C++ STL: How to iterate vector while requiring access to element and its index?

Question

I frequently find myself requiring to iterate over STL vectors. While I am doing this I require access to both the vector element and its index.

I used to do this as:

typedef std::vector<Foo> FooVec;
typedef FooVec::iterator FooVecIter;

FooVec fooVec;
int index = 0;
for (FooVecIter i = fooVec.begin(); i != fooVec.end(); ++i, ++index)
{
    Foo& foo = *i;
    if (foo.somethingIsTrue()) // True for most elements
        std::cout << index << ": " << foo << std::endl;
}

After discovering BOOST_FOREACH, I shortened this to:

typedef std::vector<Foo> FooVec;

FooVec fooVec;
int index = -1;
BOOST_FOREACH( Foo& foo, fooVec )
{
    ++index;
    if (foo.somethingIsTrue()) // True for most elements
        std::cout << index << ": " << foo << std::endl;
}

Is there a better or more elegant way to iterate over STL vectors when both reference to the vector element and its index is required?

I am aware of the alternative: for (int i = 0; i < fooVec.size(); ++i) But I keep reading about how it is not a good practice to iterate over STL containers like this.

Answer 1

for (size_t i = 0; i < vec.size(); i++)
    elem = vec[i];

Vectors are a thin wrapper over C arrays; whether you use iterators or indexes, it's just as fast. Other data structures are not so forgiving though, for example std::list.

Answer 2

You can always compute the index in the loop:

std::size_t index = std::distance(fooVec.begin(), i);

For a vector, this is quite likely to be implemented as a single pointer subtraction operation, so it's not particularly expensive.

Answer 3

Elegance is in the eye of the beholder, however do remember pointer/iterator arithmetics :)

for (FooVecIter i = fooVec.begin(); i != fooVec.end(); ++i)
{
    Foo& foo = *i;
    if (foo.somethingIsTrue()) // True for most elements
        std::cout << i - fooVec.begin() << ": " << foo << std::endl;
}

The up-side compared to the distance method is that you won't mistakenly do this for a non-random_access_iterator, so you'll always be in O(1).

Answer 4

For the specific question:

Is there a better or more elegant way to iterate over STL vectors
when both reference to the vector element and its index is required?

IMHO,

for( size_t i = 0; i < fooVec.size(); ++i ) {
    Foo & foo = fooVec[ i ];
}

is the most simple and elegant solution. Per the requirements of the question, using an iterator is not required.

I know iterators are there, they are useful, and using them is the preferred approach. But, we also have to put the solution after the problem (as in putting the carriage after the horse, nor the other way). I don't see how something becomes elegant just by using iterator, at least for this problem. But, I am not an expert, so I look forward what others have to say.