In order to keep answers from changing when I did a massive C++ refactoring of fields, I added a hack where the U/V fields had a second mask (the true u/v mask) added to the atlas metadata (this was in addition to the interpolation mask, which was the T mask). So that @fmahebert can make progress with regard to having a single "mask" metadata that the generic algorithms can use, I have removed this second mask.
This only impacts the print statements for U/V fields, not any real results.
This is, again, a bit of a hack, U/V fields should not be using the cell center mask, but this will all get sorted out when u/v de-staggering is re-implemented in the variable changes.
Due to these changes, the convert state application should not be used to do a resolution change with uocn and vocn (or you could still use it, and make sure your results don't have missing values in them, the problem at this point is illegal floating point operations where the u/v and t masks don't line up... fix that some other day)
Testing
The reference values for u and v prints for all tests had to be updated
Description
In order to keep answers from changing when I did a massive C++ refactoring of fields, I added a hack where the U/V fields had a second mask (the true u/v mask) added to the atlas metadata (this was in addition to the interpolation mask, which was the T mask). So that @fmahebert can make progress with regard to having a single "mask" metadata that the generic algorithms can use, I have removed this second mask.
This only impacts the print statements for U/V fields, not any real results.
This is, again, a bit of a hack, U/V fields should not be using the cell center mask, but this will all get sorted out when u/v de-staggering is re-implemented in the variable changes.
Due to these changes, the convert state application should not be used to do a resolution change with uocn and vocn (or you could still use it, and make sure your results don't have missing values in them, the problem at this point is illegal floating point operations where the u/v and t masks don't line up... fix that some other day)
Testing
The reference values for u and v prints for all tests had to be updated
run-ci-on-draft=true jedi-ci-test-select=clang