Counting number of distinct integers in array - c++

Question

To find the number of distinct numbers in an array from the lth to the rth index, I wrote a code block like:
int a[1000000];
//statements to input n number of terms from user in a.. along with l and r
int count=r-l+1; //assuming all numbers to be distinct
for(; l<=r; l++){
for(int i=l+1; i<=r; i++){
if(a[l]==a[i]){
count--;
break;
}
}
}
cout<<count<<'\n';
Explanation
For an array say, a=5 6 1 1 3 2 5 7 1 2 of ten elements. If we wish to check the number of distinct numbers between a[1] and a[8] that is the second and the 9th elements (including both), The logic I tried to implement would first take count=8 (no. of elements to be considered) and then it starts from a[1] that is 6 and checks for any other 6 after it, if it does find, it decreases the count by one and goes on for the next number in the row. So that if there are any more occurrence of 6 after that one, it would not be included twice.
Problem I tried small test cases and it works. But when I tried with bigger data, it did not work, so I wanted to know where would my logic fail?
Bigger data, as in integrated with other parts of the program and then used. Which gave incorrect output

Answer 1

You can try to use std::set
Basic idea is to add all the elements into your new set, and just output the size of your set.
#include <iostream>
#include <vector>
#include <set>
using namespace std;
int main()
{
int l = 1, r = 6;
int arr[] = {1, 1, 2, 3, 4, 5, 5, 5, 5};
set<int> s(&arr[l], &arr[r + 1]);
cout << s.size() << endl;
return 0;
}

Answer 2

Here is an answer that does not use std::set, although that solution is probably simpler.
#include <algorithm>
#include <vector>
int main()
{
int input[10]{5, 6, 1, 1, 3, 2, 5, 7, 1, 2}; //because you like raw arrays, I guess?
std::vector<int> result(std::cbegin(input), std::cend(input)); //result now contains all of input
std::sort(std::begin(result), std::end(result)); //result now holds 1 1 1 2 2 3 5 5 6 7
result.erase(std::unique(std::begin(result), std::end(result)), std::end(result)); //result now holds 1 2 3 5 6 7
result.size(); //gives the count of distinct integers in the given array
}
Here it is live on Coliru if you're into that.
--
EDIT: Here, have a short version of the set solution, too.
#include <set>
int main()
{
int input[10]{5, 6, 1, 1, 3, 2, 5, 7, 1, 2}; //because you like raw arrays, I guess?
std::set<int> result(std::cbegin(input), std::cend(input));
result.size();
}

Answer 3

The first question to ask with this type of problem is what is the possible range of the values. if the range of numbers N is "reasonably small", then you can use a boolean array of size N to indicate whether the number corresponding to the index is present. You iterate from l to r, setting the flag, and if the flag was not already set increment a counter.
count = 0;
for(int i=l; i<=r; i++) {
if (! isthere[arr[i]]) {
count++;
isthere[arr[i]] = TRUE;
}
}
In terms of complexity, both this approach and the one based on set are O(n), but this one is faster as there is no hashing involved. For small N, for example for numbers between 0-255, most likely this is also likely to be less memory intensive. For larger N, for example if any 32-bit integers is allowed, the set based approach is more suitable.

Answer 4

You said you didn't mind another solution. So here it is. It uses set - a structure that stores only unique elements. By the way, on the bigger data - it will much faster than solution with two cycles.
set<int> a1;
for (int i = l; i <= r; i++)
{
a1.insert(a[i]);
}
cout << a1.size();

