Resources

DSA Patterns is the pattern-recognition layer for LeetCode-style interviews: a curated catalog of techniques, multi-language templates, mental models, and on-site write-up paths. We restate problems and cite sources, we do not copy official editorials.

• Start here: refresh Big-O → roadmap foundations → pattern template → write-up + tests.
• Practice elsewhere: volume on LeetCode / NeetCode; visuals on VisuAlgo; contests on Codeforces / AtCoder; live practice on Pramp or interviewing.io.
• Go deeper with books: CLRS (theory), EPI / CtCI (interview problems), Grokking (visual intuition).

Before coding

• Restate the problem: inputs, outputs, and types.
• Ask for the constraints that change the algorithm: input size, whether you may mutate the input, and any time or memory bound.
• Walk a normal example, then an empty one and a one-element one. Check that reading with the interviewer.
• Name the pattern family: scan, search, window, graph, tree, or dynamic programming.

While coding

• Put a correct brute force on the table and state its complexity, then tighten it.
• Keep talking. Narrate the approach, the trade-off, and the alternative you did not pick.
• Use names that say what the value is.

After coding

• Trace the examples you already agreed on.
• Trace the edge cases below.
• State time and extra space, and which line dominates.
• If a tighter bound or a cleaner shape is obvious, say what it is.

Edge cases

• Numbers: zero, negatives, and the stated minimum and maximum. Fixed-width overflow in Java and C. Python ints do not overflow. Division by zero. Floats: compare with a tolerance, not ==.
• Strings: empty string and a single character. All characters the same. Spaces, punctuation, and non-ASCII. None versus "" in languages that have both.
• Arrays: empty, one element, all duplicates. Already sorted and reverse sorted. None entries, if the language allows them.
• Trees and graphs: empty, and a single node. Only left children, or only right children. Disconnected graph, self-loop, cycle. Topological sort needs a DAG. Dijkstra needs non-negative weights.
• Matrices: empty, 1×1, a single row, a single column. Square versus rectangular.

Habits that help

1. Agree on the approach before writing it down. A few quiet minutes of thinking are useful. Going silent for the rest of the interview is the failure mode.
2. Ask when a constraint is missing.
3. Test before you are asked.
4. State time and space every time.
5. Treat a hint as new information and change the solution.

Habits that hurt

1. Coding before you can restate the problem.
2. Skipping the empty input.
3. Ignoring a hint.
4. Defending a bug instead of tracing it.
5. Building a framework for a problem that needed one loop.
6. Stopping without a complexity.