setting up a conda environment

Note

This page provides instructions for creating a new conda environment using Anaconda Navigator, or using the Anaconda Command Prompt.

You only need to choose ONE of these methods for setting up the environment.

Danger

You should have already opened a GitHub account, installed git and GitHub Desktop, forked the EGM722 repository, cloned the repository to your computer, and installed Anaconda Navigator.

If you haven’t done all of these steps, please do so now before continuing.

anaconda navigator

creating an environment

From the Start menu, open Anaconda Navigator. When it opens, click on the Environments tab on the left-hand side of the screen. You should see the following:

As mentioned above, conda is a package management system. We can use conda to create different programming environments, which will enable us to keep track of and manage the specific versions of python packages that we use for this module. It also enables us to easily share our environment specifications across different computers, so that we can be sure that any processing steps that we do are consistent and reproducible.

We will be creating a new environment using the environment.yml file provided in the git repository. A .yml file is one of the ways that we can easily duplicate environments using conda or other package management systems – it provides a list of the packages for the package management system to find and install.

Open the environment.yml file in a text editor (NB – this means NotePad, Notepad++ or something similar, NOT MS Word!). It should look something like this (the exact list of dependencies may differ somewhat):

Here, you can see the name of the environment (egm722), the channels to install packages from, listed in order of preference, and the dependencies, or required packages.

From the Anaconda Navigator window, click on the Import button at the bottom of the Environments tab. An import window will open:

Click the folder icon to navigate to the environment.yml file, select it and click Open. The window should now look like this:

The Name field may not automatically populate based on the .yml file, so be sure to check that it is egm722. The Location will differ depending on where Anaconda is installed on your computer.

Click Import to start setting up the environment. Depending on your connection speed and computer’s specifications, this may take some time. When it finishes, you should see the following screen:

Note how many packages were actually installed – 161, from the 5 shown in the original .yml file.1

This is because each of those 5 packages have additional dependencies which have to be installed as well. Fortunately, almost all of this is done automatically – we don’t have to worry about tracking down each individual dependency on our own.

changing environments

The next thing to do is to switch to the Home tab in Anaconda Navigator:

You should see that you have two different environments when you open the dropdown menu next to Applications on. Whenever you want to switch environments in Anaconda Navigator, you can select the environment from this menu.

You should see that when you change between the base and egm722 environments, the list of installed applications changes – this is because we have not installed the same applications in the different environments.

setting up the command prompt

The final step covered in this manual is to install the CMD.exe Prompt for this environment:

While not strictly required, this will enable you to directly launch a Windows Command Prompt with your egm722 environment already loaded.

If you run the Anaconda Prompt from the Start Menu, it will automatically load the default (base) environment, and you will need to switch environments using the conda activate command when you want to use your egm722 environment.

Once you have the conda environment set up, you are ready to move on to configuring jupyter.

anaconda command prompt

From the Start menu, open the Anaconda Command Prompt:

When you open the command prompt, you should see (base) next to the prompt:

(base) C:\Users\bob>

Warning

The Anaconda Command Prompt automatically loads conda so that you can use it; if you don’t see (base), it likely means that you have opened the normal Command Prompt, and so you will see the following error when you try to run a command with conda:

Next, navigate to where you have cloned the repository using cd:

cd c:\Users\bob\projects\egm722

If you’re in the right place, you should see all of the files from the repository when you enter the dir command:

Now, enter the following command:

conda env create -f environment.yml

This tells conda to create a new environment, using the “recipe” provided in environment.yml. You should (eventually) see something like the following:

This step may take some time2, but if all goes well, you will eventually see the following message:

And that’s it. Once you have done this, you may want to open Anaconda Navigator and set up a new command prompt link for your environment, following the instructions in setting up the command prompt.

Once you have the conda environment set up, you are ready to move on to configuring jupyter.

notes and references

1

The exact number of packages may differ - the point is that we have only told conda to install a few of these, and conda has figured out the rest.

2

Conda can be notoriously slow, which is why there has been an effort to increase the speed/efficiency of the package “solver”: mamba. I am happy to help you get set up with mamba, but I am unsure about how well this works with Anaconda Navigator.