Lecture 6 Exercises — Unit and Functional Testing (1-week workload)

Objective: turn the VWCE C++ and Python examples from Lecture 5 into test-driven mini-projects. Work in your sspa-sandbox under ./<name.surname>/, keeping a lab log (lab-notebook.md) with commands, design notes, and failing/passing test outputs. Aim for repeatable suites: no network calls, small fixtures, deterministic assertions.

Exercise 1 — Setup and baseline¶

1. Copy the Lecture 5 materials into ./<name.surname>/:
   cp -r /path/to/course/codes/lab05/{c++,python,vwce_2024.csv,fetch_vwce.sh} codes/lab06/<name.surname>/

2. Verify the original programs still run on the full CSV (C++ and Python) and note current outputs in the lab log.

3. Ensure toolchains are ready: C++17 compiler plus gtest (system package or vendored), Python 3 with pytest. Record versions: g++ --version, pytest --version.

Exercise 2 — gtest suite for the C++ code¶

1. Create c++/tests/test_vwce.cpp (or similar) and wire a build: a simple CMakeLists.txt with find_package(GTest) + enable_testing() is fine, or a one-shot command like
   g++ -std=c++17 main.cpp tests/test_vwce.cpp -lgtest -lgtest_main -pthread -o test_vwce.

2. Add unit tests for pure functions: mean, variance (including single-value/empty vectors), daily_returns (constant and increasing prices), annualized_return (zero/positive series), max_drawdown (known mini-series with expected start/end dates), and CSV error handling (load_prices throws on missing file).

3. Add a functional/smoke test that runs the main pipeline on a tiny CSV fixture (3–5 rows) written in the test and checks the gnuplot output file has monotonically non-decreasing running peaks and matching row count.

4. Run ctest or ./test_vwce until all cases pass; fix any bugs discovered in main.cpp rather than weakening tests. Capture failures and fixes in the notebook.

Exercise 3 — pytest suite for the Python code¶

1. Under python/tests/, add conftest.py for shared fixtures (e.g., a small PricePoint list and a temporary CSV writer using tmp_path).

2. Write unit tests mirroring the C++ coverage: loader error path, mean/variance, daily/annualized returns, drawdown tuple, and gnuplot writer (verify header + running peak monotonicity).

3. Add a CLI smoke test: run python -m main --csv <tiny_csv> --out <tmp> via subprocess.run (capture stdout) and assert exit code 0 plus file creation.

4. Execute pytest -q (or pytest -q python/tests) and iterate until green. Note any discrepancies with the C++ results and align behaviors if appropriate.

Exercise 4 — Regression guards and hygiene¶

1. Add at least one regression test for a bug you fixed (mark it clearly in the test name/docstring).

2. Keep fixtures small and checked in; avoid relying on the full 2024 CSV for fast runs. If you need floating-point comparisons, use EXPECT_NEAR/pytest.approx with justified tolerances.

3. Optionally add ctest/pytest targets to a root Makefile or CI config in your sandbox to make running both suites a one-liner.

Submission checklist¶

• ./<name.surname>/ contains the C++/Python sources plus test suites and any minimal fixtures; no large generated artifacts committed.

• lab-notebook.md with commands, failing/passing outputs, and brief notes on fixes/regressions.

• A clean git history in your sandbox branch; push and open a PR to the teachers’ repository, mentioning remaining questions for Lecture 7.