362. design hit counter

Design a hit counter which counts the number of hits received in the past 5 minutes (i.e., the past 300 seconds).

Your system should accept a timestamp parameter (in seconds granularity), and you may assume that calls are being made to the system in chronological order (i.e., timestamp is monotonically increasing). Several hits may arrive roughly at the same time.

Implement the HitCounter class:

• HitCounter() Initializes the object of the hit counter system.
• void hit(int timestamp) Records a hit that happened at timestamp (in seconds). Several hits may happen at the same timestamp.
• int getHits(int timestamp) Returns the number of hits in the past 5 minutes from timestamp (i.e., the past 300 seconds).

solutions

queue w/ hashmap

My initial idea was to keep track of the timestamps in a queue, adding to the right, and then delete the oldest ones from the left when the newest timestamp seen is out of the 300 second range.

Then, as an optimization, I keep track of a running total, although equally, we could just not do that and return len(self.timestamps) in getHits.

This gives us $O(1)$ time for hit and worst-case $O(n)$ for getHits.

class HitCounter:
	def __init__(self):
		self.timestamps = deque()
		self.total = 0
	  
	def hit(self, timestamp: int) -> None:
		self.timestamps.append(timestamp)
		self.total += 1
	  
	def getHits(self, timestamp: int) -> int:
		while len(self.timestamps) and self.timestamps[0] <= (timestamp-300):
			removed_timestamp = self.timestamps.popleft()
			self.total -= 1
		return self.total

just a queue

class HitCounter:
	def __init__(self):
		self.timestamps = deque()
	  
	def hit(self, timestamp: int) -> None:
		self.timestamps.append(timestamp)
	  
	def getHits(self, timestamp: int) -> int:
		while len(self.timestamps) and self.timestamps[0] <= (timestamp-300):
			self.timestamps.popleft()
		return len(self.timestamps)

array w/ binary search

Instead of linearly removing elements that have fallen out of the window, we can just maintain the entire window, and do a binary search for the left bound whenever we call getHits.

This lets us improve the worst-case runtime of getHits to $O(\log n)$.

class HitCounter:
	def __init__(self):
		self.a = [float("-inf")]
 
	def hit(self, timestamp: int) -> None:
		self.a.append(timestamp)
 
	def getHits(self, timestamp: int) -> int:
		bottom_bound = bisect.bisect_right(self.a, timestamp-300)
		return len(self.a) - bottom_bound