Sorting a vector of (double precision) reals and o

Question

In C++ would like to sort a lengthy (2^20) vector of reals, obviously sort() does the trick. Having used R before I was used to the nice order() function which yields the permutation that leads to the sorted vector.

Example:

x = {24, 55, 22, 1}

Then the permutation

perm = {3, 2, 0, 1}

Maps the original x to the sorted x in ascending order.

I can probably implement some bubble sort which does not only sort x but performs the same transpositions on the vector {0,1,2,...} and outputs both, but I believe someone must have thought about it and especially have done it efficiently.

Answer 1

You can use std::sort to sort the list of pairs {(24, 0), (55, 2), (22, 0), (1, 1)}. It isn't particularly pretty, but I usually do something like this:

#include <vector>
#include <algorithm>
#include <utility>

typedef std::pair<double, int> Pair;

struct CmpPair
{
    bool operator()(const Pair& a, const Pair& b)
    { return a.first < b.first; }
};

void sortingPermutation(
    const std::vector<double>& values,
    std::vector<int>& permutation)
{
    std::vector<Pair> pairs;
    for (int i = 0; i < (int)values.size(); i++)
        pairs.push_back(Pair(values[i], i));

    std::sort(pairs.begin(), pairs.end(), CmpPair());

    typedef std::vector<Pair>::const_iterator I;
    for (I p = pairs.begin(); p != pairs.end(); ++p)
        permutation.push_back(p->second);
}

Here is the test:

#include <iostream>

int main()
{
    std::vector<double> values;
    values.push_back(24);
    values.push_back(55);
    values.push_back(22);
    values.push_back(1);

    std::vector<int> permutation;
    sortingPermutation(values, permutation);

    typedef std::vector<int>::const_iterator I;
    for (I p = permutation.begin(); p != permutation.end(); ++p)
        std::cout << *p << " ";
    std::cout << "\n";
}

Answer 2

Edit

Better than before approach without using helper vectors: (source on ideone):

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

template<class Vals>
void sortingPermutation(const Vals& values, std::vector<int>& v){
  int size = values.size(); 
  v.clear(); v.reserve(size);
  for(int i=0; i < size; ++i)
    v.push_back(i);

  std::sort(v.begin(), v.end(), [&values](int a, int b) -> bool { 
    return values[a] < values[b];
  });
}

int main()
{
    std::vector<double> values;
    values.push_back(24);
    values.push_back(55);
    values.push_back(22);
    values.push_back(1);

    std::vector<int> permutation;
    sortingPermutation(values, permutation);

    typedef std::vector<int>::const_iterator I;
    for (I p = permutation.begin(); p != permutation.end(); ++p)
        std::cout << *p << " ";
    std::cout << "\n";
}

I am using lambda from C++0x, but it can be replaced with simple functor object:

template<class T>
struct CmpPairs{
  CmpPairs(const std::vector<T> &v): v_(v) {}
  std::vector<T> v_;
  bool operator()(int a, int b){ return v_[a] < v_[b]; }
};

template<class T>
CmpPairs<T> CreateCmpPairs(const std::vector<T> & v) { return CmpPairs<T>(v); }
//in sortingPermutation:
std::sort(v.begin(), v.end(), CreateCmpPairs(values));

Source of old solution with std::map: ideone

Answer 3

I would say the best way would be to create a vector of ints 0..N and then sort that array with a comparison function that compares the corresponding elements of the vector you're trying to find the sorted permutation of. Something like:

#include <vector>
#include <algorithm>

template<class T> class sorter {
    const std::vector<T> &values;
public:
    sorter(const std::vector<T> &v) : values(v) {}
    bool operator()(int a, int b) { return values[a] < values[b]; }
};

template<class T> std::vector<int> order(const std::vector<T> &values)
{
    std::vector<int> rv(values.size());
    int idx = 0;
    for (std::vector<int>::iterator i = rv.begin(); i != rv.end(); i++)
        *i = idx++;
    std::sort(rv.begin(), rv.end(), sorter<T>(values));
    return rv;
}

This minimizes the allocation overhead, as we don't create any large temporary object that we sort and then extract the final permution -- the same vector that is being returned is the temp for sorting.