covers just about everything in the language in a very accessible format and in great detail. The fifth edition (released August 16, 2012) covers C++11. [Review]

  ^{* Not to be confused with C++ Primer Plus (Stephen Prata), with a significantly less favorable review.}

• Programming: Principles and Practice Using C++ (Bjarne Stroustrup, 2nd Edition - May 25, 2014) (updated for C++11/C++14) An introduction to programming using C++ by the creator of the language. A good read, that assumes no previous programming experience, but is not only for beginners.

Introductory, with previous programming experience

• A Tour of C++ (Bjarne Stroustrup) (2nd edition for C++17) The “tour” is a quick (about 180 pages and 14 chapters) tutorial overview of all of standard C++ (language and standard library, and using C++11) at a moderately high level for people who already know C++ or at least are experienced programmers. This book is an extended version of the material that constitutes Chapters 2-5 of The C++ Programming Language, 4th edition.

• Accelerated C++ (Andrew Koenig and Barbara Moo, 1st Edition - August 24, 2000) This basically covers the same ground as the C++ Primer, but does so in a quarter of its space. This is largely because it does not attempt to be an introduction to programming, but an introduction to C++ for people who've previously programmed in some other language. It has a steeper learning curve, but, for those who can cope with this, it is a very compact introduction to the language. (Historically, it broke new ground by being the first beginner's book to use a modern approach to teaching the language.) Despite this, the C++ it teaches is purely C++98. [Review]

Best practices

• Effective C++ (Scott Meyers, 3rd Edition - May 22, 2005) This was written with the aim of being the best second book C++ programmers should read, and it succeeded. Earlier editions were aimed at programmers coming from C, the third edition changes this and targets programmers coming from languages like Java. It presents ~50 easy-to-remember rules of thumb along with their rationale in a very accessible (and enjoyable) style. For C++11 and C++14 the examples and a few issues are outdated and Effective Modern C++ should be preferred. [Review]

• Effective Modern C++ (Scott Meyers) This is basically the new version of Effective C++, aimed at C++ programmers making the transition from C++03 to C++11 and C++14.

• Effective STL (Scott Meyers) This aims to do the same to the part of the standard library coming from the STL what Effective C++ did to the language as a whole: It presents rules of thumb along with their rationale. [Review]

Intermediate

• More Effective C++ (Scott Meyers) Even more rules of thumb than Effective C++. Not as important as the ones in the first book, but still good to know.

• Exceptional C++ (Herb Sutter) Presented as a set of puzzles, this has one of the best and thorough discussions of the proper resource management and exception safety in C++ through Resource Acquisition is Initialization (RAII) in addition to in-depth coverage of a variety of other topics including the pimpl idiom, name lookup, good class design, and the C++ memory model. [Review]

• More Exceptional C++ (Herb Sutter) Covers additional exception safety topics not covered in Exceptional C++, in addition to discussion of effective object-oriented programming in C++ and correct use of the STL. [Review]

• Exceptional C++ Style (Herb Sutter) Discusses generic programming, optimization, and resource management; this book also has an excellent exposition of how to write modular code in C++ by using non-member functions and the single responsibility principle. [Review]

• C++ Coding Standards (Herb Sutter and Andrei Alexandrescu) “Coding standards” here doesn't mean “how many spaces should I indent my code?” This book contains 101 best practices, idioms, and common pitfalls that can help you to write correct, understandable, and efficient C++ code. [Review]

• C++ Templates: The Complete Guide (David Vandevoorde and Nicolai M. Josuttis) This is the book about templates as they existed before C++11. It covers everything from the very basics to some of the most advanced template metaprogramming and explains every detail of how templates work (both conceptually and at how they are implemented) and discusses many common pitfalls. Has excellent summaries of the One Definition Rule (ODR) and overload resolution in the appendices. A second edition covering C++11, C++14 and C++17 has been already published. [Review]

• C++ 17 - The Complete Guide (Nicolai M. Josuttis) This book describes all the new features introduced in the C++17 Standard covering everything from the simple ones like 'Inline Variables', 'constexpr if' all the way up to 'Polymorphic Memory Resources' and 'New and Delete with over aligned Data'. [Review]

• C++ in Action (Bartosz Milewski). This book explains C++ and its features by building an application from ground up. [Review]

• Functional Programming in C++ (Ivan Čukić). This book introduces functional programming techniques to modern C++ (C++11 and later). A very nice read for those who want to apply functional programming paradigms to C++.

