A large proportion of the extract code is set up to vectorise 3D arrays. This is a legacy of the Matlab code (evidence from using Fortran-style indexing everywhere). To clean up the code, reduce the volume and make it more readable it maybe as efficient to create small subroutines of purely NumPy array calculations and decorate these with NUMBA's @jit
A large proportion of the extract code is set up to vectorise 3D arrays. This is a legacy of the Matlab code (evidence from using Fortran-style indexing everywhere). To clean up the code, reduce the volume and make it more readable it maybe as efficient to create small subroutines of purely NumPy array calculations and decorate these with NUMBA's @jit