codecov at 0%

[joshualerickson]

Can't seem to figure this out... Not sure why codecov is only picking up on a few lines of code? I'll keep working on it in another branch but help needed for sure!

[zackarno]

ok, i'll read up a little bit to see what I can do to help

[zackarno]

@joshualerickson - did some tinkering and have a couple leads - not sure if I fully understand, but have gained some insight

I think it's likely due to the testthat.R file and the skip_if_no_pypkg function in there. I believe this function allows the test to be skipped in github actions if the python packages are not installed. However, this is is also causing covr::package_coverage() to skip the tests due to the following line (at least on my computer) in the function:

have_numpy <- reticulate::py_module_available("numpy2")

This returns FALSE. If I switch "numpy2" to "numpy" it returns TRUE and the tests are run. I'm not really sure what "numpy2" is and a quick google search did not show too much on it. I had basically copied the testthat.R file from rgee repo to get github actions working initially so unfortunately do not have an understanding on where "numpy2" comes from.

After changing "numpy2" to "numpy" everywhere it occurs in the testthat.R file I received another error:

Python specified in RETICULATE_PYTHON (/usr/bin/python3) - this seems to be due to the first 2 lines after loading libraries where we call Sys.setenv(). I changed both paths to my local python environment (from tools-> global options-> python ): C:/Users/zack.arno/AppData/Local/r-miniconda/envs/rgee/python.exe . After doing this covr::package_coverage() works as expected.

Still trying to understand any potential consequences of these changes, but seems to be moving in the right direction

[joshualerickson]

Hey @zackarno, this is a really good lead! Looked into rgee repo and looks like they dropped test CI about a year ago in this commit. I'm not completely sure why but maybe because it was a bit cumbersome...

Couple options;

1). Just drop the CI. Pros are we don't have to deal with the time it will take to run each time but cons are we have to remember to run tests locally before pushing.

2). Figure it out and run it once so we get a 'code coverage' and then back to (1). This seems to be what rgee did.

3). Figure it out and try to keep it running on CI. Big issue from what I can tell is time (has to install python env every time)!

I think (2) would be nice to pursue. (3) would be great but not sure how much time we should spend on it...

Let me know what you think.

Thanks Zach.