pandas visualisation

Plotting in pandas is very easy, mainly by using Series.plot() and DataFrame.plot() methods. There is a plotting subsystem in pandas based in matplotlib that implement different types of plots (e.g. lines, bars, boxplots, kde…).

In the last months, several new projects have been create to address new use cases for visualising in pandas. For example, to generate interactive plots. Some of these projects are:

• https://hvplot.pyviz.org
• https://github.com/PatrikHlobil/Pandas-Bokeh
• https://github.com/altair-viz/pdvega/

Those libraries have been monkey patching pandas to be able to plot easily, so plotting can be done by using DataFrame.hvplot(), DataFrame.plot_bokeh()…

But a better design would be to decouple the existing plotting code in pandas into an extension registered with the pandas extension capabilities, and be able to select with an option with plotting backend the user wants to use. The resulting code would be:

pandas.set_option('plotting.backend', 'hvplot')
df.plot()

With this architecture, there are several advantages:

• Developing new plotting backends for pandas becomes much simpler
• Plotting backends share a common API
• Users of pandas don’t need to learn a new syntax for each plugin
• Migrating existing code becomes trivial, by just adding a single line of code setting the option for the backend
• Internal pandas code becomes cleaner, with the plotting code fully decoupled from the rest

Work on this is already going on, with a first PR that decoupled the current plotting code:

https://github.com/pandas-dev/pandas/pull/26414

In this sprint we will continue the work, by working on different tasks:

• Adding the option to select the backend
• Defining and documenting the pandas plotting API
• Porting existing libraries to the new plotting API
• Document current pandas functionality

We also can work on other pandas issues, for example:

• pytest-azurepipelines: https://github.com/pandas-dev/pandas/issues/26601
• Documentation warnings: https://github.com/pandas-dev/pandas/issues/24173 / https://github.com/pandas-dev/pandas/issues/24661
• PandasArray in API docs: https://github.com/pandas-dev/pandas/issues/24507
• pyarrow failure: https://github.com/pandas-dev/pandas/issues/26656
• Organise dependencies: https://github.com/pandas-dev/pandas/issues/26659
• Remove isort dependency: https://github.com/pandas-dev/pandas/issues/26661
• Feel free to propose yours

As usual, we will give priority to join the sprint to the next people:

• Experienced open source contributors
• People from underrepresented minorities in our sprints

Agenda

• 6pm: Food and networking
• 6:30pm: Presentation of the project and the sponsor
• 6:45pm: Coding
• 9pm: Retrospective presentations

The day of the sprint

• Bring your own laptop if you can
• Join the Gitter channel of the sprint

Thanks to J.P. Morgan for hosting this event!

J.P. Morgan Chase is one of the biggest and most successful Banks in the world, but equally importantly, J.P. Morgan is one of one of the world’s biggest tech companies. In 14 technology hubs worldwide, our team of 50,000+ technologists design, build and deploy everything from enterprise technology initiatives to big data and mobile solutions, as well as innovations in electronic payments, cybersecurity, machine learning, and cloud development. Our $10B+ annual investment in technology enables us to hire people to create innovative solutions that will not only transform the financial services industry, but also change the world.