Coding Challenges

57 self-contained algorithm and systems challenges. Most use file-based cases, an optimized reference, a deliberately naive benchmark control, and stubs in Python, Go, Rust, and C. A few exercises use an API test suite or a systems stress test instead. Run make in a language directory to format, build, lint, and test.

---

Quick start

# pick a challenge, pick a language
make -C 05-medium-price-streak/rust       # fmt → build → lint → test
make -C 05-medium-price-streak/go bench   # check and time your solution on large inputs

From the repo root, verify the whole bench at once:

make test        # I/O and API golden suites; I/O rotten small suites pass
make golden      # I/O and API golden tests pass; I/O benchmarks stay fast
make rotten      # I/O rotten controls pass small cases; every large case times out
make sys         # systems golden C stress tests pass
make sys-rotten  # systems controls pass sanity and fail controlled stress

I/O solver directories share these targets:

target	does
make	fmt → build → lint → test (default)
make build	compile or syntax-check
make fmt	format in place
make lint	run static analysis
make check	format, then lint
make test	correctness — small cases only, fast
make bench	correctness + speed — compare every large result, then report time
make help	list all targets

API, systems, and quiz challenges use targets suited to their test style. Their challenge README and make help list the available commands.

---

I/O challenge structure

NN-level-slug/
  README.md      ← the problem only: task, constraints, I/O, examples
  HINTS.md       ← the approach/technique — spoilers, open only when stuck
  cases/         ← NN.in / NN.out  (I/O challenges)
  golden/        ← optimized reference; always passes make test
  rotten/        ← deliberately naive benchmark control
  python/        ← stub: implement solve() in main.py
  go/            ← stub: implement solve() in main.go
  rust/          ← stub: implement solve() in src/main.rs
  c/             ← stub: implement solve() in solution.c

The C directory ships a finished scaffold — json.h, harness.h, main.c, and test.c are shared and identical everywhere. Only solution.c holds the stub: input_parse and answer_print are written for you, solve() is not.

The README.md, challenge title, directory slug, and catalog never narrow the solution search. All guidance, including rejected approaches and complexity comparisons, lives in HINTS.md.

Four challenge types:

• io — reads JSON from stdin and writes one line in the documented format
• api — implements functions checked directly by a language test suite
• sys — exposes a systems API; the test is a stress test rather than files
• quiz — asks for predictions about standalone programs, then checks them

I/O challenge input is always JSON ({"n":4,"edges":[[0,1]],"loads":[10,null]}), so parsing is real work rather than splitting whitespace.

---

Challenges

Difficulty is based on prerequisite depth, correctness edge cases, implementation burden, and the constraints enforced by make bench.

#	Name	Level	Lang
01	Maximum Subarray	easy	py go rs c
02	Modular Power	easy	py go rs c
03	Max Drawdown	easy	py go rs c
04	Vertex Load Assignment	medium	py go rs c
05	Price Streak	medium	py go rs c
06	Edit Distance	medium	py go rs c
07	Coin Change	medium	py go rs c
08	Interval Scheduling	medium	py go rs c
09	Count Inversions	medium	py go rs c
10	Route Costs	medium	py go rs c
11	Friend Groups	medium	py go rs c
12	Textbook Split	medium	py go rs c
13	Sliding Window Maximum	medium	py go rs c
14	Count Primes	medium	py go rs c
15	Huge Fibonacci	medium	py go rs c
16	String Search	medium	py go rs c
17	0/1 Knapsack	medium	py go rs c
18	Task Ordering	medium	py go rs c
19	Cheapest Road Network	medium	py go rs c
20	Longest Common Subsequence	medium	py go rs c
21	Constraint Puzzles	medium	py go
22	Unbounded Sequences	medium	py go
23	Search Suggestions	medium	py go rs c
24	Cache Eviction	medium	py go rs c
25	Running Median	medium	py go rs c
26	Dynamic Prefix Sums	medium	py go rs c
27	Weighted Job Scheduling	medium	py go rs c
28	News Feed Merge	medium	py go rs c
29	Multi-Producer Queue	hard	go rs c
30	Consistent Tick Snapshot	hard	go rs c
31	Concurrent Owner/Thief Deque	hard	rs c
32	Two-Thread Buffer	hard	go rs c
33	Lock-Free Stack Reclamation	hard	rs c
34	Reusable Spin Barrier	hard	go rs c
35	Dynamic Range Sums	hard	py go rs c
36	Matrix Chain Multiplication	hard	py go rs c
37	Prime Pair Sets	hard	py go rs c
38	Distinct Substrings	hard	py go rs c
39	Max Flow	hard	py go rs c
40	Go Concurrency Quizzes	hard	go
41	Ordered Set Queries	hard	py go rs c
42	Fragmented String Queries	hard	py go rs c
43	Order Book	hard	py go rs c
44	Affine Alignment Score	hard	py go rs c
45	K-mer Assembly	hard	py go rs c
46	CRISPR Off-Targets	hard	py go rs c
47	RNA Max Pairs	hard	py go rs c
48	Shortest Superstring	hard	py go rs c
49	Gene Region Decoder	hard	py go rs c
50	Tree Sequence Likelihood	hard	py go rs c
51	Deadline Scheduler	hard	py go rs c
52	Service Pairing	hard	py go rs c
53	Circular Genome Distance	hard	py go rs c
54	Spectrum Peptide Recovery	hard	py go rs c
55	Changing Network Queries	hard	py go rs c
56	Orthogonal Segment Crossings	hard	py go rs c
57	Causal Event Replay	hard	py go rs c

---

Adding a challenge

1. Copy template/ to the next numbered directory.
2. Write the pure specification in README.md. Put all solution guidance and solution-bearing sources in HINTS.md.
3. Add the optimized implementation in golden/ and the short, correct, naive control in rotten/. Keep every solver solve body stubbed.
4. Add at least eight small fixture pairs and at least two large pairs. Each large input must independently make the rotten implementation exceed its timeout.
5. Run the root make golden and make rotten contract checks.
6. Add the challenge to the catalog above.

---

Sources

Each challenge keeps solution-neutral attribution in its README.md. Solution-bearing references live in HINTS.md. Upstream problem sets and reference works are credited in NOTICE.

Development assisted by Claude Code (Anthropic).

---

License

GNU General Public License v3.0 — LICENSE.