home bbs files messages ]

Forums before death by AOL, social media and spammers... "We can't have nice things"

   comp.lang.c++.moderated      Moderated discussion of C++ superhackery      33,346 messages   

[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]

   Message 32,168 of 33,346   
   =?UTF-8?B?RGFuaWVsIEtyw7xnbGVy?= to All   
   Re: specializing swap for nested iterato   
   21 Apr 12 13:27:13   
   
   From: daniel.kruegler@googlemail.com   
      
   Am 21.04.2012 09:51, schrieb galexander2@nc.rr.com:   
   > I'm developing my own container class and have implemented the   
   > container's iterator as a nested class.  I want to include a   
   > specialized version of swap for the iterator.  Here's a small   
   > example of what I tried:   
   >   
   >      #include   
   >      #include   
   >      #include   
   >      #include   
   >      #include   
   >   
   >      namespace ns {   
   >        template, class   
   > Allocator = std::allocator  >   
   >        class myContainer {   
   >          public:   
   >            class iterator_implementation {   
   >              public:   
   >                inline iterator_implementation() {   
   >                }   
   >            };   
   >          public:   
   >            typedef iterator_implementation iterator;   
   >        };   
   >        template   
   >        inline void swap(typename myContainer Allocator>::iterator&  x,   
   >                         typename myContainer Allocator>::iterator&  y) {   
   >          std::cout<<  "**** specialized iterator swap ****"<<  std::endl;   
   >        }   
   >      }   
      
   The problem is located here in the signature of your function   
   template: A parameter like   
      
   typename myContainer::iterator   
      
   is always very suspicious unless designed carefully: It does not allow   
   by-argument deduction of the template parameters. I did not check the   
   FAQ, but this should actually belong to the FAQ list: A   
   type-construction like   
      
   typename /some_template/::/some_type/   
      
   where some_arguments needs to be deduced is always an undeduced   
   context, because in general the compiler cannot find the set of   
   some_arguments given a dependent type some_type: This is like finding   
   an inverse function: To find the inverse of a mathematical function   
   requires that this function is invertible (i.e. there are some   
   constraints imposed), which is a unique inverse relation. C++   
   templates are not (in general) invertible type/value functions.   
      
   >      int main(int argc, char* argv[]) {   
   >        ns::myContainer::iterator x =   
   ns::myContainer::iterator();   
   >        ns::myContainer::iterator y =   
   ns::myContainer::iterator();   
   >        using std::swap;   
   >        std::cout<<  "before swap(x, y)"<<  std::endl;   
   >        swap(x, y);   
   >        std::cout<<  "after swap(x, y)"<<  std::endl;   
   >        std::cout<<  std::endl;   
   >        std::cout<<  "before swap, std::allocator   
   >> (x, y)"<<  std::endl;   
   >        swap, std::allocator  >(x, y);   
   >        std::cout<<  "after swap, std::allocator   
   >> (x, y)"<<  std::endl;   
   >        exit(0);   
   >      }   
   >   
   > The problem is that swap(x, y) calls std::swap rather than the   
   > specialized ns:swap.  It seems that ADL isn't finding ns:swap. To   
   > call ns:swap, one has to specify template arguments as in for   
   > example swap, std::allocator >(x, y), which   
   > isn't what I want.  Is this correct behavior according to the C++   
   > standard?   
      
   Yes, that is to be expected: You swap overload can never be called by   
   argument deduction.   
      
   > Or is it a compiler bug? If this is correct standard behavior, can   
   > someone point to where this behavior is described?   
      
   This is described in 14.8.2.5 [temp.deduct.type] p4-6, in particular   
   see p5:   
      
   "The non-deduced contexts are:   
   — The nested-name-specifier of a type that was specified using a   
   qualified-id."   
      
   combined with p6:   
      
   "When a type name is specified in a way that includes a non-deduced   
   context, all of the types that comprise that type name are also   
   non-deduced."   
      
   In your example "myContainer" is the above   
   mentioned nested-name-specifier and the referred to type is   
   "myContainer::iterator".   
      
   > In this case is there a standard way to define a specialize swap for   
   > iterator short of making the iterator_implementation class its own   
   > stand-alone class rather than a nested class?   
      
   The most simple one is to add a defining in-class friend function to   
   your iterator type such as:   
      
         namespace ns {   
           template, class   
   Allocator = std::allocator >   
           class myContainer {   
             public:   
               class iterator_implementation {   
                 public:   
                   inline iterator_implementation() {   
                   }   
      
                   friend void swap(iterator_implementation& x,   
                                    iterator_implementation& y) {   
                     std::cout << "**** specialized iterator swap ****" <<   
   std::endl;   
                   }   
               };   
             public:   
               typedef iterator_implementation iterator;   
           };   
         }   
      
   HTH & Greetings from Bremen,   
      
   Daniel Krügler   
      
      
      
   --   
         [ See http://www.gotw.ca/resources/clcm.htm for info about ]   
         [ comp.lang.c++.moderated.    First time posters: Do this! ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)   

[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]


(c) 1994,  bbs@darkrealms.ca