RSH 18: Why is software, data, etc. important?

Video

Collaborative notes taken during the session

Outline

intro to RSH

• Who are we?
• What is RSH
• Who are we for? And who is this for?
• About our 2021 program: https://researchsoftwarehour.github.io/

Intro to today

• What we are doing today
• Who do we imagine you to be? (what does each person get?)
  • You may be a new or experienced researcher, working on things.
  • You may have been directed here to get an overview of what you can improve. You may be working with someone who will guide you through this
  • You may be a supervisor, who needs to ensure your students do good work

The Zen of Scientific computing

• For each topic:
  • Introduce it
  • Give an example of it going wrong
  • Give an example of it going right
  • Discussion
  • Where do we recommend someone go next?

Future

• About rest of RSH
• What's your future?

Questions/comments

Hey, yes it is good with all three of you :-)

• Oh yes! I guess my coding style is far from industry standard XD and difficult to understand for anyone including me in the future ....

  • haha. welcome!

• Nice paper: "Good enough practices in scientific computing" https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1005510

• do your arguments change if it is a one-time-code which is never published but you only describe its functionality in the paper? (including sanity tests using theoretical arguments) and which cannot be refactored in a general tool? And it is only used by one person?

  • hmmm.... great question! I think I would call this research/test code (now that I understood the definition better)
  • a one-time-code? I have never seen a code you use only once.
  • this is the general theme of our research: you have a specialized problem where you program a tailored tool for. But it is not really adapting to more than this single scenario, even increasing parameters implies a combinatorial explosion

• When does research code become production code and how?

  • as soon as it starts being used for research/publication? (unsure)
  • "Experiment(al) code"?

• Also reminded me of https://xkcd.com/2347/

• Any "Big data effort"... run once, almost impossible to run again

  • is it because of the data part? or the workflow part? or the compute resources needed?
  • a bit of everything I'd say. Often things change halfway. The cost of recomputing is proportionally larger. Debugging is equally harder. A reproducibility challenge.

• Single person project - "collaborating with your past self"

  • oh yes that's a great description

• Version control: add new level, web service for just yourself

• Packaged and published

  • which is many layers. But it helps to use standard solutions for this instead of inventing a new one.

• Where do you place tools like mypy? static code evaluators?

  • automated testing and also modular/reusable code development