Introduction to Conda Beginner

Conda is an open-source package manager and environment manager that runs on Windows, macOS, and Linux. It quickly installs, runs, and updates packages and their dependencies, while letting you create isolated environments to prevent conflicts between projects.

What is Conda?

Conda is a cross-platform tool that solves two fundamental problems in software development:

1. Package Management

   Installing, updating, and removing software packages and all their dependencies. Unlike pip (which only handles Python packages), Conda manages packages for Python, R, C++, Java, Fortran, and many other languages.

2. Environment Management

   Creating isolated environments where you can install specific versions of packages without affecting other projects or your system Python installation.

Conda vs pip vs venv vs Poetry

Understanding how Conda compares to other Python tools helps you choose the right tool for your workflow:

Multi-Language Support

One of Conda's greatest strengths is its ability to manage packages across multiple programming languages and system-level dependencies:

Conda, Anaconda, and Miniconda

These three terms are often confused. Here is how they relate:

Think of it this way: Miniconda gives you just Conda so you install only what you need, while Anaconda gives you everything pre-installed. Both use Conda as their package manager.

Why Use Conda?

Dependency Resolution

Conda uses a SAT (satisfiability) solver to find compatible versions of all packages. When you install a package, Conda checks all dependencies and their dependencies, ensuring everything works together without conflicts.

Binary Packages

Conda distributes pre-compiled binary packages, so you never need to compile C extensions from source. This is especially valuable for packages like NumPy, SciPy, and TensorFlow that depend on optimized math libraries (MKL, BLAS, LAPACK).

Cross-Platform

The same Conda commands work on Windows, macOS, and Linux. An environment.yml file created on one platform can recreate the same environment on another (with platform-appropriate binaries).

Non-Python Dependencies

Need CUDA for GPU computing? A specific version of OpenSSL? A C compiler? Conda handles these system-level dependencies that pip simply cannot manage.

# Conda installs CUDA toolkit as a regular package
$ conda install cudatoolkit=11.8

# No need to install CUDA system-wide!
# Each environment can have its own CUDA version

Quick Start Preview

Here is a taste of what working with Conda looks like:

# Create a new environment with Python 3.11
$ conda create -n myproject python=3.11

# Activate the environment
$ conda activate myproject

# Install packages
$ conda install numpy pandas scikit-learn

# Check what's installed
$ conda list

# Export for reproducibility
$ conda env export > environment.yml