R, Python, and MATLAB Wrappers

[ChrisRackauckas]

I think we are at a good spot to start considering how we can make DifferentialEquations.jl usable from R, Python, and MATLAB. For R, JuliaCall is quite well developed, with an example package for IPOPT and a JuliaCall example from Douglas Bates. This might be a good start.

Python has pyjulia which I am unsure about. MATLAB has mexjulia.

I think a good goal would be to expose basic ODEs, SDEs, DDEs, and DAEs for those languages. That would give a solid unique offering at each place to establish some ground and then use the knowledge of these tools to see what's feasible.

[ChrisRackauckas]

diffeqr is a set of R wrappers. JuliaCall is amazing and that was easy to create.

[ChrisRackauckas]

I'm running into some issues with pyjulia. Some starter code is here: JuliaPy/PyCall.jl#460 . The issues are:

How do deal with the solution object returns? We can just ignore the problem completely like diffeqr and just return t,u, or try to return an appropriate object ( JuliaPy/PyCall.jl#460 ).

How do we make function calls fast? In diffeqr we just had them write Julia code. I'm not sure how to do the same with pyjulia (JuliaPy/pyjulia#154), and I'm not sure how to interface with something else like Cython.

In a sense something is working:

from julia import Julia

jul = Julia()
jul.using("DifferentialEquations")

def f(u,p,t):
    return -u

u0 = 0.5
tspan = (0., 1.)
prob = jul.ODEProblem(f, u0, tspan)
jul.using("PyCall")
jul.using("DiffEqBase")
solve = jul.pyfunctionret(jul.solve, jul.Any, jul.PyAny)
solve(prob)

[ChrisRackauckas]

mexjulia is crashing: juliamatlab/mexjulia#58

[ChrisRackauckas]

diffeqpy works. It has almost a full interface. Needs some finishing touches. It uses DiffEqPy.jl under the hood to get the job done.

[ChrisRackauckas]

Looks like there's reports of mexjulia being almost unusable right now: juliamatlab/mexjulia#32 . We'll just not do MATLAB until a better solution comes along.