Advanced

• Modern C++ Design (Andrei Alexandrescu) A groundbreaking book on advanced generic programming techniques. Introduces policy-based design, type lists, and fundamental generic programming idioms then explains how many useful design patterns (including small object allocators, functors, factories, visitors, and multi-methods) can be implemented efficiently, modularly, and cleanly using generic programming. [Review]

• C++ Template Metaprogramming (David Abrahams and Aleksey Gurtovoy)

• C++ Concurrency In Action (Anthony Williams) A book covering C++11 concurrency support including the thread library, the atomics library, the C++ memory model, locks and mutexes, as well as issues of designing and debugging multithreaded applications. A second edition covering C++14 and C++17 has already been published. [Review]

• Advanced C++ Metaprogramming (Davide Di Gennaro) A pre-C++11 manual of TMP techniques, focused more on practice than theory. There are a ton of snippets in this book, some of which are made obsolete by type traits, but the techniques, are nonetheless useful to know. If you can put up with the quirky formatting/editing, it is easier to read than Alexandrescu, and arguably, more rewarding. For more experienced developers, there is a good chance that you may pick up something about a dark corner of C++ (a quirk) that usually only comes about through extensive experience.

• Large Scale C++ volume I, Process and architecture (John Lakos). Part one of a three-part series extending the older book 'Large Scale C++ Design'. Lakos explains battle-tested techniques to manage very big C++ software projects. If you work in a big C++ software project this is a great read, detailing the relationship between physical and logical structure, strategies for components, and their reuse.

Reference Style - All Levels

• The C++ Programming Language (Bjarne Stroustrup) (updated for C++11) The classic introduction to C++ by its creator. Written to parallel the classic K&R, this indeed reads very much like it and covers just about everything from the core language to the standard library, to programming paradigms to the language's philosophy. [Review] Note: All releases of the C++ standard are tracked in the question "Where do I find the current C or C++ standard documents?".

• C++ Standard Library Tutorial and Reference (Nicolai Josuttis) (updated for C++11) The introduction and reference for the C++ Standard Library. The second edition (released on April 9, 2012) covers C++11. [Review]

• The C++ IO Streams and Locales (Angelika Langer and Klaus Kreft) There's very little to say about this book except that if you want to know anything about streams and locales, then this is the one place to find definitive answers. [Review]

C++11/14/17/… References:

• The C++11/14/17 Standard (INCITS/ISO/IEC 14882:2011/2014/2017) This, of course, is the final arbiter of all that is or isn't C++. Be aware, however, that it is intended purely as a reference for experienced users willing to devote considerable time and effort to its understanding. The C++17 standard is released in electronic form for 198 Swiss Francs.

• The C++17 standard is available, but seemingly not in an economical form – directly from the ISO it costs 198 Swiss Francs (about $200 US). For most people, the final draft before standardization is more than adequate (and free). Many will prefer an even newer draft, documenting new features that are likely to be included in C++20.

• Overview of the New C++ (C++11/14) (PDF only) (Scott Meyers) (updated for C++14) These are the presentation materials (slides and some lecture notes) of a three-day training course offered by Scott Meyers, who's a highly respected author on C++. Even though the list of items is short, the quality is high.

• The C++ Core Guidelines (C++11/14/17/…) (edited by Bjarne Stroustrup and Herb Sutter) is an evolving online document consisting of a set of guidelines for using modern C++ well. The guidelines are focused on relatively higher-level issues, such as interfaces, resource management, memory management, and concurrency affecting application architecture and library design. The project was announced at CppCon'15 by Bjarne Stroustrup and others and welcomes contributions from the community. Most guidelines are supplemented with a rationale and examples as well as discussions of possible tool support. Many rules are designed specifically to be automatically checkable by static analysis tools.

• The C++ Super-FAQ (Marshall Cline, Bjarne Stroustrup, and others) is an effort by the Standard C++ Foundation to unify the C++ FAQs previously maintained individually by Marshall Cline and Bjarne Stroustrup and also incorporating new contributions. The items mostly address issues at an intermediate level and are often written with a humorous tone. Not all items might be fully up to date with the latest edition of the C++ standard yet.

