GSoC 2020 Work Product - Baseline Images Problem

Google Summer of Code 2020 is completed. Hurray!! This post discusses about the progress so far in the three months of the coding period from 1 June to 24 August 2020 regarding the project Baseline Images Problem under matplotlib organisation under the umbrella of NumFOCUS organization.

Project Details:

This project helps with the difficulty in adding/modifying tests which require a baseline image. Baseline images are problematic because

So, the idea is to not store the baseline images in the repository, instead to create them from the existing tests.

Creation of the matplotlib_baseline_images package

We had created the matplotlib_baseline_images package. This package is involved in the sub-wheels directory so that more packages can be added in the same directory, if needed in future. The matplotlib_baseline_images package contain baseline images for both matplotlib and mpl_toolkits. The package can be installed by using python3 -mpip install matplotlib_baseline_images.

Creation of the matplotlib baseline image generation flag

We successfully created the generate_missing command line flag for baseline image generation for matplotlib and mpl_toolkits in the previous months. It was generating the matplotlib and the mpl_toolkits baseline images initially. Now, we have also modified the existing flow to generate any missing baseline images, which would be fetched from the master branch on doing git pull or git checkout -b feature_branch.

Now, the image generation on the time of fresh install of matplotlib and the generation of missing baseline images works with the python3 -pytest lib/matplotlib matplotlib_baseline_image_generation for the lib/matplotlib folder and python3 -pytest lib/mpl_toolkits matplotlib_baseline_image_generation for the lib/mpl_toolkits folder.


We have written documentation explaining the following scenarios:

  1. How to generate the baseline images on a fresh install of matplotlib?
  2. How to generate the missing baseline images on fetching changes from master?
  3. How to install the matplotlib_baseline_images_package to be used for testing by the developer?
  4. How to intentionally change an image?


I am grateful to be part of such a great community. Project is really interesting and challenging :)

Thanks Thomas, Antony and Hannah for helping me to complete this project.