Snapshot Array
A medium-tier problem at 37% community acceptance, tagged with Array, Hash Table, Binary Search. Reported in interviews at StackAdapt and 9 others.
Snapshot Array keeps showing up at Google, TikTok, Nvidia, and Snowflake, but most candidates fail it. You're asked to design a data structure that handles set(index, val) and snap() operations on an array that could have millions of indices. The naive approach stores the entire array, then copies it on every snap(). That fails immediately on space and time limits. The trick is storing only the deltas per snapshot, indexed by a hash table, so you can retrieve any value at any snapshot in O(log n) time. StealthCoder solves it in seconds if you blank on the design.
Companies that ask "Snapshot Array"
Snapshot Array is the kind of problem that decides whether you pass. StealthCoder reads the problem on screen and surfaces a working solution in under 2 seconds. Invisible to screen share. The proctor sees nothing. Made by a working Amazon engineer who got tired of watching qualified friends bomb OAs they'd solve cold in an IDE.
Get StealthCoderThe problem tests whether you can recognize that most array indices never change. Instead of copying the full array on snap(), store only modified indices and their values in a map per snapshot. When retrieving a value at a given snapshot, binary search the history of that index to find the most recent change. The common trap is thinking you need a 2D array or a full copy on each snap(). Wrong. You need a dictionary of indices, where each index maps to a list of (snapshot_id, value) pairs. Then binary search that list. The pattern combines Hash Table for fast index lookup with Binary Search for efficient history traversal. If you haven't drilled this design problem before, StealthCoder becomes your safety net during the live OA.
Pattern tags
You know the problem.
Make sure you actually pass it.
Snapshot Array recycles across companies for a reason. It's medium-tier, and most candidates blank under the timer. StealthCoder is the hedge: an AI overlay invisible during screen share. It reads the problem and surfaces a working solution in under 2 seconds. Made by a working Amazon engineer who got tired of watching qualified friends bomb OAs they'd solve cold in an IDE. Works on HackerRank, CodeSignal, CoderPad, and Karat.
Snapshot Array interview FAQ
Why do candidates fail Snapshot Array so often?+
Most try to store the entire array state per snapshot, which eats O(n) space and time per snap() call. The problem is a design problem disguised as an array problem. You have to recognize that sparsity is the key constraint and model the data structure accordingly. Acceptance is 37%, so the naive approach burns most attempts.
Is this really asked at FAANG?+
Yes. Google, Snap, and TikTok all report it. It tests design thinking and space-time tradeoff reasoning, which matters in system design interviews. It's in the medium bucket for difficulty but hard for conceptual framing, which is why top companies use it.
What's the actual trick?+
Use a dictionary where each index key maps to a list of (snapshot, value) tuples. On set(), append to that index's list. On snap(), increment a counter. On get(), binary search the index's list to find the value at the target snapshot. This is O(log n) per get() and O(1) per set() and snap().
How does Binary Search fit into this?+
Each index has a sorted history of (snapshot_id, value) pairs. When you query get(snapshot, index), you binary search that index's history to find the most recent change at or before that snapshot. Without binary search, you'd scan the whole history linearly and blow the time limit.
Can I just use a 2D array?+
No. If the array length is 10^8 and you have 10^4 snapshots, a 2D array is 10^12 cells. The problem guarantees millions of indices and multiple snapshots. Hash Table plus binary search on sparse histories is the only viable approach.
Want the actual problem statement? View "Snapshot Array" on LeetCode →