Remove OptimizationBase.jl pieces

[ChrisRackauckas]

No description provided.

[drbergman]

This PR seems to be breaking backwards compatibility for me. I can confirm that my package compiled with v2.120.0 and then errored on v2.121.0. Error message below. It looks like this function was removed and is the cause of the error:

function init(prob::OptimizationProblem, alg, args...; kwargs...)::AbstractOptimizationCache
    if prob.u0 !== nothing && !isconcretetype(eltype(prob.u0))
        throw(NonConcreteEltypeError(eltype(prob.u0)))
    end
    _check_opt_alg(prob::OptimizationProblem, alg; kwargs...)
    cache = __init(prob, alg, args...; prob.kwargs..., kwargs...)
    return cache
end

My package uses (no SciMLBase here except in their deps)

[compat]
...
Optimization = "4.1.0"
OptimizationOptimJL = "0.4.1"
...

Any recommendations for how to get around this?

In case it helps, here's the call I make to these Optimization packages:

optf = OptimizationFunction(objfn, Optimization.AutoForwardDiff())
prob = OptimizationProblem(optf, [atan(y, x)], [x; y; a; b], lb=[-pi], ub=[pi])
sol = solve(prob, BFGS())

Finally, here's the error output:

ERROR: LoadError: MethodError: no method matching init(::SciMLBase.OptimizationProblem{true, SciMLBase.OptimizationFunction{true, ADTypes.AutoForwardDiff{nothing, Nothing}, OptimizationOptimJL.var"#obj_f#38", Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, typeof(SciMLBase.DEFAULT_OBSERVED_NO_TIME), Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing}, Vector{Float64}, Tuple{Matrix{Float64}, Vector{Float64}}, Vector{Float64}, Vector{Float64}, Nothing, Nothing, Nothing, Nothing, Base.Pairs{Symbol, Union{}, Tuple{}, @NamedTuple{}}}, ::Optim.LBFGS{Nothing, LineSearches.InitialStatic{Float64}, LineSearches.HagerZhang{Float64, Base.RefValue{Bool}}, Returns{Nothing}}; maxiters::Int64, maxtime::Int64)
Stacktrace:
 [1] solve(::SciMLBase.OptimizationProblem{true, SciMLBase.OptimizationFunction{true, ADTypes.AutoForwardDiff{nothing, Nothing}, OptimizationOptimJL.var"#obj_f#38", Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, typeof(SciMLBase.DEFAULT_OBSERVED_NO_TIME), Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing}, Vector{Float64}, Tuple{Matrix{Float64}, Vector{Float64}}, Vector{Float64}, Vector{Float64}, Nothing, Nothing, Nothing, Nothing, @Kwargs{}}, ::Vararg{Any}; kwargs::@Kwargs{maxiters::Int64, maxtime::Int64})
   @ CommonSolve ~/.julia/packages/CommonSolve/JfpfI/src/CommonSolve.jl:23
 [2] (::OptimizationOptimJL.var"#solve_nonnegative_least_squares#39"{OptimizationOptimJL.var"#obj_f#38"})(A::Matrix{Float64}, b::Vector{Float64}, solver::Optim.LBFGS{Nothing, LineSearches.InitialStatic{Float64}, LineSearches.HagerZhang{Float64, Base.RefValue{Bool}}, Returns{Nothing}})
   @ OptimizationOptimJL ~/.julia/packages/OptimizationOptimJL/pe3SX/src/OptimizationOptimJL.jl:491
 [3] macro expansion
   @ ~/.julia/packages/OptimizationOptimJL/pe3SX/src/OptimizationOptimJL.jl:502 [inlined]
 [4] macro expansion
   @ ~/.julia/packages/PrecompileTools/L8A3n/src/workloads.jl:78 [inlined]
 [5] top-level scope
   @ ~/.julia/packages/OptimizationOptimJL/pe3SX/src/OptimizationOptimJL.jl:480
 [6] include
   @ ./Base.jl:495 [inlined]
 [7] include_package_for_output(pkg::Base.PkgId, input::String, depot_path::Vector{String}, dl_load_path::Vector{String}, load_path::Vector{String}, concrete_deps::Vector{Pair{Base.PkgId, UInt128}}, source::String)
   @ Base ./loading.jl:2292
 [8] top-level scope
   @ stdin:4
in expression starting at /home/runner/.julia/packages/OptimizationOptimJL/pe3SX/src/OptimizationOptimJL.jl:1
in expression starting at stdin:4

[drbergman]

OK, it seems like OptimizationOptimJL is now a submodule of Optimization and just removing that explicit dep works.

[drbergman]

Well, it now compiles but it doesn't run. The sample at the Optimization.jl docs errors:

# Import the package and define the problem to optimize
using Optimization, Zygote
rosenbrock(u, p) = (p[1] - u[1])^2 + p[2] * (u[2] - u[1]^2)^2
u0 = zeros(2)
p = [1.0, 100.0]

optf = OptimizationFunction(rosenbrock, AutoZygote())
prob = OptimizationProblem(optf, u0, p)

sol = solve(prob, Optimization.LBFGS())

ERROR: MethodError: no method matching init(::OptimizationProblem{…}, ::Optimization.LBFGS)
The function `init` exists, but no method is defined for this combination of argument types.
Stacktrace:
 [1] solve(::OptimizationProblem{true, OptimizationFunction{…}, Vector{…}, Vector{…}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, @Kwargs{}}, ::Vararg{Any}; kwargs::@Kwargs{})
   @ CommonSolve ~/.julia/packages/CommonSolve/JfpfI/src/CommonSolve.jl:23

[drbergman]

OK, final comment on this: I have pkg> add [email protected] to my package and then set the [compat] entry to SciMLBase = "~2.120.0" to not allow it to upgrade to 2.121.0.

[ChrisRackauckas]

Okay, finally fixed this code movement. Sorry, this change has been a mess.

[drbergman]

All good now! To me, this is the exception that proves the rule: the julia ecosystem does a great job with version control and making updates smooth. And I include you in that. Thank you for all you do with SciML!