IFI/UZH Summerschool: From Scripts to Reusable Software and Reproducible Research

One of the key principles of proper scientific procedure is the act of repeating an experiment or analysis and being able to reach similar conclusions. In the optimal case, researchers should be able to reproduce results from papers that they read with very little effort. The reality is, however, often different. Most authors do not provide any source code leading to their findings, and even when they do, information is incomplete or does not reproduce the results from related works. The course provides actionable guidelines on how to convert non-reproducible scripts with hard-coded parameters into reusable software. We will exercise topics such as version control, documentation, unit testing and packaging. While most of the concepts are independent of the programming language, examples will be provided in Python.

Access course slides.

Access the course page.

Program:

• Motivation / Introduction (20 minutes, 5 minutes Q&A)
• Workflow for code organization (extensibility) (30 minutes, provide solution, 10 minutes Q&A):
  • Disentangling score production from analysis (tabbing and plotting)
• Command-line parameters (The Missing Semester CSAIL/MIT) (30 minutes talk, 30 minutes of hands-on + Q&A):
  • Argument Parsing with argparse
  • Configuration files: YAML
  • Producing tables (using package "tabulate")
  • Producing plots (using package "matplotlib")
• Version control (Git/GitHub) (30 minutes of talk, 10 minutes of hands-on and Q&A)
• Deployment (python packaging, for sharing) (30 minutes of talk, 30 minutes hands-on + Q&A)
• Software documentation (30 minutes, 30 minutes hands-on)
• Unit testing (20 minutes talk, no hands-on)