Open dbrakenhoff opened 1 day ago
But all notebooks (all TTim models) call both the besselnumba and invlapnumba routines. Wouldn't that be part of testing?
If these contain numba functions it could be that coverage is not recognizing them because they are compiled outside of python scope. I think a possible solution would be to use add the NUMBA_DISABLE_JIT=1
environment variable, but it might slow things down a bit.
besselnumba.py
andinvlapnumba.py
seem to be the cause of the low coverage at the moment:There is a test_bessel.py file in ttim/src/ but it compares fortran compiled bessel functions to their numba counterparts. Adapting this file to test the numba funcs against stored results would probably significantly improve coverage, and actually test the besselnumba code.