Assignment 3a: Package Management Concepts

Why virtual environments exist and how pip, conda, and uv compare.

Why Package Management Matters

When you install a package, it brings its own dependencies. For example, installing requests actually installs 5 packages: certifi, charset-normalizer, idna, urllib3, and requests itself. This is called a dependency tree.

The core problem is version conflicts — Project A might need numpy 1.26 while Project B needs numpy 2.0. Installing both into the same Python environment breaks one of them.

Virtual Environments

A virtual environment gives each project its own isolated Python installation. I proved this by creating two environments with different numpy versions:

# env-numpy1 had numpy 1.26.4
# env-numpy2 had numpy 2.4.3
# Both coexisted without conflict

pip vs uv Speed Comparison

I timed both package managers installing the same 4 packages:

• pip: 5.766 seconds
• uv: 1.627 seconds

uv was nearly 4x faster! uv uses a SAT solver for dependency resolution and caches aggressively.

Requirements Files

A requirements file pins exact versions so anyone can recreate the environment:

uv pip freeze > requirements.txt
uv pip install -r requirements.txt

This ensures reproducibility — the same packages, same versions, every time.