Format project

docs/make.jl

@@ -3,7 +3,7 @@ using Documenter, DocumenterMarkdown, NLSolvers
 makedocs(
     doctest = false,
     sitename = "NLSolvers.jl",
-    pages = ["index.md", "optimization.md",],
+    pages = ["index.md", "optimization.md"],
     format = Markdown(),
 )
 

src/nlsolve/problem_types.jl

@@ -36,7 +36,7 @@ end
 function jacobian(nleq::NEqProblem{<:ScalarObjective}, J, x)
     gradient(nleq.R, J, x)
 end
-function value_jacobian(nleq::NEqProblem{<:ScalarObjective, <:Any, <:Any, OutOfPlace}, F, J, x)
+function value_jacobian(nleq::NEqProblem{<:ScalarObjective,<:Any,<:Any,OutOfPlace}, F, J, x)
     nleq.R.fg(J, x)
 end
 

src/nlsolve/spectral/dfsane.jl

@@ -59,15 +59,22 @@ To re-implement the setup in the numerical section of [PAPER] use
 
 DFSANE(nexp=2,memory=10, sigma_limit=(1e-10, 1e10), decrease=1e-4, sigma_0=1.0))
 """
-struct DFSANE{T,Σ,TAU,N, G, A}
+struct DFSANE{T,Σ,TAU,N,G,A}
     memory::T
     sigma_limits::Σ
     taus::TAU
     nexp::N
     γ::G
     σ0::A
 end
-DFSANE(; memory = 4, sigma_limits = (1e-5, 1e5), taus=(1 / 10, 1 / 2), nexp=2,decrease = 1 / 10000, sigma_0 = 1.0) = DFSANE(memory, sigma_limits, taus,nexp, decrease, sigma_0)
+DFSANE(;
+    memory = 4,
+    sigma_limits = (1e-5, 1e5),
+    taus = (1 / 10, 1 / 2),
+    nexp = 2,
+    decrease = 1 / 10000,
+    sigma_0 = 1.0,
+) = DFSANE(memory, sigma_limits, taus, nexp, decrease, sigma_0)
 
 init(::NEqProblem, ::DFSANE; x, Fx = copy(x), y = copy(x), d = copy(x), z = copy(x)) =
     (; x, Fx, y, d, z)
@@ -122,7 +129,7 @@ function solve(
         ηk = ρ2F0 / (1 + iter)^2
         φ(α) = norm(F(Fx, (z .= x .+ α .* d)))^nexp
         φ0 = fx
-        α, φα = find_steplength(RNMS(method.γ,method.σ0), φ, φ0, fbar, ηk, τmin, τmax)
+        α, φα = find_steplength(RNMS(method.γ, method.σ0), φ, φ0, fbar, ηk, τmin, τmax)
         if isnan(α) || isnan(φα)
             status = :linesearch_failed
             break

src/nlsolve/trustregions/newton.jl

@@ -48,7 +48,7 @@ function solve(
     approach::TrustRegion{<:Union{SR1,DBFGS,BFGS,Newton},<:Any,<:Any},
     options::NEqOptions,
 )
-  if !(mstyle(prob) === InPlace()) && !(approach.spsolve isa Dogleg)
+    if !(mstyle(prob) === InPlace()) && !(approach.spsolve isa Dogleg)
         throw(
             ErrorException(
                 "solve() not defined for OutOfPlace() with Trustregion for NEqProblem",
@@ -76,5 +76,5 @@ function solve(
         time = res.info.time,
         iter = res.info.iter,
     )
-    return  ConvergenceInfo(approach, newinfo, options)
+    return ConvergenceInfo(approach, newinfo, options)
 end

