Ad Hoc Polymorphism and Heterogeneous Containers with Value Semantics

Ad hoc polymorphism and heterogeneous containers with value semantics

Different alternatives

It is possible. There are several alternative approaches to your problem. Each one has different advantages and drawbacks (I will explain each one):

Create an interface and have a template class which implements this interface for different types. It should support cloning.
Use boost::variant and visitation.

Blending static and dynamic polymorphism

For the first alternative you need to create an interface like this:

class UsableInterface 
{
public:
    virtual ~UsableInterface() {}
    virtual void use() = 0;
    virtual std::unique_ptr<UsableInterface> clone() const = 0;
};

Obviously, you don't want to implement this interface by hand everytime you have a new type having the use() function. Therefore, let's have a template class which does that for you.

template <typename T> class UsableImpl : public UsableInterface
{
public:
    template <typename ...Ts> UsableImpl( Ts&&...ts ) 
        : t( std::forward<Ts>(ts)... ) {}
    virtual void use() override { use( t ); }
    virtual std::unique_ptr<UsableInterface> clone() const override
    {
        return std::make_unique<UsableImpl<T>>( t ); // This is C++14
        // This is the C++11 way to do it:
        // return std::unique_ptr<UsableImpl<T> >( new UsableImpl<T>(t) ); 
    }

private:
    T t;
};

Now you can actually already do everything you need with it. You can put these things in a vector:

std::vector<std::unique_ptr<UsableInterface>> usables;
// fill it

And you can copy that vector preserving the underlying types:

std::vector<std::unique_ptr<UsableInterface>> copies;
std::transform( begin(usables), end(usables), back_inserter(copies), 
    []( const std::unique_ptr<UsableInterface> & p )
    { return p->clone(); } );

You probably don't want to litter your code with stuff like this. What you want to write is

copies = usables;

Well, you can get that convenience by wrapping the std::unique_ptr into a class which supports copying.

class Usable
{
public:
    template <typename T> Usable( T t )
        : p( std::make_unique<UsableImpl<T>>( std::move(t) ) ) {}
    Usable( const Usable & other ) 
        : p( other.clone() ) {}
    Usable( Usable && other ) noexcept 
        : p( std::move(other.p) ) {}
    void swap( Usable & other ) noexcept 
        { p.swap(other.p); }
    Usable & operator=( Usable other ) 
        { swap(other); }
    void use()
        { p->use(); }
private:
    std::unique_ptr<UsableInterface> p;
};

Because of the nice templated contructor you can now write stuff like

Usable u1 = 5;
Usable u2 = std::string("Hello usable!");

And you can assign values with proper value semantics:

u1 = u2;

And you can put Usables in an std::vector

std::vector<Usable> usables;
usables.emplace_back( std::string("Hello!") );
usables.emplace_back( 42 );

and copy that vector

const auto copies = usables;

You can find this idea in Sean Parents talk Value Semantics and Concepts-based Polymorphism. He also gave a very brief version of this talk at Going Native 2013, but I think this is to fast to follow.

Moreover, you can take a more generic approach than writing your own Usable class and forwarding all the member functions (if you want to add other later). The idea is to replace the class Usable with a template class. This template class will not provide a member function use() but an operator T&() and operator const T&() const. This gives you the same functionality, but you don't need to write an extra value class every time you facilitate this pattern.

A safe, generic, stack-based discriminated union container

The template class boost::variant is exactly that and provides something like a C style union but safe and with proper value semantics. The way to use it is this:

using Usable = boost::variant<int,std::string,A>;
Usable usable;

You can assign from objects of any of these types to a Usable.

usable = 1;
usable = "Hello variant!";
usable = A();

If all template types have value semantics, then boost::variant also has value semantics and can be put into STL containers. You can write a use() function for such an object by a pattern that is called the visitor pattern. It calls the correct use() function for the contained object depending on the internal type.

class UseVisitor : public boost::static_visitor<void>
{
public:
    template <typename T>
    void operator()( T && t )
    {
        use( std::forward<T>(t) );
    }
}

void use( const Usable & u )
{
    boost::apply_visitor( UseVisitor(), u );
}

Now you can write

Usable u = "Hello";
use( u );

And, as I already mentioned, you can put these thingies into STL containers.

std::vector<Usable> usables;
usables.emplace_back( 5 );
usables.emplace_back( "Hello world!" );
const auto copies = usables;

The trade-offs

You can grow the functionality in two dimensions:

Add new classes which satisfy the static interface.
Add new functions which the classes must implement.

