agu-ossi / 2018-agu-oss Goto Github PK

We have to decide which services/program/templates to use for the various parts in the example we develop. I assume that participants will just follow these but they are of course free to choose their own.

git Hosting

Possibilities: github, gitlab, bitbucket, ?

I guess github is the obvious choice here although I personally really like the fully integrated approach gitlab is taking.

Programming language for the example

Python 3.7?

License

MIT/BSD

There are good arguments against the GPL variants in the context of scientific software but we of course also have to mention these.

Code of Conduct

https://www.contributor-covenant.org/ but there are many options.

We should probably also mention the various controversial arguments that have popped up over the last years over code of conducts and so on.

Unit testing in Python

I feel like pytest is the only sensible options these days.

CI

travis CI, circle CI, ... gitlab also has a really powerful CI system but it works best when using gitlab

circle CI tends to finish a lot faster which might be useful. But I can live with either choice.

Documentation

readthedocs is probably the only option for hosting. For writing the documentation we could either using sphinx or mkdocs (both work on readthedocs).

We should have a license on this repo. I was thinking CC-BY for the text, etc, and BSD/MIT for the code. Any thoughts?

We should probably send out an email to all participants with instructions for setting up their computers. We don't want to waste a bunch of time creating accounts and installing Anaconda.

Hopefully AGU will provide a way to reach out to participants. The workshop system seems a bit weird. They don't show the names of the conveners on the website.

We should have a simple thematic example for the code that we write. I would suggest that we keep the software footprint small and rely only on numpy, scipy and matplotlib (plus dev requirements of pytest, sphinx)

What is an example that would be interesting and simple to test? We could go with a time-series analysis problem? Stack a signal, fourier transform and apply a filter, and transform back? (or even just one of those steps?)

I pasted our tentative schedule on the README. Do we need to revise this or will we keep it as is?

Checklist for things to bring to the workshop:

• Thumbdrives with Anaconda and git for Win/Linux/Mac

We have 4h 20min scheduled for the workshop. We should probably reserve 20min for introductions and setup. That would leave us with 4h to work with. How would you like to format the workshop?

The initial thought was to walk through setting up a repository with some code, tests, CI, and readthedocs.

Should we have this repo be the end product that they hope to achieve? We can doctor the history to have each commit be one of the lessons or use tags to mark the points at which they should be.