• cppreference.com (C++03/11/14/17/…) (initiated by Nate Kohl) is a wiki that summarizes the basic core-language features and has extensive documentation of the C++ standard library. The documentation is very precise but is easier to read than the official standard document and provides better navigation due to its wiki nature. The project documents all versions of the C++ standard and the site allows filtering the display for a specific version. The project was presented by Nate Kohl at CppCon'14.

Classics / Older

Note: Some information contained within these books may not be up-to-date or no longer considered best practice.

• The Design and Evolution of C++ (Bjarne Stroustrup) If you want to know why the language is the way it is, this book is where you find answers. This covers everything before the standardization of C++.

• Ruminations on C++ - (Andrew Koenig and Barbara Moo) [Review]

• Advanced C++ Programming Styles and Idioms (James Coplien) A predecessor of the pattern movement, it describes many C++-specific “idioms”. It's certainly a very good book and might still be worth a read if you can spare the time, but quite old and not up-to-date with current C++.

• Large Scale C++ Software Design (John Lakos) Lakos explains techniques to manage very big C++ software projects. Certainly, a good read, if it only was up to date. It was written long before C++ 98 and misses on many features (e.g. namespaces) important for large-scale projects. If you need to work on a big C++ software project, you might want to read it, although you need to take more than a grain of salt with it. Not to be confused with the extended and later book series Large Scale C++ volume I-III.

• Inside the C++ Object Model (Stanley Lippman) If you want to know how virtual member functions are commonly implemented and how base objects are commonly laid out in memory in a multi-inheritance scenario, and how all this affects performance, this is where you will find thorough discussions of such topics.

• The Annotated C++ Reference Manual (Bjarne Stroustrup, Margaret A. Ellis) This book is quite outdated in the fact that it explores the 1989 C++ 2.0 version - Templates, exceptions, namespaces, and new casts were not yet introduced. Saying that however, this book goes through the entire C++ standard of the time explaining the rationale, the possible implementations, and features of the language. This is not a book to learn programming principles and patterns on C++, but to understand every aspect of the C++ language.

• Thinking in C++ (Bruce Eckel, 2nd Edition, 2000). Two volumes; is a tutorial-style free set of intro level books. Downloads: vol 1, vol 2. Unfortunately, they're marred by a number of trivial errors (e.g. maintaining that temporaries are automatic const), with no official errata list. A partial 3^rd party errata list is available at http://www.computersciencelab.com/Eckel.htm, but it is apparently not maintained.

• Scientific and Engineering C++: An Introduction to Advanced Techniques and Examples (John Barton and Lee Nackman) It is a comprehensive and very detailed book that tried to explain and make use of all the features available in C++, in the context of numerical methods. It introduced at the time several new techniques, such as the Curiously Recurring Template Pattern (CRTP, also called Barton-Nackman trick). It pioneered several techniques such as dimensional analysis and automatic differentiation. It came with a lot of compilable and useful code, ranging from an expression parser to a Lapack wrapper. The code is still available online. Unfortunately, the books have become somewhat outdated in the style and C++ features, however, it was an incredible tour-de-force at the time (1994, pre-STL). The chapters on dynamics inheritance are a bit complicated to understand and not very useful. An updated version of this classic book that includes move semantics and the lessons learned from the STL would be very nice.

Overview

Why do we need the copy-and-swap idiom?

Any class that manages a resource (a wrapper, like a smart pointer) needs to implement The Big Three. While the goals and implementation of the copy-constructor and destructor are straightforward, the copy-assignment operator is arguably the most nuanced and difficult. How should it be done? What pitfalls need to be avoided?

The copy-and-swap idiom is the solution, and elegantly assists the assignment operator in achieving two things: avoiding code duplication, and providing a strong exception guarantee.

How does it work?

Conceptually, it works by using the copy-constructor's functionality to create a local copy of the data, then takes the copied data with a swap function, swapping the old data with the new data. The temporary copy then destructs, taking the old data with it. We are left with a copy of the new data.

In order to use the copy-and-swap idiom, we need three things: a working copy-constructor, a working destructor (both are the basis of any wrapper, so should be complete anyway), and a swap function.

A swap function is a non-throwing function that swaps two objects of a class, member for member. We might be tempted to use std::swap instead of providing our own, but this would be impossible; std::swap uses the copy-constructor and copy-assignment operator within its implementation, and we'd ultimately be trying to define the assignment operator in terms of itself!

