Switch to DifferentialEquations.jl

[ChrisRackauckas]

ODE.jl is slowly being phased out, so for performance I would suggest switching to DifferentialEquations.jl. DifferentialEquations.jl has a low dependency mode so that way you don't need to depend on everything at once. Instead of ode45(f,u0,tspan;kwargs...), you can simply translate things to sol = solve(ODEProblem(f,u0,tspan),Tsit5();kwargs) and then use the sol type.

Doing it like this, you only need a dependency on using DiffEqBase, OrdinaryDiffEq. DiffEqBase was in ODE.jl anyways, and OrdinaryDiffEq.jl is the new solver you're using. This setup also gives you access to a whole load of other algorithms as well if you wanted the user to pass in the algorithm type. Note that here tspan must be a tuple and the types are respected, so a change to tspan = (0.0,1.0) instead of tspan = [0,1] is required.

If you're interested, I'll setup the PR.

[baggepinnen]

Hello!

Just to see that I get things right, the user would then send in the object Tsit5 in the example above?

A PR would be very generous =) I have just made sure that tests pass on travis to make it a bit easier.