Count subarrays divisible by K - c++

Question

Given a sequence of n positive integers we need to count consecutive sub-sequences whose sum is divisible by k.
Constraints : N is up to 10^6 and each element up to 10^9 and K is up to 100
EXAMPLE : Let N=5 and K=3 and array be 1 2 3 4 1
Here answer is 4
Explanation : there exists, 4 sub-sequences whose sum is divisible by 3, they are
3
1 2
1 2 3
2 3 4
My Attempt :
long long int count=0;
for(int i=0;i<n;i++){
long long int sum=0;
for(int j=i;j<n;j++)
{
sum=sum+arr[j];
if(sum%k==0)
{
count++;
}
}
}
But obviously its poor approach. Can their be better approach for this question? Please help.
Complete Question: https://www.hackerrank.com/contests/w6/challenges/consecutive-subsequences

Answer 1

Here is a fast O(n + k) solution:
1)Lets compute prefix sums pref[i](for 0 <= i < n).
2)Now we can compute count[i] - the number of prefixes with sum i modulo k(0 <= i < k).
This can be done by iterating over all the prefixes and making count[pref[i] % k]++.
Initially, count[0] = 1(an empty prefix has sum 0) and 0 for i != 0.
3)The answer is sum count[i] * (count[i] - 1) / 2 for all i.
4)It is better to compute prefix sums modulo k to avoid overflow.
Why does it work? Let's take a closer a look at a subarray divisible by k. Let's say that it starts in L position and ends in R position. It is divisible by k if and only if pref[L - 1] == pref[R] (modulo k) because their differnce is zero modulo k(by definition of divisibility). So for each fixed modulo, we can pick any two prefixes with this prefix sum modulo k(and there are exactly count[i] * (count[i] - 1) / 2 ways to do it).
Here is my code:
long long get_count(const vector<int>& vec, int k) {
//Initialize count array.
vector<int> cnt_mod(k, 0);
cnt_mod[0] = 1;
int pref_sum = 0;
//Iterate over the input sequence.
for (int elem : vec) {
pref_sum += elem;
pref_sum %= k;
cnt_mod[pref_sum]++;
}
//Compute the answer.
long long res = 0;
for (int mod = 0; mod < k; mod++)
res += (long long)cnt_mod[mod] * (cnt_mod[mod] - 1) / 2;
return res;
}

Answer 2

That have to make your calculations easier:
//Now we will move all numbers to [0..K-1]
long long int count=0;
for(int i=0;i<n;i++){
arr[i] = arr[i]%K;
}
//Now we will calculate cout of all shortest subsequences.
long long int sum=0;
int first(0);
std::vector<int> beg;
std::vector<int> end;
for(int i=0;i<n;i++){
if (arr[i] == 0)
{
count++;
continue;
}
sum += arr[i];
if (sum == K)
{
beg.push_back(first);
end.push_back(i);
count++;
}
else
{
while (sum > K)
{
sum -= arr[first];
first++;
}
if (sum == K)
{
beg.push_back(first);
end.push_back(i);
count++;
}
}
}
//this way we found all short subsequences. And we need to calculate all subsequences that consist of some short subsequencies.
int party(0);
for (int i = 0; i < beg.size() - 1; ++i)
{
if (end[i] == beg[i+1])
{
count += party + 1;
party++;
}
else
{
party = 0;
}
}
So, with max array size = 10^6 and max size of rest = 99, you will not have overflow even if you will need to summ all numbers in simple int32.
And time you will spend will be around O(n+n)