src/objectives.jl

@@ -20,15 +20,15 @@ struct ScalarObjective{Tf,Tg,Tfg,Tfgh,Th,Thv,Tbf,P}
     param::P
 end
 ScalarObjective(;
-     f = nothing,
-     g = nothing,
-     fg = nothing,
-     fgh = nothing,
-     h = nothing,
-     hv = nothing,
-     batched_f = nothing,
-     param = nothing,
- ) = ScalarObjective(f, g, fg, fgh, h, hv, batched_f, param)
+    f = nothing,
+    g = nothing,
+    fg = nothing,
+    fgh = nothing,
+    h = nothing,
+    hv = nothing,
+    batched_f = nothing,
+    param = nothing,
+) = ScalarObjective(f, g, fg, fgh, h, hv, batched_f, param)
 has_param(so::ScalarObjective) = so.param === nothing ? false : true
 function value(so::ScalarObjective, x)
     if has_param(so)
@@ -118,7 +118,8 @@ struct VectorObjective{TF,TJ,TFJ,TJv}
     FJ::TFJ
     Jv::TJv
 end
-VectorObjective(; F=nothing, J=nothing, FJ=nothing, Jv=nothing) = VectorObjective(F, J, FJ, Jv)
+VectorObjective(; F = nothing, J = nothing, FJ = nothing, Jv = nothing) =
+    VectorObjective(F, J, FJ, Jv)
 
 ## If prob is a NEqProblem, then we can just dispatch to least squares MeritObjective
 # if fast JacVec exists then maybe even line searches that updates the gradient can be used??? 

test/nlsolve/interface.jl

@@ -447,17 +447,14 @@ end
 @testset "scalar nlsolves" begin
     function ff(x)
         x^2
-     end
-
-     function fgg(Jx, x)
-         x^2, 2x
-     end
-
-     prob_obj = NLSolvers.ScalarObjective(
-         f=ff,
-         fg=fgg,
-     )
-
+    end
+
+    function fgg(Jx, x)
+        x^2, 2x
+    end
+
+    prob_obj = NLSolvers.ScalarObjective(f = ff, fg = fgg)
+
     prob = NEqProblem(prob_obj; inplace = false)
 
     x0 = 0.3
@@ -468,7 +465,7 @@ function solve_static()
     function F_rosenbrock_static(Fx, x)
         Fx1 = 1 - x[1]
         Fx2 = 10(x[2] - x[1]^2)
-        return @SVector([Fx1,Fx2])
+        return @SVector([Fx1, Fx2])
     end
     function J_rosenbrock_static(Jx, x)
         Jx11 = -1
@@ -489,11 +486,11 @@ function solve_static()
         nothing,
     )
 
-    prob_static =  NEqProblem(obj; inplace=false)
+    prob_static = NEqProblem(obj; inplace = false)
     x0_static = @SVector([-1.2, 1.0])
     res = solve(prob_static, x0_static, TrustRegion(Newton(), Dogleg()), NEqOptions())
 end
 
 solve_static()
 alloced = @allocated solve_static()
-@test alloced == 0
+@test alloced == 0

test/nlsolve/problems.jl

@@ -45,7 +45,7 @@ NLE_PROBS["rosenbrock"]["static"] = Dict()
 function F_rosenbrock_static(Fx, x)
     Fx1 = 1 - x[1]
     Fx2 = 10(x[2] - x[1]^2)
-    return @SVector([Fx1,Fx2])
+    return @SVector([Fx1, Fx2])
 end
 function J_rosenbrock_static(Jx, x)
     Jx11 = -1

test/optimize/param.jl

@@ -2,26 +2,26 @@
     N_test = 9
     w = rand(N_test)
     function obji(x, data)
-        s=sum((x .- data) .^ 2)
+        s = sum((x .- data) .^ 2)
         s
     end
     function g!(G, x, data)
         G .= 2 * (x .- data)
         G
     end
     function h!(H, x, data)
-        H .=  2 * I
+        H .= 2 * I
         H
     end
     param1 = [1, 2, 3]
     sc = ScalarObjective(; f = obji, g = g!, h = h!, param = param1)
     op = OptimizationProblem(sc)
     sol1 = solve(op, [3.0, 4.0, 4.0], BFGS(), OptimizationOptions())
-    @test sol1.info.solution ≈  param1
+    @test sol1.info.solution ≈ param1
 
     param2 = [10, 20, 30]
     sc = ScalarObjective(; f = obji, g = g!, h = h!, param = param2)
     op = OptimizationProblem(sc)
     sol2 = solve(op, [3.0, 4.0, 4.0], BFGS(), OptimizationOptions())
-    @test sol2.info.solution ≈  param2
-end
+    @test sol2.info.solution ≈ param2
+end