Count number of unique elements using thrust unique_by_key when the set of values is a tuple

Here is one possible approach:

As you've already pointed out, thrust::unique_by_key returns a pair of iterators. A pair in thrust is something like a tuple, and we can use the thrust tuple element access mechanism (thrust::get<...>) to retrieve individual members of the pair.

Therefore, if we retrieve the first element of the pair, this corresponds to the result end iterator for the keys supplied to the thrust::unique_by_key algorithm. We can subtract from this the begin iterator for the keys, to retrieve the length of the keys in the result (which is the same as the length of the values in the result).

Here is a worked example:

$ cat t390.cu
#include <thrust/device_vector.h>
#include <thrust/sort.h>
#include <thrust/execution_policy.h>
#include <thrust/unique.h>


int main()
{
    const int N = 8;

        thrust::device_vector<int> keys1(N);
        thrust::device_vector<int> keys2(N);
        thrust::device_vector<float> value_keys(N);

        keys1[0] = 1; keys1[1] = 1; keys1[2] = 2; keys1[3] = 2;
        keys1[4] = 2; keys1[5] = 3; keys1[6] = 3; keys1[7] = 3;


        keys2[0] = 4; keys2[1] = 1; keys2[2] = 7; keys2[3] = 2;
        keys2[4] = 6; keys2[5] = 8; keys2[6] = 3; keys2[7] = 5;

        value_keys[0] = -0.01; value_keys[1] = 1.1; value_keys[2] = -0.07; value_keys[3] = 2.1;
        value_keys[4] = 5.2; value_keys[5] = -0.08; value_keys[6] = 3.2; value_keys[7] = 5.1;


        auto end = thrust::unique_by_key(thrust::device, keys1.begin(), keys1.end(),
          thrust::make_zip_iterator(thrust::make_tuple(value_keys.begin(), keys2.begin())));
        int result_size = thrust::get<0>(end) - keys1.begin();

    std::cout << "Unique PAIRS:"<< std::endl;
    std::cout << "keys1, value_keys, keys2:" << std::endl;
    for (int i = 0; i < result_size; i++) {
        std::cout << keys1[i] <<'\t' << value_keys[i] <<'\t' << keys2[i] << std::endl;
    }

}
$ nvcc -o t390 t390.cu -std=c++11
$ ./t390
Unique PAIRS:
keys1, value_keys, keys2:
1       -0.01   4
2       -0.07   7
3       -0.08   8
$

If you don't want to use c++11 auto, then you could adapt the above example by defining the returned pair explicitly yourself. This is determined by looking at the two input iterator types that the thrust::unique_by_key algorithm is using, and creating a pair out of them. Here is an example:

$ cat t390.cu
#include <thrust/device_vector.h>
#include <thrust/sort.h>
#include <thrust/execution_policy.h>
#include <thrust/unique.h>

typedef thrust::zip_iterator<thrust::tuple<thrust::device_vector<float>::iterator, thrust::device_vector<int>::iterator> > my_iter;
typedef thrust::pair<thrust::device_vector<int>::iterator, my_iter> my_pair;


int main()
{
    const int N = 8;

        thrust::device_vector<int> keys1(N);
        thrust::device_vector<int> keys2(N);
        thrust::device_vector<float> value_keys(N);

        keys1[0] = 1; keys1[1] = 1; keys1[2] = 2; keys1[3] = 2;
        keys1[4] = 2; keys1[5] = 3; keys1[6] = 3; keys1[7] = 3;


        keys2[0] = 4; keys2[1] = 1; keys2[2] = 7; keys2[3] = 2;
        keys2[4] = 6; keys2[5] = 8; keys2[6] = 3; keys2[7] = 5;

        value_keys[0] = -0.01; value_keys[1] = 1.1; value_keys[2] = -0.07; value_keys[3] = 2.1;
        value_keys[4] = 5.2; value_keys[5] = -0.08; value_keys[6] = 3.2; value_keys[7] = 5.1;


        my_pair end = thrust::unique_by_key(thrust::device, keys1.begin(), keys1.end(),
          thrust::make_zip_iterator(thrust::make_tuple(value_keys.begin(), keys2.begin())));
        int result_size = thrust::get<0>(end) - keys1.begin();

    std::cout << "Unique PAIRS:"<< std::endl;
    std::cout << "keys1, value_keys, keys2:" << std::endl;
    for (int i = 0; i < result_size; i++) {
        std::cout << keys1[i] <<'\t' << value_keys[i] <<'\t' << keys2[i] << std::endl;
    }

}
$ nvcc -o t390 t390.cu
$ ./t390
Unique PAIRS:
keys1, value_keys, keys2:
1       -0.01   4
2       -0.07   7
3       -0.08   8
$