Question

I recently learned about the right way to work with reverse iterators in C++ (specifically when you need to erase one). (See this question and this one.)

This is how you’re supposed to do it:

typedef std::vector<int> IV;
for (IV::reverse_iterator rit = iv.rbegin(), rend = iv.rend();
     rit != rend; ++rit)
{
  // Use 'rit' if a reverse_iterator is good enough, e.g.,
  *rit += 10;
  // Use (rit + 1).base() if you need a regular iterator e.g.,
  iv.erase((rit + 1).base());
}

But I think thought this is much better (Don’t do this, not standards compliant, as MooingDuck points out):

for (IV::iterator it = iv.end(), begin = iv.begin();
     it-- != begin; )
{
  // Use 'it' for anything you want
  *it += 10;
  iv.erase(it);
}

Cons:

• You tell me. What’s wrong with it?
• It’s not standards compliant, as MooingDuck points out. That pretty much overrules any of the possible advantages below.

Pros:

• Uses a familiar idiom for reverse for-loops
• Don’t have to remember (or explain) the +1
• Less typing
• Works for std::list too: it = il.erase(it);
• If you erase an element, you don’t have to adjust the iterator
• If you erase, you don’t have to recompute the begin iterator

Answer 1

The reason for reverse iterators is that the standard algorithms do not know how to iterate over a collection backwards. For example:

#include <string>
#include <algorithm>
std::wstring foo(L"This is a test, with two letter a's involved.");
std::find(foo.begin(), foo.end(), L'a'); // Returns an iterator pointing
                                        // to the first a character.
std::find(foo.rbegin(), foo.rend(), L'a').base()-1; //Returns an iterator
                                                 // pointing to the last A.
std::find(foo.end(), foo.begin(), L'a'); //WRONG!! (Buffer overrun)

Use whichever iterator results in clearer code.