id,summary,reporter,owner,description,type,status,milestone,component,version,severity,resolution,keywords,cc
4277,Ease access to underlying singleton_pool,Henning.Stroeker@…,Chris Newbold,"It is not convenient to call ''singleton_pool<...>::release_memory()'' if you use
''pool_allocator'' or ''fast_pool_allocator''. The reason is that the exact singleton_pool type has to be specified:
{{{
typedef boost::pool_allocator<int> myAllocator;
std::vector<int, myAllocator> v;
...  //do something and empty v again
boost::singleton_pool<boost::pool_allocator_tag, sizeof(int)>::release_memory();
}}}
In fact the call to ''release_memory'' might need even more template arguments (because of user allocators etc). And: The developer has to know all template parameters. If the vector declaration changes the call to ''release_memory'' has to be changed too, otherwise it would still compile but have no effect. This is error-prone.
It would be much easier if ''pool_allocator'' and ''fast_pool_allocator'' would provide a typedef for the used singleton_pool. Then the above code would look like this:
{{{
typedef boost::pool_allocator<int> myAllocator;
std::vector<int, myAllocator> v;
...  //do something and empty v again
myAllocator::singleton_pool_type::release_memory();
}}}

The implementation in pool_alloc.hpp would look like this:

{{{

template <typename T,
    typename UserAllocator,
    typename Mutex,
    unsigned NextSize>
class pool_allocator
{
  public:
    typedef T value_type;
    typedef UserAllocator user_allocator;
    typedef Mutex mutex;

    ...

    typedef singleton_pool<pool_allocator_tag, sizeof(T), UserAllocator, Mutex, NextSize> singleton_pool_type;

...
}}}
",Feature Requests,closed,Boost 1.43.0,pool,Boost 1.44.0,Optimization,wontfix,,boostpool@…