WikiCite 2017/Jupyter notebooks on Wikimedia sites

From Meta, a Wikimedia project coordination wiki


Jupyter notebooks (en, d) are a way to share computational and data science workflows, and they have begun to be referenced on Wikimedia projects, e.g. in the English Wikipedia article on Jacobi elliptic functions. In this project, we are looking at making them reproducible, e.g. as in this example that anyone can run from basically any device.

PAWS is a service run on Wikimedia Labs that allows to run Jupyter notebooks. Here, we are looking into running Jupyter notebooks referenced from Wikipedia et al. on PAWS infrastructure. Before we can do that, we want to make sure that the notebooks actually run through properly, and a previous attempt to replicate a collection of Jupyter notebooks did not look promising in this regard (see an initial write-up).

We are now looking into automating the testing of Jupyter notebooks in terms of whether they run through, and what the errors are if they don't.

A first prototype of such a validator tool has been created at the WikiCite 2017 hackathon by Laurentius.