Open peterdesmet opened 4 months ago
This is a difficult one:
On the one hand, having a uniform repo structure (using cookiecutter for example) would greatly improve each repository's readability. On the other hand, making it a requirement (i.e. 'MUST'), would probably be more restrictive than necessary. Moreover, since some legacy packages are in the scope of B-Cubed, you risk being 'non-compliant' over a style thing.
We could think of a bare minimum set of files (given that we already require a CODE_OF_CONDUCT.md, LICENCE....) and folders (data, test, ...).?
Another option could be to create 'B-Cubed'-templates for each of the programming languages and require people to start from these?
Our guide suggests a repo structure for Python code https://docs.b-cubed.eu/dev-guide/#python-repo-structure
Similarly, for (R and Python) analysis, we could suggest https://cookiecutter-data-science.drivendata.org/
@mtrekels do you think we should add any of this as a requirement? At minimum, I think data should be placed in a
data
directory (not the root).