Open TimTaylor opened 3 months ago
I'm unsure if the GitHub api allows you to inspect the directory structures and avoid a shallow clone
Yes, this is technically possible
We had a related discussion yesterday. In particular with the example of data.table which doesn't use a testing framework.
I think it then potentially opens the question of "is it still best practice?" :thinking:. I chatted with data.table contributors recently who shared how the unusual infrastructure made it more difficult for them to contribute.
My view is probably that:
no tests < tests with no testing framework < testing framework
On a slightly more general note, "best practices" is bound to be somewhat subjective. I tried to resolve this by linking to what we use as reference but I believe some room will always be left for interpretation.
On a slightly more general note, "best practices" is bound to be somewhat subjective. I tried to resolve this by linking to what we use as reference but I believe some room will always be left for interpretation.
Yeah I appreciate you've tried to avoid judgement in the analysis and I agree from a contribution perspective a framework tends to make it much easier.
That said the output table could possibly be split to have both "Uses testing" and "Uses testing framework" percentages?
... Maybe :thinking:
Not sure how feasible this is without shallow cloning from https://github.com/cran but the idea is that you only check for a non-empty
tests/
directory. I'm unsure if the GitHub api allows you to inspect the directory structures and avoid a shallow clone :man_shrugging: