Preventing DeReLiCT code

Oftentimes, especially when doing research, we begin the coding portion of a project without that much planning.

In some projects, coding is an analysis tool: it lets us investigate data, perform statistical calculations, and create plots and visualisations. We might write snippets of code in an interactive notebook1 and quickly prototype. These snippets (or a tidied-up subset) might end up shared in supplementary information alongside a published research article. In other situations, code might be a more foundational part of the project, for example when building a numerical model. This code may be written more formally, and you might expect it to be shared or used more widely. Another case is where code is a tool for sharing and presenting our results: this can be again for building visualisations (but this time, publication-worthy, polished visualisations) or can scale up as far as web-apps or dashboards.

The requirements for each of these situations is quite different, and it can be very challenging to manage code across these: in many situations, you might need to dip in to more than one of these categories.

In part 2 of this course, we’ll dive in to strategies for setting a project up for success from the start, but in this section we are going to look at tools and approaches that you can apply to an already-existing, stressfully messy project. While good planning is fantastically useful, and of course on completing this course you’ll implement it going forward, it’s never too late to add some tools to your project to help it run more smoothly.

DeReLiCT Code

Because I love an acronym, and because I find having a checklist makes things much less overwhelming, I’ve distilled down what I think are the key features of software development into five categories:

  1. Dependencies
  2. Repository
  3. License
  4. Citation
  5. Testing

Ten minute SOS

Help! I’ve only got 10 minutes to improve my code

We know the feeling. If your code is falling down around you and you need some quick help right now, then let’s just look at these two points:

1. Dependencies

Write down a list of the libraries you’ve used in your code in a requirements.txt file.

If you have a minute more, let’s set up a quick virtual environment with conda - if you have Miniforge installed already, you can do this in five minutes. See a walkthrough here: “art of conda” blog post.

2. Repository

Let’s initialise a git repository in your working folder. You’ll need to have git installed on your system; from your project directory, run the following from the command line:

git init

If this sounds nonsensical right now, do not fear! It might take a little bit longer than ten minutes, but if you read through the Dependencies and Repository section you will understand better what’s going on in both these sections!

Back to top

Footnotes

  1. Interactive notebooks, such as Jupyter notebooks for Python and R, allow for code and code output to be embedded in a file alongside text, images and other content. Notebooks are often used for rapid prototyping and plotting, as you can quickly see results alongside your code.↩︎