Planned Future Development¶
Development of NodePy is based on research needs. The following is a list of capabilities or features that would naturally fit into the package but have not yet been implemented.
- Time-stepping for multistep methods
- Selection of startup method
- Variable step size multistep methods
- Properties of particular multistep + startup method combinations
- Adaptive step size and order selection
- Time stepping for implicit methods
- Interpolants (dense output)
- Adaptive order (extrapolation and deferred correction)
Many common semi-discretizations of PDEs will be implemented as \(ivp\) objects. Initially this will be implemented purely in Python and limited to simple 1D PDEs (e.g. advection, diffusion), since time-stepping for multi-dimensional or nonlinear PDEs will be too slow in Python. Eventually we plan to support wrapped Fortran and C semi-discretizations.
- Additional classes of multi-stage, multistep methods.
- Analysis of geometric integrators.
- Partitioned rooted trees, additive and partitioned RK methods.
- Unit tests.
- An automatically-generated encyclopedia of solvers.