In our previous post we learned about Python virtual environments and how conda makes it easy for us as scientists to use virtual environments and access hundreds of useful packages.

Despite the number of packages directly available through conda, not all packages can be downloaded this way. In this post we will learn how to add such packages to our Python virtual environments.

What if conda cannot find the package I am looking for? Conda channels

Thankfully, there is a large community of Anaconda users that help fill the gap by providing different channels from which we can download our desired package. Think of channels as branches of the Anaconda library. The same Python packages can be contained in several channels; pick a channel that contains the version of the package that is compatible with your system (32-bit vs 64-bit; linux, win, os-x), preferably one that has been downloaded many times as this provides some reassurance that the package works reliably.

As an example, lets pretend we wanted to include the pygame package to our environment. Note that we start by activating the virtual environment we created in our previous post – it is called sci_sound.

(base) $ conda activate sci_sound
(sci_sound) $ conda install pygame
Collecting package metadata: done
Solving environment: failed

PackagesNotFoundError: The following packages are not available from current channels:

  - pygame

Current channels:

  - https://conda.anaconda.org/conda-forge/linux-64
  - https://conda.anaconda.org/conda-forge/noarch
  - https://repo.anaconda.com/pkgs/main/linux-64
  - https://repo.anaconda.com/pkgs/main/noarch
  - https://repo.anaconda.com/pkgs/free/linux-64
  - https://repo.anaconda.com/pkgs/free/noarch
  - https://repo.anaconda.com/pkgs/r/linux-64
  - https://repo.anaconda.com/pkgs/r/noarch

To search for alternate channels that may provide the conda package you're
looking for, navigate to

    https://anaconda.org

and use the search bar at the top of the page.

As you can see, pygame is not available for direct download. But if we navigate to https://anaconda.org and search for pygame, we see the following search results:

If we click the most popular channel, called CogSci, we see the following information:

We need to run conda install -c cogsci pygame to install pygame. Let’s do that!

(sci_sound) $ conda install -c cogsci pygame 
Collecting package metadata: done
Solving environment: done

## Package Plan ##

  environment location: /home/martin/anaconda3/envs/sci_sound

  added / updated specs:
    - pygame


The following packages will be downloaded:

    package                    |            build
    ---------------------------|-----------------
    numpy-1.16.3               |   py27he5ce36f_0         4.2 MB  conda-forge
    pip-19.1                   |           py27_0         1.8 MB  conda-forge
    setuptools-41.0.1          |           py27_0         650 KB  conda-forge
    ------------------------------------------------------------
                                           Total:         6.7 MB

The following NEW packages will be INSTALLED:

  pygame             cogsci/linux-64::pygame-1.9.2a0-py27_0

The following packages will be DOWNGRADED:

  certifi                                   2019.3.9-py37_0 --> 2019.3.9-py27_0
  numpy                               1.16.3-py37he5ce36f_0 --> 1.16.3-py27he5ce36f_0
  pip                                           19.1-py37_0 --> 19.1-py27_0
  python                                   3.7.3-h5b0a415_0 --> 2.7.15-h721da81_1008
  setuptools                                  41.0.1-py37_0 --> 41.0.1-py27_0
  wheel                                       0.33.1-py37_0 --> 0.33.1-py27_0


Proceed ([y]/n)? 

Who-oh! That is more than we bargained for. Do you see what went wrong? The package that we were going to download is listed as:

pygame             cogsci/linux-64::pygame-1.9.2a0-py27_0

The key thing to notice is the py27_0 part. It means the pygame package we were about to download was written to work with Python 2.7. As you may or may not known, the creators of Python made a breaking change from Python 2 to Python 3. That means that packages coded using Python 2 often won’t work when they are run using Python 3. This is why, to create a virtual environment that is compatible with the version of pygame we asked conda to download, conda is telling us that it has to downgrade several packages, including Python! Specifically, conda wants to downgrade Python from version 3.7.3 to version 2.7.15.

No one from the Anaconda community has included a version of pygame in one of the channels that is compatible with Python 3. While we could change everything over to Python 2.7, this is not a great solution.

Is there a way to easily obtain a version of pygame that is compatible with Python 3?

What if conda cannot find the package I am looking for? Pip install

pip is a package-management system used to install and manage software packages written in Python. Many packages can be found in the default source for packages and their dependencies — Python Package Index (PyPI). The nice thing of working with Anaconda is that, once we have activated our Python virtual environment, conda is aware of pip and packages that are installed using it.

Thus, to install a Python 3 compatible version of pygame, we can run a simply pip command:

(sci_sound) $ pip install pygame
Collecting pygame
  Downloading https://[...]/pygame-1.9.6-cp37-cp37m-manylinux1_x86_64.whl (11.4MB)
     |████████████████████████████████| 11.4MB 20.6MB/s 
Installing collected packages: pygame
Successfully installed pygame-1.9.6

Conclusion

In this post we learned how to look for a Python package that is not directly available throught the conda install command. Specifically, we learned about https://anaconda.org and how to search and (almost) install packages available in conda channels. Finally, we learned how to install Python packages using the pip command. In our next post, we will review how to create a new conda environment, and learn how to see what environments exists on our system, see what packages are install in each environment, and also delete unwanted environments.