Re: [boost] Inconsistency between std::minmax and boost::minmax

Subject: Re: [boost] Inconsistency between std::minmax and boost::minmax
From: Eric Niebler (eric_at_[hidden])
Date: 2012-02-24 22:41:26

On 2/24/2012 4:58 PM, Marshall Clow wrote:
> On Feb 24, 2012, at 2:00 PM, Jeffrey Lee Hellrung, Jr. wrote:
> 
>> On Fri, Feb 24, 2012 at 1:52 PM, Marshall Clow
>> <mclow.lists_at_[hidden]>wrote:
>> 
>>> So, C++11 has been approved, and in the new standard library,
>>> there is:
>>> 
>>> template <typename T> std::pair<const T&, const T&> minmax (
>>> const T& a, const T& b )
>>> 
>>> but in Boost, we've got: template <typename T> tuple< T const&,
>>> T const& > minmax(T const& a, T const& b) {
>>> 
>>> I can easily change boost minmax to return a pair instead of a
>>> 2-element tuple, but that's an interface change, and might
>>> break existing code.
>>> 
>>> On the other hand, giving them the same interface will make it
>>> easier for people to move to C++11's library functions.
>>> 
>>> Any opinions? ideas?
>>> 
>> 
>> I could be wrong about this, but wouldn't that be problematic for
>> C++03 std::pair, which (typically) barfs when instantiated with
>> reference types?
> 
> Unfortunately, all I can find is
>> late after the review, as I finally scrounged to add the library
>> for a release, Eric Niebler noted the bad behavior of std::pair
>> for minmax and suggested to use Boost.tuple instead.
> 
> Does anyone have a link to  this discussion?

Not offhand, but I can tell you that my objection was precisely
because C++03 std::pair cannot hold references.

-- 
Eric Niebler
BoostPro Computing
http://www.boostpro.com