Re: [boost] [range] range_size metafunction deprecated?

From: shunsuke (pstade.mb_at_[hidden])
Date: 2007-03-01 06:11:08

Thorsten Ottosen wrote:
> shunsuke wrote:
>> Thorsten Ottosen wrote:
>>
>>> shunsuke wrote:
>>>
>>>> Hi,
>>>>
>>>> I noticed Range Library Core proposal
>>>> (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n2068.html)
>>>> removed 'range_size' metafunction.
>>>> It was replaced by 'range_difference'?
>>>> (But, cvs head is not affected.)
>>> Right. A proposal for the standard and boost is two different things.
>>
>> The 'boost::range_size' of 'iterator_range' is defined as 'std::size_t'.
>> Can it guarantee to represent any difference_type of iterator?
> 
> Probably not.
> 
> Anyway, I plan to change boost::size() to return iterator_difference.
> 
> (If there is one thing I hate about C++ containers it has to be the use
> of unsigned.)
> 
>>>> If so, why 'std::size' too is proposed?
>>>> 'std::distance' seems enough.
>>> So that you may have the guarantee that the function is O(1) complexity.
>>
>> 'std::distance'(24.3.4/1) also seems to guarantee it?
> 
> Only for random access iterators. std::size() would only accept random 
> access iterators.

Hmm, when should I use 'std::size'?
Whenever the argument is known to be RandomAccessRange,
the 'std::distance' is known to be O(1) complexity.
Thus, I will always use 'std::distance'.
(I actually replaced all the 'boost::size' in my codes
  by 'boost::distance'.)
Any pitfalls?

-- 
Shunsuke Sogame