(Not only that, but unqualified calls to swap will use our custom swap operator, skipping over the unnecessary construction and destruction of our class that std::swap would entail.)

An in-depth explanation

The goal

Let's consider a concrete case. We want to manage, in an otherwise useless class, a dynamic array. We start with a working constructor, copy-constructor, and destructor:

#include <algorithm> // std::copy
#include <cstddef> // std::size_t

class dumb_array
{
public:
    // (default) constructor
    dumb_array(std::size_t size = 0)
        : mSize(size),
          mArray(mSize ? new int[mSize]() : nullptr)
    {
    }

    // copy-constructor
    dumb_array(const dumb_array& other)
        : mSize(other.mSize),
          mArray(mSize ? new int[mSize] : nullptr)
    {
        // note that this is non-throwing, because of the data
        // types being used; more attention to detail with regards
        // to exceptions must be given in a more general case, however
        std::copy(other.mArray, other.mArray + mSize, mArray);
    }

    // destructor
    ~dumb_array()
    {
        delete [] mArray;
    }

private:
    std::size_t mSize;
    int* mArray;
};

This class almost manages the array successfully, but it needs operator= to work correctly.

A failed solution

Here's how a naive implementation might look:

// the hard part
dumb_array& operator=(const dumb_array& other)
{
    if (this != &other) // (1)
    {
        // get rid of the old data...
        delete [] mArray; // (2)
        mArray = nullptr; // (2) *(see footnote for rationale)

        // ...and put in the new
        mSize = other.mSize; // (3)
        mArray = mSize ? new int[mSize] : nullptr; // (3)
        std::copy(other.mArray, other.mArray + mSize, mArray); // (3)
    }

    return *this;
}

And we say we're finished; this now manages an array, without leaks. However, it suffers from three problems, marked sequentially in the code as (n).

1. The first is the self-assignment test.
   This check serves two purposes: it's an easy way to prevent us from running needless code on self-assignment, and it protects us from subtle bugs (such as deleting the array only to try and copy it). But in all other cases it merely serves to slow the program down, and act as noise in the code; self-assignment rarely occurs, so most of the time this check is a waste.
   It would be better if the operator could work properly without it.

2. The second is that it only provides a basic exception guarantee. If new int[mSize] fails, *this will have been modified. (Namely, the size is wrong and the data is gone!)
   For a strong exception guarantee, it would need to be something akin to:

    dumb_array& operator=(const dumb_array& other)
    {
        if (this != &other) // (1)
        {
            // get the new data ready before we replace the old
            std::size_t newSize = other.mSize;
            int* newArray = newSize ? new int[newSize]() : nullptr; // (3)
            std::copy(other.mArray, other.mArray + newSize, newArray); // (3)
   
            // replace the old data (all are non-throwing)
            delete [] mArray;
            mSize = newSize;
            mArray = newArray;
        }
   
        return *this;
    }
   

3. The code has expanded! Which leads us to the third problem: code duplication.

Our assignment operator effectively duplicates all the code we've already written elsewhere, and that's a terrible thing.

In our case, the core of it is only two lines (the allocation and the copy), but with more complex resources this code bloat can be quite a hassle. We should strive to never repeat ourselves.

(One might wonder: if this much code is needed to manage one resource correctly, what if my class manages more than one?
While this may seem to be a valid concern, and indeed it requires non-trivial try/catch clauses, this is a non-issue.
That's because a class should manage one resource only!)

A successful solution

As mentioned, the copy-and-swap idiom will fix all these issues. But right now, we have all the requirements except one: a swap function. While The Rule of Three successfully entails the existence of our copy-constructor, assignment operator, and destructor, it should really be called "The Big Three and A Half": any time your class manages a resource it also makes sense to provide a swap function.

We need to add swap functionality to our class, and we do that as follows†:

class dumb_array
{
public:
    // ...

    friend void swap(dumb_array& first, dumb_array& second) // nothrow
    {
        // enable ADL (not necessary in our case, but good practice)
        using std::swap;

        // by swapping the members of two objects,
        // the two objects are effectively swapped
        swap(first.mSize, second.mSize);
        swap(first.mArray, second.mArray);
    }

    // ...
};

(Here is the explanation why public friend swap.) Now not only can we swap our dumb_array's, but swaps in general can be more efficient; it merely swaps pointers and sizes, rather than allocating and copying entire arrays. Aside from this bonus in functionality and efficiency, we are now ready to implement the copy-and-swap idiom.

