When to use
DFS (stack/recursion) explores one branch fully before backtracking. BFS (queue) explores level-by-level.
Rules
- If the problem asks "shortest" or "minimum steps" on an unweighted graph, ALWAYS choose BFS
- If it asks "all paths," "can we reach," or "count islands," DFS is simpler
- NEVER use BFS for cycle detection — DFS is the right tool
- Both visit each node once: O(V+E) time
DFS use cases
Cycle detection, topological sort, path existence, connected components, backtracking puzzles, flood fill.
BFS use cases
Shortest unweighted path, level-order traversal, nearest neighbor, minimum steps.
Example
"Count islands" → DFS flood-fill on 2D grid. "Shortest path in maze" → BFS with queue tracking (row, col, distance).