Currently, codecov does not account for numba decorators, leading to a lower coverage than expected. This is because Numba code is compiles, and hence cannot be analyzed by the codecov tool.
This is a known and unsolved problem with Codecov and Numba. link to open issue
Solutions are:
1) Disable Numba JIT during pytest stage. I believe this is unfeasible since the Kalman filtering would be too slow ( > 1 hr )
2) Ignore numba jit functions in the coveragerc file. This is a workaround that ignores these decorated functions, and does not include them in the coverage calclations.
This is the main cause of low coverage in kalman.py, data.py and solve.py
Currently, codecov does not account for numba decorators, leading to a lower coverage than expected. This is because Numba code is compiles, and hence cannot be analyzed by the codecov tool. This is a known and unsolved problem with Codecov and Numba. link to open issue Solutions are: 1) Disable Numba JIT during pytest stage. I believe this is unfeasible since the Kalman filtering would be too slow ( > 1 hr ) 2) Ignore numba jit functions in the coveragerc file. This is a workaround that ignores these decorated functions, and does not include them in the coverage calclations.
This is the main cause of low coverage in
kalman.py
,data.py
andsolve.py