Unit Testing for Reproducible Science

Overview

This seminar teaches the basics of unit testing and how to use unit testing for reproducible research.

Seminar Topic

A unit test is a separate piece of code that exercises your code to determine if it’s functioning as expected. With unit tests in place, it’s possible to isolate changes in your results to changes to your input data or changes to your model. Without unit tests in place, it’s hard to determine if changes in your results are a side effect of the code not doing what you expect it to do rather than a “legitimate” result.

Beyond it’s role in reproducible research, unit testing helps you write better code by forcing you to think through edge cases, factor your code into easily-testable units, and ensure that code changes don’t break existing functionality.

In this seminar, you will learn what unit testing is, why it’s important, and how to write (and run!) unit tests for your code.

Audience

The seminar is intended for researchers and scientists, and it assumes that most of your code is developed in Jupyter notebooks. If that’s not you, feel free to skip the lesson on testing Jupyter notebooks.

How to use this repository

First, create a fork of this repository on GitHub here and then clone your fork locally with:

git clone [email protected]:{username}/unit-testing-seminar.git

To build the docs and run the examples included in this seminar, you need python3.7 and you need to install a few dependencies. Create a virtual environment and install the required dependencies into it.

If you have pipenv installed, you can run:

$ cd unit-testing-seminar
$ pipenv install
$ pipenv shell

If you don’t have pipenv, you can install the dependencies into any other virtual environment by running:

$ cd unit-testing-seminar
$ pip install -r requirements.txt

Either of these will install all of the dependencies for you. Now, you can build the docs locally and run the examples. To build the docs:

$ sphinx-build -b html _source _build
$ (cd _build; python -m http.server 8000)

Now you can navigate to localhost:8000 in your browser and see the built doctumentation.