wildtreetech / epfl-osip-2017 Goto Github PK

Collecting ideas for an afternoon workshop on "open science aspects of research code".

Time slot: 2pm - 6.30pm with a 60min break for a talk on IP.

• participants have whole range of skills and background in different disciplines
• not hardcore coders but researchers
• git, github, continuous integration, documentation generation
• jupyter notebooks
• there are lots of tools which ones should I use?
• recommend a cookie-cutter repo to start from to reduce overhead
• Gael's "premature software engineering" list
• pick python for concreteness but make sure to highlight transferable parts

From talking with @mdeff:

I think that Python / Jupyter notebooks and tools for collaboration and
publishing are the most important topics. We can stick to that. Short
sessions is an interesting idea. To maximize engagement, I was thinking
of having one project for the whole afternoon, e.g. you start with a
small analysis in a notebook (or some code the participants bring), add
versioning, go to a script, introduce tests, add generated
documentation, and finish the day with a github repo.

Learning goals. As a participant I will learn ...

• ... how to setup a "virtual env" for my project and be able to name advantages of doing so for each new project
• ... what a linter does and how to connect it to the Atom editor
• ... how to use jupyter notebooks to perform interactive exploration
• ... how to package my data sanity checks so that they are easy to run
• ... how to automate running my data sanity checks
• ... how to use git and GitHub as a "dropbox" for my code
• ... be able to name some of the tradeoffs between just putting my code out there and trying to create a open-source project with external users
• ... to recognise when code should be moved from my notebook to a module

Students should think about a project they would like to "open up".

We will ask them to write a README and a ROADMAP, etc and all this is much better if you have a project you care about instead of Tim's fake bike counting project.

Projects don't have to be science! Could be an event, or something as well.

best practices according to #binder workshop participants

https://github.com/wildtreetech/advanced-comp-2017/blob/master/install.md

need slight modification for this workshop but should be good to go.

Add conda env creation and activation commands to the material