Installing Python

There is no single official way to install Python, which is the honest reason this page exists. Below are the two routes that work, when to pick each, and the handful of conventions that separate an installation that lasts a semester from one that breaks in week three.

If you want the short answer: install Miniforge, create one environment per project, and never install anything into base.

1. Two routes

For this course either works, and conda is the safer default because it handles the compiled libraries underneath NumPy and SciPy for you, on Windows in particular.

A word on distributions, since the names are confusing:

• Anaconda is the batteries-included distribution: Python plus several hundred preinstalled packages and a graphical Navigator. Convenient, large, and, since the 2024 change to its terms, subject to a paid licence for larger organisations. Check with your employer before putting it on a work laptop.
• Miniforge is the minimal installer that ships conda configured to use the community-run conda-forge channel. Same tool, no licensing question, and you install only what you need. This is what I use.
• Miniconda is Anaconda’s own minimal installer, and carries the same terms as Anaconda.

2. Route A: conda, step by step

Step 1: download and install

• Go to conda-forge.org/download for Miniforge, or anaconda.com/download if your institution provides Anaconda.
• Pick the installer matching your operating system and CPU architecture (Apple Silicon and Intel Macs are different downloads; on Windows, take the 64-bit one).

Installers are per-OS and per-architecture. Take Python 3.x. Python 2 reached end of life in January 2020 and is gone.

• Run the installer and accept the defaults. On Windows, do not tick “add to PATH”; use the Miniforge Prompt / Anaconda Prompt shortcut instead, which is what the installer is telling you when it warns you off. Installation takes a few minutes, or up to half an hour for full Anaconda.

If you installed Anaconda, you now also have the Navigator, a graphical front end to environments and applications:

The Navigator, if you installed Anaconda. Everything it does is also one command away, and the command line is the version that fits in a README.

Step 2: create an environment

This is the step people skip and later regret. An environment is an isolated set of packages with its own Python. One per project means a broken dependency in one place cannot break anything else, and the environment is throwaway, which is what makes it safe to experiment.

conda create -n course python=3.12 numpy pandas matplotlib scikit-learn jupyterlab
conda activate course

Your prompt should now be prefixed with (course). Rules of thumb:

• Never install into base. Keep it for conda itself.
• Prefer conda install over pip install inside a conda environment, and when you must mix, install everything you can with conda first, then pip. Mixing in the other order is the most common way to corrupt an environment.
• Record what you use: conda env export --from-history > environment.yml. That file, committed next to your code, is what makes the work reproducible.

To start again from scratch, which is always allowed:

conda deactivate
conda env remove -n course

3. Route B: plain Python, venv and pip

If you would rather not have conda, install Python from python.org (on Windows, tick Add python.exe to PATH), then:

python -m venv .venv
source .venv/bin/activate        # Windows: .venv\Scripts\activate
pip install numpy pandas matplotlib scikit-learn jupyterlab
pip freeze > requirements.txt

Faster modern alternative, if you like your tools sharp: uv replaces venv, pip and pip-tools in one binary and resolves environments in seconds.

uv venv && uv pip install numpy pandas matplotlib scikit-learn jupyterlab

4. Check that it works

python -c "import sys, numpy; print(sys.version); print(numpy.__version__)"

If that prints a 3.x version and a NumPy version, you are done. If the shell answers command not found or ‘python’ is not recognised, your environment is not activated. Go back and run conda activate course (or source .venv/bin/activate) in the terminal you are actually using.

5. Where you will write code

Jupyter, for exploration and reports

A notebook is a sequence of cells mixing text, equations, code and results. This is the format used throughout the course, because it keeps the reasoning and the output in the same document.

jupyter lab        # or: jupyter notebook

It opens in your browser. Use Help > Keyboard Shortcuts early; Shift+Enter to run a cell and Esc then A/B to insert one above/below will cover most of your day.

The file browser. New notebook: click New (or the Launcher) and pick your Python 3 kernel.

The initiation ritual. print("hello world!"), Shift+Enter.

One catch worth knowing before it bites you: the notebook’s kernel is a specific Python, and it is not always the environment you think. Check with import sys; print(sys.executable) in a cell. If your environment is missing from the kernel list, register it:

conda activate course
python -m ipykernel install --user --name course --display-name "Python (course)"

An editor, for code you keep

Notebooks are excellent for exploring and poor for anything reused: cells run out of order, and the hidden state is invisible. Once code stabilises, move it into .py files and import it.

• VS Code with the Python and Jupyter extensions is the common default, and runs notebooks natively.
• Spyder is closer to Matlab or RStudio: an editor, a console and a variable explorer side by side. F5 runs the file, F9 runs the selection while keeping previous variables in memory.
• PyCharm if you want a full IDE, free for students.

The command line, for everything else

Start an interactive session:

python              # or: ipython, which is the same thing but pleasant

>>> print("hello world!")
hello world!
>>> quit()          # or Ctrl-D

Run a script:

python script.py

And from inside IPython or a notebook, when you want a script’s variables left in your session afterwards:

%run script.py

6. When something goes wrong

Next: why Python in the first place, or go straight to the course notebooks.