[boost] Re: Boost memory management guidelines

From: David Abrahams (dave_at_[hidden])
Date: 2003-08-30 10:16:38

"Peter Dimov" <pdimov_at_[hidden]> writes:

> Howard Hinnant wrote:
>> On Saturday, August 30, 2003, at 9:31 AM, Peter Dimov wrote:
>>
>>> The stateful allocator problem is a thorny issue too, since to allow
>>> stateful allocators, one needs to specify whether the stored
>>> allocator is part of the logical state of the object. IE
>>>
>>> vector<T, A> v1(a1);
>>> vector<T, A> v2(a2);
>>>
>>> v1 = v2; // which allocator does v1 hold now?
>>
>> Just to throw in my .02:
>>
>> v1 holds the same allocator as it was constructed with.  Rationale:
>> Allocator state should stay with the memory it allocates so that it
>> can deallocate it when the time comes (much like the shared_ptr
>> philosophy).  Since vector assignment does not transfer ownership of
>> memory, it should not assign allocators.
>>
>> But consider:
>>
>> swap(v1, v2);
>>
>> Here memory ownership is swapped between the vectors.  Therefore,
>> imho, allocator state should also be swapped.
>
> This is all very reasonable; the allocator is not part of the logical state
> (merely an implementation detail) and therefore v1 and v2 can be equivalent
> (Assignable postcondition) even though they have different allocators.
>
> If I want to assign the allocator I'll have to use
>
> vector<T, A>(v2).swap(v1);
>
> On the other hand, the alternative seems reasonable, too. The allocator is
> part of the state (accessor is available), therefore after v1 = v2 v1 holds
> a2. If I want to retain the allocator I can use
>
> v1.assign(v2.begin(), v2.end());
>
> which works for any v2, not just the specific vector<T, A>.

And, FWIW, I find the latter much more reasonable, since as Peter
mentions, there is a get_allocator function.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com