Re: Traversing multiple vectors "in parallel"

Raphael Simon wrote:


The important thing about an iterator is mainly that it provides operator++,
operator!= (and maybe operator==) and unary operator*, some algorithms
might need a nested value_type typedef and maybe some further code to make
iterator_traits work. Haven't played with the latter though, so I can't
tell you more.

Now:

struct my_iterator
{
   typedef std::vector<unsigned>::iterator iterator_base;
   my_iterator( iterator_base i1, iterator_base i2):
      m_i1(i1), m_i2(i2)
   {}

   unsigned operator*() const
   { return *m_i1+*m_i2; }
   void operator++() //TODO: correct returntype
   {
      ++m_i1;
      ++m_i2;
   }
   friend bool operator==(my_iterator const& i1, my_iterator const& i2)
   {
      bool b1 = i1.m_i1==i2.m_i1;
      bool b2 = i1.m_i2==i2.m_i2;
      // if this fires, iterators are shifted, i.e. the sequences
      // have different sizes.
      assert(b1==b2);
      return b1;
   }
   friend bool operator!=(my_iterator const& i1, my_iterator const& i2)
   { return !(i1==i2); }
private:
   iterator_base m_i1;
   iterator_base m_i2;
};

vector<unsigned> v1, v2;
....
for( my_iterator it(v2.begin(), v2.begin()), end(v1.end(), v2.end());
     it!=end;
     ++it)
{
    std::cout << "sum = " << *it << std::endl;
}

Now, Boost's zip_iterator rules. This here is basically the same, but much,
much more premitive.

Uli