Answer 5

In the below process I'm giving process of counting unique numbers. In this technique you just get unique elements in an array. this process will update your array with garbage value. So in this process you can't use this array (that we will use) further anymore. This array will automatically resize with distinct elements.
#include <stdio.h>
#include <iostream>
#include <algorithm> // for using unique (library function)
int main(){
int arr[] = {1, 1, 2, 2, 3, 3};
int len = sizeof(arr)/sizeof(*arr); // finding size of arr (array)
int unique_sz = std:: unique(arr, arr + len)-arr; // Counting unique elements in arr (Array).
std:: cout << unique_sz << '\n'; // Printing number of unique elements in this array.
return 0;
}
If you want to handle that problem (That I told before), you can follow this process. You can handle this by coping your array in another array.
#include <stdio.h>
#include <iostream>
#include <algorithm> // for using copy & unique (library functions)
#include <string.h> // for using memcpy (library function)
int main(){
int arr[] = {1, 1, 2, 2, 3, 3};
int brr[100]; // we will copy arr (Array) to brr (Array)
int len = sizeof(arr)/sizeof(*arr); // finding size of arr (array)
std:: copy(arr, arr+len, brr); // which will work on C++ only (you have to use #include <algorithm>
memcpy(brr, arr, len*(sizeof(int))); // which will work on C only
int unique_sz = std:: unique(arr, arr+len)-arr; // Counting unique elements in arr (Array).
std:: cout << unique_sz << '\n'; // Printing number of unique elements in this array.
for(int i=0; i<len; i++){ // Here is your old array, that we store to brr (Array) from arr (Array).
std:: cout << brr[i] << " ";
}
return 0;
}

Answer 6

Personally, I'd just use standard algorithms
#include<algorithm>
#include <iostream>
int main()
{
int arr[] = {1, 1, 2, 3, 4, 5, 5, 5, 5};
int *end = arr + sizeof(arr)/sizeof(*arr);
std::sort(arr, end);
int *p = std::unique(arr, end);
std::cout << (int)(p - arr) << '\n';
}
This obviously relies on being allowed to modify the array (any duplicates are moved to the end of arr). But it is easy to create a copy of an array if needed and work on the copy.

Answer 7

TL;DR: Use this:
template<typename InputIt>
std::size_t countUniqueElements(InputIt first, InputIt last) {
using value_t = typename std::iterator_traits<InputIt>::value_type;
return std::unordered_set<value_t>(first, last).size();
}
There are two approaches:
Insert everything into a set, count the set. Because you don't care about the order you can use a std::unordered_set which will be faster than std::set. std::set is implemented as a tree which does a lot of allocations so it can be slow.
Use std::sort. If you want to preserve the original array you'll need to make a copy of it.
Here is a complete example.
#include <algorithm>
#include <cstdint>
#include <vector>
#include <unordered_set>
#include <iostream>
template<typename RandomIt>
std::size_t countUniqueElementsSort(RandomIt first, RandomIt last) {
if (first == last)
return 0;
std::sort(first, last);
std::size_t count = 1;
auto val = *first;
while (++first != last) {
if (*first != val) {
++count;
}
val = *first;
}
return count;
}
template<typename InputIt>
std::size_t countUniqueElementsSet(InputIt first, InputIt last) {
using value_t = typename std::iterator_traits<InputIt>::value_type;
return std::unordered_set<value_t>(first, last).size();
}
int main() {
std::vector<int> v = {1, 3, 4, 4, 3, 6};
std::cout << countUniqueElementsSet(v.begin(), v.end()) << "\n";
std::cout << countUniqueElementsSort(v.begin(), v.end()) << "\n";
int v2[] = {1, 3, 4, 4, 3, 6};
std::cout << countUniqueElementsSet(v2, v2 + 6) << "\n";
std::cout << countUniqueElementsSort(v2, v2 + 6) << "\n";
}
Using that loop in the sort version should be faster than std::unique.
The complexity of 2. is worse than 1. - the average case is O(N) vs O(N log N). But it avoids allocation so may end up being faster for small arrays or ones that are already sorted or mostly already sorted.
You should definitely not use std::set, and probably not use std::unique (though it does lead to fewer lines of code, and won't make that much difference to performance so up to you).
In any case, in most cases you should go with the set version - it's a lot simpler simpler and should be faster in almost all cases.
As other people have mentioned, if you know your input domain is small you can use a bool array instead of an unordered_set.