Hi @nperez
The huge "transform + chunk" (plus many accompanying errors) development is still ongoing, but I open the merge request now to leave the record here.
-
transform + chunk, lat/lon selector is list of values -
transform + chunk, lat/lon selector is vector of indices -
transform + chunk, lat/lon selector is vector of values -
transform + chunk, lat/lon selector is list of indices -
transform + chunk, lat/lon selector is 'all' -
transform + chunk, lon selector is goes_across_prime_meridian (e.g., the original lon is [0, 360] but CircularSort(-180, 180)) *NOTE: Return error message. The current algorithm doesn't support this case. -
The consistency between lat/lon selector is list of values and vector of values (for both transform and no transform) -
The consistency between lat/lon selector is list of indices and vector of indices (for both transform and no transform) -
transform + chunk, the number of transform_extra_cells -
transform + chunk, the crop options: Region, TRUE, and FALSE (TRUE has problem now) -
reordering, lon selector is list of indices -
reordering, lat selector is indices in a reversed order (e.g., c(640:1)) -
reordering, lat/lon selector is 'all' -
transform, lat/lon selector is indices, parameter aiat(I'd like to leave it for future development)
The usage of "vector of values" and "list of indices" is rare, especially "vector of values", which doesn't make much sense to me. But I'll still check if the logic is right.
Another issue is, we have more heavy unit tests now (because they use Compute() and the global domain is tested), the pipeline kept failing even if R CMD check was finished. It failed at covr::package_coverage(). The log file didn't show any errors. I ran all the unit tests and covr::package_coverage() at workstation and they all passed (very time-consuming though). I'm not sure if the failure is because the tests are too heavy, but the pipeline passed after I removed covr::package_coverage() in .gitlab-ci.yml . In my opinion, the coverage check for startR functions is not important so it's okay to remove it. Please let me know if you have a better solution, and also any suggestions for these developments. Thanks!
Cheers,
An-Chi