DSA Trainer
Binary Search
← All problems
PremiumBinary SearchMedium

Koko Eating Bananas

Asked at:AmazonGoogleMeta

Koko loves bananas and wants to eat all of them before the guards return in h hours. There are n piles of bananas. Each hour, Koko picks one pile and eats up to k bananas from it. If a pile has fewer than k bananas, she eats the whole pile and sits idle for the rest of that hour. Find the minimum eating speed k (bananas per hour) that allows her to finish all piles within h hours.

This is a premium problem.

Unlock Koko Eating Bananas with the full guided learning path and every other premium problem.

Problem

Koko loves to eat bananas. There are n piles of bananas. The guards have gone and will come back in h hours. Koko can decide her bananas-per-hour eating speed of k. Return the minimum integer k such that she can eat all the bananas within h hours.

Input

An integer array `piles` where `piles[i]` is the number of bananas in the i-th pile, and an integer `h`, the number of hours available.

Output

An integer `k`, the minimum eating speed (bananas per hour) that gets through all piles in at most h hours.

Examples

Input: piles = [3, 6, 7, 11], h = 8

Output: 4

At speed 4: pile 3 → 1hr, pile 6 → 2hrs, pile 7 → 2hrs, pile 11 → 3hrs. Total = 8hrs. At speed 3 it takes 10hrs, too slow.

Input: piles = [30, 11, 23, 4, 20], h = 5

Output: 30

Only 5 hours but 5 piles, must finish each pile in exactly 1 hour. Speed must be at least max(piles) = 30.

Input: piles = [30, 11, 23, 4, 20], h = 6

Output: 23

One extra hour gives Koko a bit more flexibility. 23 bananas/hr works.

The brute-force approach

Try every possible speed from 1 to max(piles). For each speed k, compute the total hours needed. Return the first k that finishes within h hours.

def hours_needed(k):
    return sum(ceil(pile / k) for pile in piles)

for k in range(1, max(piles) + 1):
    if hours_needed(k) <= h:
        return k

O(max(piles) × n), you try up to max(piles) speeds and compute hours for each in O(n). For large piles (up to 10^9), this is far too slow. The valid speeds form a monotonic property: if speed k works, every speed k+1, k+2, … also works. That monotonicity is exactly what binary search exploits.

Time: O(max(piles) × n)Space: O(1)

Spotting the pattern

This is a Binary Search problem. The key question to ask yourself:

If I can check whether a given speed works in O(n), how do I find the minimum working speed without trying every possibility?

Answering that is where it clicks, and it's exactly what the guided walkthrough below builds with you: the pattern reasoning, a progressive hint ladder that never spoils the answer, a row-by-row dry run, the optimized solution, and an in-browser editor to run your code against real test cases.

Premium

Unlock the full guided solution

The Binary Search walkthrough for Koko Eating Bananas: the progressive hint ladder, a row-by-row dry run, the optimized code, and an in-browser runner, plus the full guided learning path and every other premium problem.

Already finished Big-O and Hash Maps free? This picks up right where the path leads: same format, harder patterns, the ones that show up in every FAANG-style screen.

What you get

  • The full guided learning path: concept lessons, drills, graded tests, and cold-read capstones for every pattern
  • Stop going blank on algorithm problems. The hint ladder walks you to the answer without spoiling it.
  • Actually internalize the trace: dry runs walk state row by row so it clicks before you write a line
  • Know when to reach for each pattern. Every problem names the trigger so you spot it next time.
  • Write and run JS or Python in the browser with real test cases on every problem
  • Access every new problem, unit, and pattern guide as they ship weekly
B
Big_Wolverine_7575Founding Member· unprompted on Reddit

“This app finally made it click. I tried NeetCode, CTCI, and YouTube for years and still couldn't solve Two Sum. The beginner mental models for the patterns are genuinely so helpful. THANK YOU.

Price locks in for life · First 100 only

Founding Member Deal

$29 / year

or $6 / month

LeetCode Premium is $159/year and doesn't teach you anything.

Less than one Udemy course, with guided problems instead of passive lectures.

One extra month of job searching costs more than a full year here.

Your price is locked in for life.

When the 100th founding spot closes, the price goes up for everyone who joins after. Your rate never changes, no matter how much the curriculum grows.

  • The full guided learning path, every unit
  • All 150 problems across 33 patterns, plus every new one
  • New problems and units weekly
  • Cancel anytime

Cancel anytime. Not useful within 7 days? Email for a full refund. Secured by Stripe.