Installing Packages: pip, conda, and dependencies

Before You Start

You should know: - How to create a new virtual environment using venv or conda. - How to activate your environment so your terminal runs within the isolated sandbox.

You will learn: - What a package manager is and why Python relies on them. - How to search for, install, and uninstall packages using pip and conda. - How to freeze your environment into a requirements.txt file so your work is reproducible.

Introduction

Out of the box, Python is small. It knows how to do basic arithmetic, manipulate strings, and handle text files. But if you ask pure Python to invert a 1000x1000 matrix or extract temperature values from a GeoTIFF, it won’t know how.

For scientific computing, you must expand Python’s vocabulary by downloading packages (also called libraries or modules). Packages are bundles of code written by other programmers that solve specific mathematical or geographic problems.

Instead of searching the web, downloading .zip files, and dragging them into folders manually, you use a package manager. A package manager is a command-line tool that automatically finds, downloads, installs, and links the correct version of a library into your active virtual environment.

There are two primary package managers in the Python ecosystem: pip and conda.

The Standard Pathway: pip

pip stands for “Pip Installs Packages”, and it is the default program installed with standard Python. It pulls packages from an enormous central repository called the Python Package Index (PyPI).

Before using pip, you must always ensure your virtual environment is activated. If you run a pip command without an active sandbox, it will permanently inject the package into your global system!

Installing a package

Let’s install numpy, the foundational library for doing advanced math in Python.

In your terminal, run:

pip install numpy

Your terminal will print out a rapid sequence of logs. It is reaching out to PyPI, finding the most recent stable version of numpy, downloading the files, and placing them neatly into your .venv folder.

You can install multiple packages at once:

pip install pandas matplotlib scipy

Checking what is installed

To see a list of every package currently inside your sandbox, run:

pip list

You will notice things in this list that you did not explicitly ask for. This is because package managers resolve dependencies. When you installed pandas, pip realized that pandas requires pytz (a timezone library) in order to function, so it fetched it for you automatically.

Uninstalling a package

If you made a mistake or no longer need a library, you can safely remove it:

pip uninstall numpy

The Data Science Pathway: conda

If you installed Miniforge or Anaconda in Chapter 4, your primary package manager is conda. It acts very similarly to pip, but instead of pulling from PyPI, it pulls from curated channels specifically optimized for heavy data science workflows.

Installing with conda

Before installing, ensure your environment is active (e.g., conda activate my-geo-project).

To install numpy, you use syntax identical to pip:

conda install numpy

When you run this, conda will think for a moment. It calculates a “solve”—mapping out exactly how this new package will interact with the C++ binaries and existing packages on your system. It will prompt you with a list of actions and ask you to confirm [y/N]. Type y and hit Enter.

The conda-forge channel

In geography, you will often need highly specialized spatial libraries (like geopandas or rasterio). The best place to find these is the community-maintained conda-forge channel.

Miniforge checks this channel by default, but if you are running standard Anaconda, you must explicitly tell it where to look:

conda install -c conda-forge geopandas

The Golden Rule of Reproducibility

Imagine completing an intense data analysis over six months. You publish your paper and send your Python script to a reviewer. The reviewer runs your script, but it immediately crashes. They have numpy version 1.25, but your code implicitly relied on a mathematical function that only existed in numpy version 1.18.

To prevent this, you explicitly record the state of your sandbox using a requirements file.

Exporting your environment

If you used pip, run:

pip freeze > requirements.txt

This takes the exact list of every installed package, along with their highly specific version numbers, and writes it to a plain text file named requirements.txt.

If you used conda, the equivalent command exports a YAML file:

conda env export > environment.yml

Rebuilding an environment

When your reviewer receives your folder, they don’t have to guess what you used. They simply create an empty virtual environment and instruct the package manager to read the text file.

Rebuilding with pip:

pip install -r requirements.txt

Rebuilding with conda:

conda env create -f environment.yml

The package manager will instantly reconstruct the exact mathematical ecosystem you used six months ago.

Verify Your Work

Let’s test the workflow end-to-end. 1. Open your terminal and navigate to your coding_projects folder. 2. Activate your virtual environment. 3. Install numpy, pandas, and matplotlib. 4. Run pip list (or conda list) to visually verify they were installed. 5. Export your active environment configuration into a requirements.txt (or environment.yml) file. Use the ls command to confirm the file was successfully created in your directory.

With an active sandbox and the ability to download powerful geographical libraries, you are now ready to write real scientific code.