CASE_STUDY.md

🧩 Case Study — Seaborn & Matplotlib Visual Lab

Context

Many visualization projects stay as notebooks: hard to reuse, inconsistent exports, and difficult to validate in CI.

This repo packages a visualization learning workflow into an interactive Streamlit app that:

• builds plots from UI controls,
• shows the generated code,
• exports consistent PNGs (and ZIP galleries),
• and remains testable and maintainable.

---

Problem

• Learning visualization often requires repetitive boilerplate and copy/paste.
• Notebooks are great for exploration, but weak for sharing a consistent interactive experience.
• Visualization code can break when dependencies change (Seaborn/Streamlit API changes).

---

Solution

A Streamlit “visual lab” with:

• dataset loading (Seaborn catalog),
• Seaborn builders (high-level API patterns),
• Matplotlib builders (low-level control),
• side-by-side comparison,
• a gallery for exporting outputs.

---

Key engineering decisions

1) CI-friendly entrypoint (safe imports)

The app avoids expensive work at import-time by keeping heavy logic inside functions. This makes import app safe for smoke tests and GitHub Actions.

2) Offline resilience

If Seaborn’s dataset catalog is unavailable, the app falls back to a small built-in dataset to keep the UI usable.

3) Quality gates

• ruff for formatting/linting
• pytest smoke tests
• pre-commit hooks (optional)
• GitHub Actions to validate PRs

---

Validation

• Local run: streamlit run app.py
• Lint/format: python -m ruff check . --fix and python -m ruff format .
• Tests: python -m pytest -q

---

Result

A clean, reproducible visualization playground that is:

• interactive (UI-driven),
• explainable (code shown),
• exportable (PNG/ZIP),
• and stable under CI (tests + safe imports).

---

Next steps (optional)

• Add a small performance budget (cache heavy computations).
• Add a “gallery metadata” export (JSON) for reproducible plot settings.
• Provide a minimal Docker deploy guide (if needed).