I just wanted to know if anyone has experience with testing ray backends with test coverage.
I know coverage is not always the best option to measure quality, however it is something which gives some hints (at least if you have coverage below 50%)
Due to the nature of ray just running tests with cov enabled does not give the desired results. Everything inside a ray backend is not covered according to the coverage metrics.
I tried to mock out ray already sometime before, to only test the function itself, however was not that successful at that time, and with the new deployment decorators it might be even more complicated. Also when the backend itself gets an handle on other backends it will again get more complicated
So my question is, did someone try something in that direction already (hopefully with success? ). Or do you have any ideas on how to approach that?