Overview

Analytic code can be divided up into logical units. You might have a script which prepares and cleans data, and another which outputs a summary descriptive table.

In OpenSAFELY, each logical unit is called an action. Actions can be scripts, Jupyter notebook generators, or specialised functions provided by the framework.

An OpenSAFELY project must refer to its actions in a pipeline. This is a file which defines all the actions in a project, how they should be run, and how their outputs should be saved.

Dividing your analysis into actions and describing them in a pipeline has a few purposes:

• It aids readability and bug-finding
• It allows common code to be reused without copy-and-paste
• The OpenSAFELY pipeline system ensures your actions run efficiently and quickly
• It allows reviewers to see your intent and check your code for privacy and security accordingly