Setting Up Your Environment

In this course we will use the free Miniforge distribution of Python and Python packages. Miniforge is available for Windows, MacOS, and Linux. Download the latest release version for your operating system from the above page, and follow the instructions to install the base. Mac users: “Apple Silicon” refers to the M chips on newer Macs. Please be sure to install the standard Miniforge3 and not Miniforge-pypy.

There are two somewhat incompatible versions of Python; version 2.7 is obsolete but still sometimes used. Version 3 is the supported version. We will use Python 3, preferably version 3.9 or above. If you find that you need to use Python 2.7, Miniforge makes it easy to run both versions side-by-side through environments. At the time of this writing, Miniforge3 installs Python 3.10.4 by default with Release 23.0-0, but will later switch to Python 3.12.0.

When installing, we strongly recommend staying with the defaults unless you are experienced and very sure you want to make changes. In particular, it is best to install it “Only for You.”

On Linux, the installation is a shell script that you must run with

bash ~/Downloads/Miniforge3-Linux-x86_64.sh 

Be aware that it will alter your .bashrc or other shell resource file unless you request that it not do so.

Multi-user Linux systems such as UVA HPC may have Miniforge or other Python distributions preinstalled. See the instructions for your local system, such as ours. Typically, on a system such as UVA’s HPC resource, you must load a module before you can use the distribution.

On Windows, Miniforge will install a miniforge prompt which will appear in your Apps menu under the Miniforge folder. It can be used much like a terminal on Linux or Mac. It will be aware of python, pip, mamba, and conda commands, but unless you add those to your personal PATH environment variable, which is usually not recommended, other CMD or PowerShell windows will not find those commands.

The default installation of Miniforge creates a base environment consisting of a very minimal number of packages. It uses a package manager to install further packages. You can choose conda or mamba to install packages. In general, we recommend mamba as it is usually much faster than conda.

If you wish to install some of the more popular packages into your base environnment, open a terminal (Linux and MacOS) or a Miniforge prompt (Windows). A good “starter pack” would be NumPy, Matplotlib, Pandas, Seaborn, Jupyterlab, and Spyder. We would also like to upgrade our Python from the 3.10.4 in our current release version. At your prompt type, all on one line,

(base)$mamba install python=3.11 numpy pandas scipy mkl seaborn numba jupyterlab ipython ipywidgets spyder

This will install many more packages than just those requested, since each will have packages on which they depend.

You may also decide to keep your base minimal and set up at least one new environment that contains more packages.

In order to work with Python scripts, you will need a development environment. We will emphasize Jupyterlab as well as integrated development environments, particularly Spyder. Miniforge environments do not include either by default, so they must be installed.

conda activate yourenv
mamba install jupyterlab 

In an environment such as an HPC cluster, you may need to install spyder into each environment much like jupyter-lab

conda activate yourenv
mamba install spyder

The applications may be started directly from a Miniforge shell

jupyter-lab

This will open a tab on your default browser.

or

spyder 

Spyder and JupyterLab Apps

If using a system you control, you may wish to install Jupyterlab and Spyder as standalone applications. You may need administrator privileges for these installations.

Using a standalone application can save some space and may be more convenient and less prone to conflicts.

For Jupyterlab, install the Jupyterlab Desktop. The installation method depends on your operating system but should be similar to other software.

Download the appropriate installer from their site and follow the instructions to install. It will insert a launcher into your desktop environment from which it can be started. This could be the Applications menu for Windows, or the Launcher for Mac OS or Linux.

To use the Jupyterlab Desktop, the jupyterlab package must be installed in the environment. The nbconvert package is recommended as well.

The Spyder standalone application can be downloaded and installed from its Website. It will attempt to autodetect your operating system and offer the correct version for download, but other versions can be chosen. The Spyder standalone application requires that at least the ipython package must be installed into an environment. Installing the jupyterlab and nbconvert packages should be adequate to use both applications.