CareerCup

Use KD BST tree for storing these events and perform the range search. It gurantees to be O(nlogn) times in worst case can go upto n.

Sort by start date, probably into a multimap. Than one more path is enough to mark overlaps.

What about creating an arrany of "int Day[30]". Initialize with zero. Increment the value at each index once a date a found on that day.

Create a binary search tree which stores the end points of each time interval. Each node should also store what time interval its value comes from.

Notation: dates are represented in MMDD format.
Nodes: (1010, 0), (1015, [0, 2]), (1012, 1), (1014, 1), (1024, 2), (1023, 3), (1103, 3), (1004, 4), (1005, 5).

After constructing this BST, for each time interval, perform a search looking for a value that is contained in the given interval AND does not belong to the given interval. These searches should be directed towards branches that may lead to time conflicts, i.e. if we got the interval [1015, 1024] and we face a node with value 1010, naturally we need to conduct the search to the branch in the right.

Assuming that we have the ability to create a balanced BST, the construction phase is O(nlogn) and the search phase is also O(nlogn).