Does StochasticDiffEq need the full OrdinaryDiffEq dependency?

[oameye]

Now that OrdinaryDiffEq has been split up, does StochasticDiffEq need the full OrdinaryDiffEq library?

[ytdHuang]

any update for this issue ?

Cause StochasticDiffEq depends on the entire OrdinaryDiffEq instead of OrdinaryDiffEqCore, and installs a lot more libraries and packages.

[ChrisRackauckas]

No movement on it yet, and I don't htink I can get to it this month. But maybe the first change is that, you probably could make it just depend on OrdinaryDiffEqCore, OrdinaryDiffEqNonlinearSolve and that might be a minimal amount. We should split StochasticDiffEq into sublibs like OrdinaryDiffEq in order to fully minimize it, but that would be a bit more work.

[ytdHuang]

@ChrisRackauckas

Thank you for the information.

I'm already using SciMLBase as the minimal dependency of my package (QuantumToolbox.jl).

For the StochasticDiffEq, I just want to set a default algorithm for our function to use stochastic differential equation, something like:

import SciMLBase: solve, SDEProblem
import StochasticDiffEq: StochasticDiffEqAlgorithm, SRA1

function sde_function(...; alg::StochasticDiffEqAlgorithm = SRA1(), ...)
    ...
    prob = SDEProblem(...)
    sol = solve(prob, alg, ...)
    ...
end

I noticed that, for the StochasticDiffEqAlgorithm, I can use SciMLBase.AbstractSDEAlgorithm instead.

Would it be possible to not import StochasticDiffEq: SRA1, but also solve with SDE algorithms?

[ChrisRackauckas]

We could just move the default handling for the SDE solver choice from DifferentialEquations.jl to here.