In the first approach I presented it is easier to add new classes. The second approach makes it easier to add new functionality.

In the first approach it it impossible (or at least hard) for client code to add new functions. In the second approach it is impossible (or at least hard) for client code to add new classes to the mix. A way out is the so-called acyclic visitor pattern which makes it possible for clients to extend a class hierarchy with new classes and new functionality. The drawback here is that you have to sacrifice a certain amount of static checking at compile-time. Here's a link which describes the visitor pattern including the acyclic visitor pattern along with some other alternatives. If you have questions about this stuff, I'm willing to answer.

Both approaches are super type-safe. There is not trade-off to be made there.

The run-time-costs of the first approach can be much higher, since there is a heap allocation involved for each element you create. The boost::variant approach is stack based and therefore is probably faster. If performance is a problem with the first approach consider to switch to the second.

Can I have polymorphic containers with value semantics in C++?

Since the objects of different classes will have different sizes, you would end up running into the slicing problem if you store them as values.

One reasonable solution is to store container safe smart pointers. I normally use boost::shared_ptr which is safe to store in a container. Note that std::auto_ptr is not.

vector<shared_ptr<Parent>> vec;
vec.push_back(shared_ptr<Parent>(new Child()));

shared_ptr uses reference counting so it will not delete the underlying instance until all references are removed.

Can I Have Polymorphic Containers With Value Semantics in C++11?

Just for fun, based on James's comment about a template-based system, I came up with this Vector-like implementation. It's missing lots of features, and may be buggy, but it's a start!

#include <iostream>
#include <vector>
#include <boost/shared_ptr.hpp>

template <typename T>
class Vector
{
public:
    T &operator[] (int i) const { return p[i]->get(); }
    template <typename D>
    void push_back(D &x) { p.push_back(ptr_t(new DerivedNode<D>(x))); }

private:
    class Node
    {
    public:
        virtual T &get() = 0;
    };

    template <typename D>
    class DerivedNode : public Node
    {
    public:
        DerivedNode(D &x) : x(x) {}
        virtual D &get() { return x; }
    private:
        D x;
    };

    typedef boost::shared_ptr<Node> ptr_t;
    std::vector<ptr_t> p;
};

///////////////////////////////////////

class Parent
{
    public:
        Parent() : parent_mem(1) {}
        virtual void write() const { std::cout << "Parent: " << parent_mem << std::endl; }
        int parent_mem;
};

class Child : public Parent
{
    public:
        Child() : child_mem(2) { parent_mem = 2; }
        void write() const { std::cout << "Child: " << parent_mem << ", " << child_mem << std::endl; }

        int child_mem;
};

int main()
{
    Vector<Parent> v;

    v.push_back(Parent());
    v.push_back(Child());

    v[0].write();
    v[1].write();
}

Run-time polymorphic class with value semantics

Yes, it is possible, but of course there must be some hidden pointer, and the actual data must be stored on the heap. The reason is that the actual size of the data cannot be known at compile-time, and then can't be on the stack.

The idea is to store the actual implementation through a pointer of a polymorphic class ValueImpl, that provides any virtual method you need, like increments() or print(), and in addition a method clone(), so that your class Data is able to implement the value semantics:

class ValueImpl
{
public:
    virtual ~ValueImpl() {};
    virtual std::unique_ptr<ValueImpl> clone() const { return new ValueImpl(); }
    virtual void increments() {}
    virtual void print() const { std::cout << "VoidValue "; }
};

class Value
{
private:
    ValueImpl * p_; // The underlying pointer

public:
    // Default constructor, allocating a "void" value
    Value() : p_(new ValueImpl) {}

