What is time vs space complexity?

Time complexity bounds operations vs input size; space bounds extra memory. Interviewers want you to state best/average/worst when relevant and tie choices to constraints.

When is a hash map a good default?

When you need O(1) expected lookups by key and can afford O(n) extra space. Watch hash collisions, ordering (use `LinkedHashMap`/tree if sorted keys matter), and key lifetime.

Two-sum intuition with a hash set/map.

For each element, check if target - x exists. One pass after building the map, or single pass with complement lookup.

Reverse a linked list iteratively.

Maintain prev, curr, next; redirect `curr.next` to `prev` and advance. O(n) time, O(1) space.

Valid parentheses with a stack.

Push opening brackets; on closing, pop if matches; else invalid. End with empty stack.

BFS vs DFS — when do you pick each?

BFS for shortest path on unweighted graphs / level order. DFS for cycles, topo sort, exhaustive paths with pruning.

Detect a cycle in a directed graph.

White/Gray/Black DFS states or Kahn indegree for topo. If DFS hits a gray node, cycle exists.

Binary search: common pitfalls?

Off-by-one on `mid`, integer overflow — use `lo + (hi-lo)/2`. Know lower_bound vs upper_bound variants.

Merge two sorted arrays in place (concept).

Typical trick: fill from back when gap exists to avoid overwriting needed values.

DP: what makes a subproblem definition valid?

Optimal substructure + overlapping subproblems; state captures all decisions needed; transition is correct for all prior states.

Word break / partition-style DP pattern.

`dp[i]` = can segment prefix `i`; for each word from dictionary try match ending at `i`.

Dijkstra vs Bellman-Ford at a high level.

Dijkstra needs non-negative edges; Bellman-Ford tolerates negative edges (no neg cycles) but slower.

DSA Interview Questions | Beginner to Advanced | InterviewForge AI