...one of the most highly
regarded and expertly designed C++ library projects in the
world.
— Herb Sutter and Andrei
Alexandrescu, C++
Coding Standards
The documentation is divided into an explanation for each container. When containers have the same interface, that common interface is explained only once, but links are always provided to more relevant information. Please make sure you understand the Clonable concept and the Clone Allocator concept.
Refinement of
The Clonable concept is introduced to formalize the requirements for copying heap-allocated objects. A type T might be Clonable even though it is not Assignable or Copy Constructible. Notice that many operations on the containers do not even require the stored type to be Clonable.
Notation
Type | Object (const or non-const) | Pointer | Describes |
T | a | ptr | A Clonable type |
Valid expressions
Expression | Type | Semantics | Postcondition |
new_clone(a); | T* | Allocate a new object that can be considered equivalent to the a object | typeid(*new_clone(a)) == typeid(a) |
delete_clone(ptr); | void | Deallocate an object previously allocated with allocate_clone(). Must not throw |
In the <boost/ptr_container/clone_allocator.hpp> header a default implementation of the two functions is given:
namespace boost { template< class T > inline T* new_clone( const T& t ) { return new T( t ); } template< class T > void delete_clone( const T* t ) { checked_delete( t ); } }
Notice that this implementation makes normal Copy Constructible classes automatically Clonable unless operator new() or operator delete() are hidden.
The two functions represent a layer of indirection which is necessary to support classes that are not Copy Constructible by default. Notice that the implementation relies on argument-dependent lookup (ADL) to find the right version of new_clone() and delete_clone(). This means that one does not need to overload or specialize the function in the boost namespace, but it can be placed together with the rest of the interface of the class. If you are implementing a class inline in headers, remember to forward declare the functions.
Warning: We are considering the removal of default implementation above. Therefore always make sure that you overload the functions for your types and do not rely on the defaults in any way.
The Clone Allocator concept is introduced to formalize the way pointer containers control memory of the stored objects (and not the pointers to the stored objects). The clone allocator allows users to apply custom allocators/deallocators for the cloned objects.
More information can be found below:
Notation
Type | Object (const or non-const) | Describes |
T | a | A type |
T* | ptr | A pointer to T |
Valid expressions
Expression | Type | Semantics | Postcondition |
CloneAllocator::allocate_clone(a); | T* | Allocate a new object that can be considered equivalent to the a object | typeid(*CloneAllocator::allocate_clone(a)) == typeid(a) |
CloneAllocator::deallocate_clone(ptr); | void | Deallocate an object previously allocated with CloneAllocator::allocate_clone() or a compatible allocator. Must not throw. |
The library comes with two predefined clone allocators.
This is the default clone allocator used by all pointer containers. For most purposes you will never have to change this default.
Definition
namespace boost { struct heap_clone_allocator { template< class U > static U* allocate_clone( const U& r ) { return new_clone( r ); } template< class U > static void deallocate_clone( const U* r ) { delete_clone( r ); } }; }
Notice that the above definition allows you to support custom allocation schemes by relying on new_clone() and delete_clone().
This class provides a way to remove ownership properties of the pointer containers. As its name implies, this means that you can instead use the pointer containers as a view into an existing container.
Definition
namespace boost { struct view_clone_allocator { template< class U > static U* allocate_clone( const U& r ) { return const_cast<U*>(&r); } template< class U > static void deallocate_clone( const U* ) { // empty } }; }
The library consists of the following types of classes:
The pointer container adapters are used when you want to make a pointer container starting from your own "normal" container. For example, you might have a map class that extends std::map in some way; the adapter class then allows you to use your map class as a basis for a new pointer container.
The library provides an adapter for each type of standard container highlighted as links below:
The pointer containers of this library are all built using the adapters. There is a pointer container for each type of "normal" standard container highlighted as links below.
As of version 1.34.0 of Boost, the library supports serialization via Boost.Serialization.
Of course, for serialization to work it is required that the stored type itself is serializable. For maps, both the key type and the mapped type must be serializable.
When dealing with serialization (and serialization of polymophic objects in particular), pay special attention to these parts of Boost.Serialization:
Output/saving requires a const-reference:
// // serialization helper: we can't save a non-const object // template< class T > inline T const& as_const( T const& r ) { return r; } ... Container cont; std::ofstream ofs("filename"); boost::archive::text_oarchive oa(ofs); oa << as_const(cont);
See Compile time trap when saving a non-const value for details.
Derived classes need to call base_object() function:
struct Derived : Base { template< class Archive > void serialize( Archive& ar, const unsigned int version ) { ar & boost::serialization::base_object<Base>( *this ); ... } };
For details, see Derived Classes.
You need to use BOOST_CLASS_EXPORT to register the derived classes in your class hierarchy:
BOOST_CLASS_EXPORT( Derived )
See Export Key and Object Tracking for details.
Remember these three issues and it might save you some trouble.
The purpose of the class is simply to tell the containers that null values should be allowed. Its definition is trivial:
namespace boost { template< class T > struct nullable { typedef T type; }; }
Please notice that nullable has no effect on the containers interface (except for is_null() functions). For example, it does not make sense to do
boost::ptr_vector< boost::nullable<T> > vec; vec.push_back( 0 ); // ok vec.push_back( new boost::nullable<T> ); // no no! boost::nullable<T>& ref = vec[0]; // also no no!
There are three exceptions that are thrown by this library. The exception hierarchy looks as follows:
namespace boost { class bad_ptr_container_operation : public std::exception { public: bad_ptr_container_operation( const char* what ); }; class bad_index : public bad_ptr_container_operation { public: bad_index( const char* what ); }; class bad_pointer : public bad_ptr_container_operation { public: bad_pointer(); bad_pointer( const char* what ); }; }
As of version 1.34.0 of Boost, the library allows you to disable exceptions completely. This means the library is more fit for domains where exceptions are not used. Furthermore, it also speeds up a operations a little. Instead of throwing an exception, the library simply calls BOOST_ASSERT.
To disable exceptions, simply define this macro before including any header:
#define BOOST_PTR_CONTAINER_NO_EXCEPTIONS 1 #include <boost/ptr_container/ptr_vector.hpp>
It is, however, recommended that you define the macro on the command-line, so you are absolutely certain that all headers are compiled the same way. Otherwise you might end up breaking the One Definition Rule.
If BOOST_NO_EXCEPTIONS is defined, then BOOST_PTR_CONTAINER_NO_EXCEPTIONS is also defined.
Navigate:
Copyright: | Thorsten Ottosen 2004-2007. Use, modification and distribution is subject to the Boost Software License, Version 1.0 (see LICENSE_1_0.txt). |
---|