    // Construct a Value given an actual implementation:
    // This allocates memory on the heap, hidden in clone()
    // This memory is automatically deallocated by unique_ptr
    Value(const ValueImpl & derived) : p_(derived.clone()) {}

    // Destruct the data (unique_ptr automatically deallocates the memory)
    ~Value() {}

    // Copy constructor and assignment operator:
    // Implements a value semantics by allocating new memory 
    Value(const Value & other) : p_(other.p_->clone()) {}
    Value & operator=(const Value & other) 
    {
        if(&other != this)
        {
            p_ = std::move(other.p_->clone());
        }
        return *this;
    }

    // Custom "polymorphic" methods
    void increments() { p_->increments(); }
    void print()      { p_->print(); }
};

The contained pointer is stored inside a C++11 std::unique_ptr<ValueImpl> to ensure the memory is released when destroyed or assigned a new value.

The derived implementations can finally be defined the following way:

class IntValue : public ValueImpl
{
public:
    IntValue(int k) : k_(k) {}
    std::unique_ptr<IntValue> clone() const
    {
        return std::unique_ptr<IntValue>(new IntValue(k_)); 
    }
    void increments() { k_++; }
    void print() const { std::cout << "Int(" << k_ << ") "; }

private:
    int k_;
};

class DoubleValue : public ValueImpl
{
public:
    DoubleValue(double x) : x_(x) {}
    std::unique_ptr<DoubleValue> clone() const
    {
        return std::unique_ptr<DoubleValue>(new DoubleValue(k_)); 
    }
    void increments() { x_ += 1.0; }
    void print() const { std::cout << "Double(" << x_ << ") "; }

private:
    int x_;
};

Which is enough to make the code snippet in the question works without any modification. This provides run-time polymorphism with value semantics, instead of the traditional run-time polymorphism with pointer semantics provided built-in by the C++ language. In fact, the concept of polymorphism (handling generic objects that behave differently according to their true "type") is independent from the concept of pointers (being able to share memory and optimize function calls by using the address of an object), and IMHO it is more for implementation details that polymorphism is only provided via pointers in C++. The code above is a work-around to take advantage of polymorphism when using pointers is not "philosophically required", and hence ease memory management.

Note: Thanks for CaptainObvlious for the contribution and his evolved code available here that I partially integrated. Not integrated are:

To ease the creation of derived implementations, you may want to create an intermediate templated class
You may prefer to use an abstract interface instead of my non-abstract base class

Inheritance-free polymorphism

Your basic idea (do not require inheritance) is good. I would recommend using Adobe.Poly instead. When you use 1 std::function per single operation you have N sort-of virtual table (pointers), plus potentially N heap allocations (depending if SBO (Small Buffer Optimization) can be applied or not).

You are also very likely to get into object life-time management problems. In your implementation you assume that the real object lives longer than the "interface". Sooner or later you will get it wrong. This is why I would encourage a value-semantic approach. Adobe.Poly gives you that.

With Adobe.Poly you only get one vtable (pointer). It also implements SBO: potentially not a single allocation.

I would not necessarily go with Boost.TypeErasure. It requires learning another "language" for specifying interfaces, which exploits lots of meta programming, and as of today it does not implement SBO.

Adobe.Poly is not well documented. See this post for examples of how you use it. Also, see this paper on how it is implemented.

std::variant vs pointer to base class for heterogeneous containers in C++

std::variant<A,B,C> holds one of a closed set of types. You can check whether it holds a given type with std::holds_alternative, or use std::visit to pass a visitor object with an overloaded operator(). There is likely no dynamic memory allocation, however, it is hard to extend: the class with the std::variant and any visitor classes will need to know the list of possible types.

On the other hand, BaseClass* holds an unbounded set of derived class types. You ought to be holding std::unique_ptr<BaseClass> or std::shared_ptr<BaseClass> to avoid the potential for memory leaks. To determine whether an instance of a specific type is stored, you must use dynamic_cast or a virtual function. This option requires dynamic memory allocation, but if all processing is via virtual functions, then the code that holds the container does not need to know the full list of types that could be stored.

Ad Hoc Polymorphism and Heterogeneous Containers with Value Semantics