Callback function cannot change the state of the optimization

[oameye]

Describe the bug 🐞

Using certain solvers, one cannot change the state of the optimization in the callback function

Expected behavior

The state to be updateable in the callback function

Minimal Reproducible Example 👇

using Optimization, OptimizationOptimJL

rosenbrock(x, p) = (p[1] - x[1])^2 + p[2] * (x[2] - x[1]^2)^2
x0 = zeros(2)
p = [1.0, 100.0]

optf = OptimizationFunction(rosenbrock, AutoFiniteDiff())
prob = Optimization.OptimizationProblem(optf, x0, p)
function callback(state, loss)
    state.u .= 0
    false
end
sol = solve(prob, Optimization.LBFGS(), callback=callback) # changes state
sol = solve(prob, OptimizationOptimJL.LBFGS(), callback=callback) # does not change state

Environment (please complete the following information):

• Output of using Pkg; Pkg.status()

  [7f7a1694] Optimization v4.0.5
  [36348300] OptimizationOptimJL v0.4.1

• Output of versioninfo()

Julia Version 1.10.7
Commit 4976d05258 (2024-11-26 15:57 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: Windows (x86_64-w64-mingw32)
  CPU: 16 × 12th Gen Intel(R) Core(TM) i5-1240P
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-15.0.7 (ORCJIT, alderlake)
Threads: 10 default, 0 interactive, 5 GC (on 16 virtual cores)
Environment:
  JULIA_EDITOR = code
  JULIA_NUM_THREADS = 10

[Vaibhavdixit02]

This seems like a constraint from the downstream package, so we can probably not do anything about it here. For the two specific optimizers above I would recommend the Optimization.LBFGS anyway and it seems it works as you expect too.

[ChrisRackauckas]

We should probably find a way to communicate back to the user that this kind of modification is not doing anything then. Make it an immutable?