Problem
With async/await you can start several tasks and let them run at the same time. A SemaphoreSlim often caps how many run at once.
Tasks numbered 0 to n − 1 take durations[i] time units. At time 0 start as many tasks as the limit allows, in index order. Whenever tasks finish, record them (if several finish at the same moment, lower index first) and immediately start the next waiting tasks, in index order, to fill the free slots.
Implement parallel_task_order(durations, limit).
Input
durations: How long each task takes.
limit: The maximum number of tasks running at the same time.
Output
Return the task indices in the order they finish.
Constraints
1 ≤ n ≤ 100,000
1 ≤ durations[i] ≤ 10^6; 1 ≤ limit ≤ n
Examples
Input:
limit = 2
durations = [4, 2, 3, 1]
Expected output:
expected = [1, 0, 2, 3]
Input:
long long limit = 2;
std::vector<long long> durations = {4, 2, 3, 1};
Expected output:
std::vector<long long> expected = {1, 0, 2, 3};
Input:
long limit = 2;
long[] durations = new long[] {4, 2, 3, 1};
Expected output:
long[] expected = new long[] {1, 0, 2, 3};
Explanation
Tasks 0 and 1 start. Task 1 ends at 2 and task 2 starts (ends at 5). Task 0 ends at 4 and task 3 starts (ends at 5). Tasks 2 and 3 finish together, lower index first: [1, 0, 2, 3].
Input:
limit = 3
durations = [3, 1, 2]
Expected output:
expected = [1, 2, 0]
Input:
long long limit = 3;
std::vector<long long> durations = {3, 1, 2};
Expected output:
std::vector<long long> expected = {1, 2, 0};
Input:
long limit = 3;
long[] durations = new long[] {3, 1, 2};
Expected output:
long[] expected = new long[] {1, 2, 0};
Explanation
All tasks run at once and finish at 1, 2 and 3: [1, 2, 0].
How to approach it
Try on your own first, then open one hint at a time.
Hint 1
Keep the running tasks ordered by finish time, then by index.
Hint 2
Each step jumps to the next finish time and collects every task ending then.
Hint 3
Refill free slots with the next waiting tasks in index order.
Expected complexity
O(n log limit) time, O(limit) extra space
References
This challenge is adapted for CodeUtility. Follow the statement and constraints above.