Git, testing and automation for research¶
Research code and documents go wrong in quiet ways: a number typed by hand, a figure from last month's data, a script that only ran on one laptop. This guide explains the tools that catch those faults, starting from nothing.
It's based on what I built for my own work, including real mistakes these tools caught. You can see them working in the pamphlet project and its build guide.
What you will learn¶
-
Git records every version, so that you can go back to any point.
-
GitHub shares work, and lets people report problems and suggest changes.
-
Tests check that the code still gives the right answers.
-
Continuous integration runs those checks by itself on every change.
-
Pipelines rebuild everything from the sources, so that nothing drifts apart.
-
Backups and releases protect what matters.
Prefer one long page? Read the whole guide on one page. The example files include everything shown here.