Skip to content

Latest commit

 

History

History
59 lines (45 loc) · 2.58 KB

anaconda-on-windows.md

File metadata and controls

59 lines (45 loc) · 2.58 KB
Raw

Anaconda on Windows

  1. Download Miniconda (light-weight version of Anaconda). Choose Miniconda3 Windows 64-bit.
  2. Install. Use the default recommended settings.
  3. From the Start menu, search and open the Anaconda Prompt.

Working with virtual environments

When you first open the Anaconda Prompt, you should see (base) in front of the prompt. This means that you are currently in the default "base" environment.

A best practice is to keep the base environment as minimal as possible and create a new separate environment every time you start a new project:

$ conda create -n my-new-project python=3.9 git pip

In the above, we created a new environment called my-new-project with Python 3.9, Git, and Pip (the standard package installer for Python).

To activate the environment:

$ conda activate my-new-project

After this, the prompt should change from (base) to (my-new-project).

To exit the current environment:

$ conda deactivate

To delete the environment:

$ conda env remove -n my-new-project

To list existing environments:

$ conda env list

Why Anaconda?

I find Anaconda to be the easiest way to get started with Python on Windows. In many Linux systems, Python comes pre-installed and many purists prefer the standard tools to manage virtual environments like venv and virtualenv.

Still, there are cases where Anaconda is very convenient on Linux. For example, the system may have an old version of Python but we don't have access to sudo or admin privileges to update it (e.g. we are using a company laptop, or logged into a compute cluster). Anaconda lets us install most data science tools without admin permissions. This is because Anaconda is not merely an environment manager, but a platform on which we can do more things. For example, we can install R and Java JDK on it.

Avoid mixing conda with pip to manage packages

Both Anaconda and Pip can be used to install Python packages: In Anaconda: conda install my-package, In Pip: pip install my-package. However, it is recommend that you stick to one method (preferably Pip). My main workflow is as follows:

# Create a new virtual environment with Python, Git, and Pip
$ conda create -n awesome-project python=3.9 git pip
$ conda activate awesome-project
# Install packages using Pip
$ pip install awesome-package
$ ...

That is, use Anaconda only for the environment management (create, activate, deactivate, delete, etc.) and setup of high-level tools (e.g. Python, Git, Pip, Java, etc.).