Template declaration problem?

[Bruce Eckel]

I tried a number of different attempts at forward declaration, and all were
rejected by 1.1. Can you give me a hint?

In the meantime, I've cleaned up the program significantly. It now compiles
and runs both with KCC and VC++ (latest). Here it is:

//: C20:Ring.cpp
// Making a "ring" data structure from the STL
#include <iostream>
#include <list>
#include <string>
using namespace std;

template<class T>
class Ring {
  list<T> lst;
  friend class iterator;
public:
  class iterator: public std::iterator<
    std::bidirectional_iterator_tag,T,ptrdiff_t>{
    list<T>::iterator it;
    list<T>* r;
  public:
    iterator(list<T>& lst,
      const list<T>::iterator& i)
      : r(&lst), it(i) {}
    bool operator==(const iterator& x) const

      return it == x.it;
    }
    bool operator!=(const iterator& x) const

      return !(*this == x);
    }
    list<T>::reference operator*() const

      return *it;
    }
    iterator& operator++() {
      ++it;
      if(it == r->end())
        it = r->begin();
      return *this;
    }
    iterator operator++(int) {
      iterator tmp = *this;
      ++*this;
      return tmp;
    }
    iterator& operator--() {
      if(it == r->begin())
        it = r->end();
      --it;
      return *this;
    }
    iterator operator--(int) {
      iterator tmp = *this;
      --*this;
      return tmp;
    }
    iterator insert(const T& x){
      return iterator(*r, r->insert(it, x));
    }
    iterator erase() {
      return iterator(*r, r->erase(it));
    }
  };
  void push_back(const T& x) {
    lst.push_back(x);
  }
  iterator begin()

    return iterator(lst, lst.begin());
  }
  int size() { return lst.size(); }
};

int main() {
  Ring<string> rs;
  rs.push_back("one");
  rs.push_back("two");
  rs.push_back("three");
  rs.push_back("four");
  rs.push_back("five");
  Ring<string>::iterator it = rs.begin();
  it++; it++;
  it.insert("six");
  it = rs.begin();
  // Twice around the ring:
  for(int i = 0; i < rs.size() * 2; i++)
    cout << *it++ << endl;
} ///:~

================================
Bruce Eckel     http://www.BruceEckel.com
Contains free electronic books: "Thinking in Java" and "Thinking in C++ 2e"


-----Original Message-----
From: Alexandre Oliva <oliva@dcc.unicamp.br>
To: Bruce Eckel <Eckel@CrestedButte.net>
Cc: egcs-bugs@cygnus.com <egcs-bugs@cygnus.com>; egcs-patches@cygnus.com
<egcs-patches@cygnus.com>; jason@cygnus.com <jason@cygnus.com>
Date: Tuesday, September 15, 1998 5:04 PM
Subject: Re: Template declaration problem?


>Bruce Eckel <Eckel@CrestedButte.net> writes:
>
>> This compiles and runs with KCC, but has some interesting problems
>> compiling with egcs 1.1:
>
>In fact, there are a few bugs in your code that prevent egcs from
>compiling it correctly.  First, when you declare:
>
>> template<class T> class Ring {
>>   friend class iterator;
>
>Unless Ring<T>::iterator is already declared, this declaration refers
>to a class `iterator' in the outer namespace scope, not a nested
>class.  So you have to insert a forward declaration of nested class
>`iterator' before the friend declaration.
>
>The lack of this declaration causes a name lookup problem in egcs, but
>class Ring::iterator won't be granted friendship with the declaration
>as it is.
>
>>   class iterator : public list<T>::iterator {
>[snip]
>>       return node == x.node;
>
>Note that `node' is not a standardized member of list<T>::iterator;
>although it works with the STL distributed with egcs 1.1, it does not
>work with the one in the latest snapshot of egcs!
>
>>   iterator insert (iterator position,const T& x){
>>     return iterator(r.insert(position, x));
>>   }
>
>The only constructor of class Ring<T>::iterator takes more than one
>argument; isn't the first argument missing?  Same problem in method
>erase.
>
>
>The name lookup problem I refer to only affects friend declarations
>that introduce new names in outer namespaces; if the class iterator
>had been forward-declared outside the template definition, the correct
>class name would be looked up in the member functions of Ring<T>.
>
>Jason, Mark, any idea about how to fix this problem?  May I install
>a preprocessed version of the attached test case in g++.pt?
>
>--
>Alexandre Oliva
> mailto:oliva@dcc.unicamp.br mailto:aoliva@acm.org
> http://www.dcc.unicamp.br/~oliva
>Universidade Estadual de Campinas, SP, Brasil
>