The work on the differential geometry module has not progressed much this week. I have fixed some minor issues, docstrings and naming conventions, however I have not done much with respect to the implementation of form fields as there are still some questions about the design to be ironed out.
Instead I focused on studying two of the features that SymPy presents and that I will use heavily. The first is the simplification routine that I will discuss another time. The second one is the different solvers implement in SymPy.
First of all, I have a very hard time getting used to the various output that the solvers provide. The algebraic equations solver, for example, can return either an empty list or a None instance if there is no solution. If there are solutions it can return a list of tuples of solutions, or a list of dictionaries of solutions, or a list of solutions if there is only one variable to solve for, or the solution itself if the solution is unique… Thankfully, a remedy for this was implemented by Christopher Smith. In pull request 1324 he provided some flags that force the solver to return the solutions in a canonical form. I am very grateful for his work and I hope that in the not too distant future what he has done will become the default behavior. I also hope that the solver will get refactored, because internally it is still a mess of different possible outputs that are canonicalized only at the very end. It is possible that I will work on this later.
Then there is the ODE module. I already need this solver in order to work with the integral curves that my code produces[1]. It is a very advanced solver written a few years ago by Aaron Meurer as part of his GSoC project. However, it still does not support systems of ODEs or solving for initial conditions. With Aaron’s help I have started those. The main difficulty is that I am covering only the simplest cases, however the new API must be futureproof. Moreover, here I again have a problem with the various outputs that can be produced by the ODE solver. Solutions are always returned as Equation instances (which is necessary, as some solutions can be in implicit form), however if there are multiple solutions they are returned in a list, while single solutions are returned themselves (not in a list). Anyway, the structure of the ODE module is straightforward so this should not be too hard to work around.
This week I will probably finish my work with the ODE solver and proceed to the form fields.
[1] The code on the differential geometry side is ready, however before showing it I will first extend the ODE solver in order to have more interesting examples.