How to write this for-loop using std::for_each or std::transform?

This is more of a learning question. Is there a way I can write the following for-loop using std::for_each or std::transform? If not, is there anything in boost that can help on this? The loop simply flattens a vector of vectors into one long vector.

vector<vector<int> > int_vectors;

// ... fill int_vectors

vector<int> ints;

for (vector<vector<int> >::const_iterator iter = int_vectors.begin(); iter != int_vectors.end(); ++iter) {
   ints.insert开发者_开发百科(ints.end(), iter->begin(), iter->end());
}

I wouldn't change this to use one of the algorithms unless you have a compiler that supports lambdas. It's completely clear as written. Even if your compiler did support lambdas, I'd probably not change this code.

One relatively straightforward option would be to write a flattening iterator. I wrote one for demonstration in an answer to another question.

If you really want a one-liner and can use bind (boost::bind from Boost, std::tr1::bind from TR1, and std::bind from C++0x will all work), then here is how that would look. I warn you in advance: it is horrible.

Edit: Technically this is also illegal. The type of a Standard Library member function is unspecified, so you cannot (portably or correctly) take the address of such a member function. If you could correctly take the address of a Standard Library member function, this is what it would look like:

typedef std::vector<int>::iterator (std::vector<int>::*IteratorGetter)();

std::for_each(int_vectors.begin(), int_vectors.end(),
    std::bind(
        std::bind(
            &std::vector<int>::insert<std::vector<int>::iterator>, 
            &ints, 
            std::bind((IteratorGetter)&std::vector<int>::end, &ints), 
            _1, 
            _2
        ),
        std::bind((IteratorGetter)&std::vector<int>::begin, _1),
        std::bind((IteratorGetter)&std::vector<int>::end, _1)
    )
);

(Yes, that is technically one "line of code" as it is a single statement. The only thing I have extracted is a typedef for the pointer-to-member-function type used to disambiguate the overloaded begin and end functions; you don't necessarily have to typedef this, but the code requires horizontal scrolling on Stack Overflow if I don't.)

If your compiler supports lambdas this is fairly simple. The typedefs help with readability.

typedef std::vector<int> VI;
typedef std::vector<VI> VVI;

VVI int_vectors;
VI ints;

std::for_each(int_vectors.begin(), int_vectors.end(), [&](VI &vec) {
    ints.insert(ints.end(), vec.begin(), vec.end());
});

I kno macros aren't great practice but you can do some clever things with them:

#include <algorithm>
#include <vector>
#include <iostream>
#include <iterator>

using namespace std;
#define all(v) (v).begin(), (v).end()
#define foreach(it, v) for(auto it = (v).begin(); it != (v).end(); ++it)

void main() {
    vector<int> vi;
    vector<vector<int>> vvi;

    foreach(it, vvi) {
        copy(all(*it), back_inserter(vi));
    }
}

You can also use std::copy and back_inserter to do what you're trying to achieve.