Virtual environments in Python using conda

Explains why virtual environments are essential when working on scientific projects and how to manage them in Python using conda.

January 30, 2024

What is a virtual environment?

One can think of an virtual environment as a self-contained space that has its own Python interpreter and packages. If you install a new package from within the environment, the package and the dependencies are installed only for this environment. Once you leave the environment the package is not available. Also if you update Python or any other package within the environment, the versions outside of the envirvonment remain unchanged. Conversely, updates made outside the environment do not affect the packages and Python versions within the environment.

When conducting a scientific project using Python, using such virtual environments makes a lot of sense. In fact, it is usually a good idea to create one virtual environment for each or your projects.

Why using virtual environments for scientific projects

At first glance this might sound stupid: why would you even invest time to set up an environment that then does not ‘benefit’ from any updates you install outside of it? The main reason is stability and persistence of your code. Packages evolve over time and some updates require you to change your code. This means that old code might stop working after an update has been installed. For scientific projects this can be a huge pain because for you it is usually not relevant whether the study you conducted two years ago works with the Python version that is currently the state of the art. What matters is that you can still replicate the study today and get the same results as you did two years ago. If you have used a virtual environment for your study, you just enter it now and run your code and you get the same results. If you didn’t you might need to change the code to comply with new package and Python updates. This can be terribly frustrating, so saving you from this calamity is sufficient reason to use virtual environments right from the start - especially because it is really easy.

How to create a new virtual environments for a new project

There are two main approaches to create and organize virtual environments, conda and venv. This post focuses on conda. The steps needed to manage virtual environments with venv are covered in a separate post.

Assume you created a working directory and a VSC workspace. Before you begin, make sure conda and Python is up to date:

conda update conda
conda update python

Now we can start setting up the virtual environment. The first step is to decide which Python version we want to use for our project. Say we want to use Python 3.11. Then we would use:

conda create -n my-env python=3.11 anaconda

assuming that our environment should get the name my-env. You will need to confirm that the Python version and all needed packages will be installed within the new environment. The the environment is installed, but not yet activated. To activate it from the terminal you can write

conda activate minveb-abm-env

and to deactivate and to return to your base environment just do

conda deactivate

As indicated above, if you install packages while the environment is active the packages will be installed within this environment only. If you want to install them into the environment ‘from the outside’ you can do this via

conda install -n [env-name] [package]

Without the part -n [env-name] the package will be installed into the currently active environment. If you have not activated a specific environment, this will be the base environment.

Finally, we can verify whether the environment exists:

conda info --envs

Here the new environment should now appear. It is now ready to be used. Note that the currently active environment is marked with an asterix. This should currently be the base environment. So to use the new environment you still need to activate it.

Use the environment for your project

One way to use the environment is to make sure it is activated whenever you do something for your project. A better way is to tell your development tool to use it as a default. In VSC, for instance, you can choose the command Python: Select interpreter to select your new environment. Note that it only appears here if you specified an interpreter for it, i.e. used the python=x.x part of the command above. Also, directly after the creation environment might not appear. Do Developer: Reload Window first, then you should be able to see is.

Some further information

Here I collect some additional information that I found useful when dealing with virtual environments with conda:

How to delete an environment

If you want to delete an environment you do not longer need you can do this via

conda remove -n ENVNAME -all

but you should be careful with this.

Create environments with specified packages and package version

For obvious reasons you might want to specify the particular versions of the packages you use in a new environment. The syntax to do so is straightforward and follows the following pattern:

conda create -n myenv python=3.6 scipy=0.15.0 pandas

This creates a new environment called ‘myenv’ with Python version 3.6, the package scipi in version 0.15.0 and the package pandas in the most recent version available.

What if the environment cannot be registered?

In case you get a warning that the environment cannot be registered because the path .envs is not writable either make it writable using chmod or create the environment using admin rights, i.e. using the sudo command. The latter is not recommended. On macOS the writing allowance might be wrong, so you need to give you writing allowance:

sudo chown -R username /home/username/path-to-conda-folder/

sudo chmod -R +x /home/username/path-to-conda-folder/

Adjust the directories according to the warnings that are provided it you install an environment as described below.

How to save the environment in a .yml file

To save the versions of all relevant packages and to recover the virtual environment when needed you can save the relevant information in a file from which you can later re-create the environment anew.

To save the current environment in a file environment.yml do:

conda env export > environment.yml

Across platforms you may not want to include all packages, but only those that you installed explicitly (thereby excluding platform-specific dependencies):

 conda env export --from-history  > environment.yml

To create a new environment from an existing file environment.yml file simply call:

conda env create -f environment.yml

Note that the name of the new environment is determined by the .yml (first line).


BibTeX citation:
  author = {Gräbner-Radkowitsch, Claudius},
  title = {Virtual Environments in {Python} Using Conda},
  date = {2024-01-30},
  url = {},
  langid = {en}
For attribution, please cite this work as:
Gräbner-Radkowitsch, Claudius. 2024. “Virtual Environments in Python Using Conda.” January 30, 2024.