Re: [boost] embedded_vector

Subject: Re: [boost] embedded_vector
From: Tim Blechmann (tim_at_[hidden])
Date: 2010-02-09 08:25:57

>>>>> [...] in contrary to std::vector, which is only required to
>>>>> support random-access iterators (which are implementation-defined), my
>>>>> class guaranties to allocate the data in one consecutive memory 
>>>>> area. [...]
>>>>>
>>>>
>>>> As of C++03, std::vector is contiguous, so I'm not convinced by the
>>>> utility of this.
>>>
>>> interesting ... it seems, my assumption was based on c++98
>>
>> I don't think that was changed from 98 to 03.
>> but your description reminded me of something I do think would be a 
>> useful addition to Boost.Array.

i cannot comment on c++03, since i don't have access to the iso
specification, but for c++0x it seems to be the case. n2800 tells me the
following:

23.2.6.1:
The elements of a vector are stored contiguously, meaning that if v is a
vector<T, Alloc> where T is some type other than bool, then it obeys the
identity &v[n] == &v[0] + n for all 0 <= n < v.size().

in addition to that, 23.2.6.3 defines:
pointer       data();
const_pointer data() const;

these two paragraphs would give me the guaranties, that i need for my
application.

>> a vector that resides on the stack as long as it fits in there(like a 
>> boost::array) but still supports dynamic allocation in case the vector 
>> size exceeds the reserved space.
>>
>> my implementation of this:
>>
>> template<class T,std::size_t EmbeddedSize,bool Expand=true>
>> class embedded_vector;
> 
> See also here for a lib in the review queue:
> 
> http://www.cs.aau.dk/~nesotto/boost/trunk/libs/auto_buffer/doc/html/index.html#auto_buffer.motivation

interesting ... for EmbeddedSize==0 or
StackBufferPolicy==boost::store_n_objects<0> both implementations would
match the guaranties of my class ...

cheers, tim

-- 
tim_at_[hidden]
http://tim.klingt.org
Silence is only frightening to people who are compulsively
verbalizing.
  William S. Burroughs