Without further ado, our assignment operator is:

dumb_array& operator=(dumb_array other) // (1)
{
    swap(*this, other); // (2)

    return *this;
}

And that's it! With one fell swoop, all three problems are elegantly tackled at once.

Why does it work?

We first notice an important choice: the parameter argument is taken by-value. While one could just as easily do the following (and indeed, many naive implementations of the idiom do):

dumb_array& operator=(const dumb_array& other)
{
    dumb_array temp(other);
    swap(*this, temp);

    return *this;
}

We lose an important optimization opportunity. Not only that, but this choice is critical in C++11, which is discussed later. (On a general note, a remarkably useful guideline is as follows: if you're going to make a copy of something in a function, let the compiler do it in the parameter list.‡)

Either way, this method of obtaining our resource is the key to eliminating code duplication: we get to use the code from the copy-constructor to make the copy, and never need to repeat any bit of it. Now that the copy is made, we are ready to swap.

Observe that upon entering the function that all the new data is already allocated, copied, and ready to be used. This is what gives us a strong exception guarantee for free: we won't even enter the function if construction of the copy fails, and it's therefore not possible to alter the state of *this. (What we did manually before for a strong exception guarantee, the compiler is doing for us now; how kind.)

At this point we are home-free, because swap is non-throwing. We swap our current data with the copied data, safely altering our state, and the old data gets put into the temporary. The old data is then released when the function returns. (Where upon the parameter's scope ends and its destructor is called.)

Because the idiom repeats no code, we cannot introduce bugs within the operator. Note that this means we are rid of the need for a self-assignment check, allowing a single uniform implementation of operator=. (Additionally, we no longer have a performance penalty on non-self-assignments.)

And that is the copy-and-swap idiom.

What about C++11?

The next version of C++, C++11, makes one very important change to how we manage resources: the Rule of Three is now The Rule of Four (and a half). Why? Because not only do we need to be able to copy-construct our resource, we need to move-construct it as well.

Luckily for us, this is easy:

class dumb_array
{
public:
    // ...

    // move constructor
    dumb_array(dumb_array&& other) noexcept ††
        : dumb_array() // initialize via default constructor, C++11 only
    {
        swap(*this, other);
    }

    // ...
};

What's going on here? Recall the goal of move-construction: to take the resources from another instance of the class, leaving it in a state guaranteed to be assignable and destructible.

So what we've done is simple: initialize via the default constructor (a C++11 feature), then swap with other; we know a default constructed instance of our class can safely be assigned and destructed, so we know other will be able to do the same, after swapping.

(Note that some compilers do not support constructor delegation; in this case, we have to manually default construct the class. This is an unfortunate but luckily trivial task.)

Why does that work?

That is the only change we need to make to our class, so why does it work? Remember the ever-important decision we made to make the parameter a value and not a reference:

dumb_array& operator=(dumb_array other); // (1)

Now, if other is being initialized with an rvalue, it will be move-constructed. Perfect. In the same way C++03 let us re-use our copy-constructor functionality by taking the argument by-value, C++11 will automatically pick the move-constructor when appropriate as well. (And, of course, as mentioned in previously linked article, the copying/moving of the value may simply be elided altogether.)

And so concludes the copy-and-swap idiom.

Footnotes

*Why do we set mArray to null? Because if any further code in the operator throws, the destructor of dumb_array might be called; and if that happens without setting it to null, we attempt to delete memory that's already been deleted! We avoid this by setting it to null, as deleting null is a no-operation.

†There are other claims that we should specialize std::swap for our type, provide an in-class swap along-side a free-function swap, etc. But this is all unnecessary: any proper use of swap will be through an unqualified call, and our function will be found through ADL. One function will do.

‡The reason is simple: once you have the resource to yourself, you may swap and/or move it (C++11) anywhere it needs to be. And by making the copy in the parameter list, you maximize optimization.

††The move constructor should generally be noexcept, otherwise some code (e.g. std::vector resizing logic) will use the copy constructor even when a move would make sense. Of course, only mark it noexcept if the code inside doesn't throw exceptions.