# Combining \`Optim\` and \`FieldVector\` from \`StaticArrays\`
**URL:** https://discourse.julialang.org/t/combining-optim-and-fieldvector-from-staticarrays/131472
**Category:** Optimization (Mathematical)
**Tags:** optim, staticarrays
**Created:** [August 8, 2025, 10:10am UTC](https://discourse.julialang.org/t/combining-optim-and-fieldvector-from-staticarrays/131472 "2025-08-08T10:10:16Z")
**Posts on this page:** 4
**Page:** 1
### Author:  [@NoFishLikeIan](https://discourse.julialang.org/u/NoFishLikeIan)
#### Post date: [August 8, 2025, 10:10am UTC](https://discourse.julialang.org/t/combining-optim-and-fieldvector-from-staticarrays/131472/1 "2025-08-08T10:10:16Z")
I’m getting a `MethodError` in `LBFGS` when trying to use a custom mutable struct that inherits from `FieldVector` for an optimization problem.
```julia
using StaticArrays
using Optimization, OptimizationOptimJL
mutable struct Point{T} <: FieldVector{2, T}
x::T
y::T
end
function rosenbrock(v, p)
(p[1] - v[1])^2 + p[2] * (v[2] - v[1]^2)^2
end
v₀ = Point(0., 0.)
p = (1., 100.)
f = OptimizationFunction(rosenbrock, Optimization.AutoForwardDiff())
prob = OptimizationProblem(f, v₀, p)
sol = solve(prob, LBFGS()) # ERROR: MethodError: no method matching Optim.LBFGSState(::Point{…}, ...)
```
My first attept was to write a new `rosenbrock` method as
```julia
function rosenbrock(v::Point, p)
rosenbrock(SVector(v.x, v.y), p)
end
```
but of course this still fails. Is there a good way to get this to work?
---
### Author:  [@gdalle](https://discourse.julialang.org/u/gdalle)
#### Post date: [August 8, 2025, 1:36pm UTC](https://discourse.julialang.org/t/combining-optim-and-fieldvector-from-staticarrays/131472/2 "2025-08-08T13:36:18Z")
Hi! Can you share the complete stack trace?
---
### Author:  [@matthias314](https://discourse.julialang.org/u/matthias314)
#### Post date: [August 8, 2025, 8:41pm UTC](https://discourse.julialang.org/t/combining-optim-and-fieldvector-from-staticarrays/131472/3 "2025-08-08T20:41:29Z")
I get
```julia-auto
julia> sol = solve(prob, LBFGS()) # ERROR: MethodError: no method matching Optim.LBFGSState(::Point{…}, ...)
ERROR: MethodError: no method matching Optim.LBFGSState(::Point{…}, ::Point{…}, ::Point{…}, ::Vector{…}, ::Vector{…}, ::Vector{…}, ::Point{…}, ::Point{…}, ::Point{…}, ::Float64, ::MVector{…}, ::Vector{…}, ::Int64, ::Point{…}, ::MVector{…}, ::Float64)
The type `Optim.LBFGSState` exists, but no method is defined for this combination of argument types when trying to construct it.
Closest candidates are:
Optim.LBFGSState(::Tx, ::Tx, ::G, ::Vector{T}, ::Tdx, ::Tdg, ::Tx, ::Tx, ::Tx, ::T, ::Any, ::Any, ::Int64, ::Tx, ::Tx, ::T) where {Tx, Tdx, Tdg, T, G}
@ Optim ~/.julia/packages/Optim/7krni/src/multivariate/solvers/first_order/l_bfgs.jl:139
```
where the penultimate argument is of type `MVector` instead of `Point`. I guess this is caused by the default method of `similar` for `FieldVector`. With
```julia-auto
Base.similar(::Type{<:Point}, ::Type{T}) where T = Point(zero(T), zero(T))
```
it works:
```julia-auto
julia> sol = solve(prob, LBFGS())
retcode: Success
u: 2-element Point{Float64} with indices SOneTo(2):
0.999999999999928
0.9999999999998559
```
One could also add a constructor for `Point` that doesn’t initialize the fields.
---
### Author:  [@NoFishLikeIan](https://discourse.julialang.org/u/NoFishLikeIan)
#### Post date: [August 9, 2025, 10:15am UTC](https://discourse.julialang.org/t/combining-optim-and-fieldvector-from-staticarrays/131472/4 "2025-08-09T10:15:56Z")
You are absolutely right, this works! It is suggested in the `FieldVector` [documentation](https://juliaarrays.github.io/StaticArrays.jl/v0.7/pages/api.html#FieldVector-1) to write a `Base.similar` method. I missed it.