Virtual Environments and pip

A virtual environment is an isolated Python installation for a single project. Each project gets its own set of packages, so upgrading NumPy for one project can never break another.

Without isolation, all packages install globally and version conflicts become unavoidable across data science projects.

The built-in venv module creates an environment in a folder you name (commonly .venv).

Run this inside your project root. It creates a self-contained copy of the interpreter and a place for packages.

Activation puts the environment's python and pip at the front of your PATH. Your prompt usually shows (.venv) when active.

On Windows the activation script lives under Scripts. PowerShell and cmd use slightly different commands.

When you are done, deactivate restores your normal shell PATH. The environment files remain on disk for next time.

With the environment active, pip install places packages inside .venv only. Install the data science core in one line.

Use == to lock an exact version, which keeps results reproducible. Use >= for a minimum.

pip freeze prints every installed package with its exact version. Redirect it into requirements.txt to capture the full environment.

A teammate (or your CI server) recreates the exact environment with -r. This is the heart of reproducible data science.

It is a plain text file, one pinned package per line. You can edit it by hand to relax or tighten versions.

Never commit the environment folder; it is large and machine-specific. Commit requirements.txt instead and ignore the folder in .gitignore.

Test your understanding of dependency management.

You learned the reproducibility workflow:

• python -m venv .venv creates an isolated environment
• source .venv/bin/activate / deactivate toggle it
• pip install adds packages locally
• pip freeze > requirements.txt captures versions
• pip install -r requirements.txt restores them
• Add .venv/ to .gitignore