fix all the warns. (#16)

* fix all the warns.

* Fix according to comments in PR.

* fix all the warns.

Author: pkofod

src/OptimTestProblems.jl

@@ -2,7 +2,8 @@ module OptimTestProblems
 
 export MultivariateProblems, UnivariateProblems
 
-include("optim_tests/multivariate/multivariate.jl")
+
+include("optim_tests/multivariate/MultivariateProblems.jl")
 include("optim_tests/univariate/bounded.jl")
 
 # Deprecation stuff

…optim_tests/multivariate/multivariate.jl …sts/multivariate/MultivariateProblems.jlsrc/optim_tests/multivariate/multivariate.jl renamed to src/optim_tests/multivariate/MultivariateProblems.jl src/optim_tests/multivariate/multivariate.jl renamed to src/optim_tests/multivariate/MultivariateProblems.jl

@@ -1,6 +1,6 @@
 module MultivariateProblems
 
-import Base.gradient
+import LinearAlgebra: dot
 
 export UnconstrainedProblems
 

src/optim_tests/multivariate/constrained.jl

@@ -1,6 +1,7 @@
 module ConstrainedProblems
 
-using ..OptimizationProblem,..ConstraintData
+import ..OptimizationProblem
+import ..ConstraintData
 
 examples = Dict{AbstractString, OptimizationProblem}()
 

src/optim_tests/multivariate/more_testing.jl

@@ -55,7 +55,7 @@ function extrosenbrock_hessian!(storage,x,param)
 end
 
 function _extrosenbrockproblem(N::Int;
-                               initial_x::AbstractArray{T} = repmat([-1.2,1],Int(N/2)),
+                               initial_x::AbstractArray{T} = repeat([-1.2,1]; inner = round(Int, N/2)),
                                name::AbstractString = "Extended Rosenbrock ($N)") where T
     @assert mod(N,2) == 0
     OptimizationProblem(name,
@@ -65,7 +65,7 @@ function _extrosenbrockproblem(N::Int;
                         extrosenbrock_hessian!,
                         nothing, # Constraints
                         initial_x,
-                        ones(initial_x),
+                        fill(T(1), length(initial_x)),
                         zero(T),
                         true,
                         false,
@@ -148,7 +148,7 @@ function extpowell_hessian!(storage,x,param)
 end
 
 function _extpowellproblem(N::Int;
-                           initial_x::AbstractArray{T} = repmat(float([3,-1,0,1]),Int(N/4)),
+                           initial_x::AbstractArray{T} = repeat(float([3,-1,0,1]); inner = round(Int, N/4)),
                            name::AbstractString = "Extended Powell ($N)") where T
     @assert mod(N,4) == 0
     OptimizationProblem(name,
@@ -158,7 +158,7 @@ function _extpowellproblem(N::Int;
                         extpowell_hessian!,
                         nothing, # Constraints
                         initial_x,
-                        zeros(initial_x),
+                        fill(zero(T), length(initial_x)),
                         zero(T),
                         true,
                         false,
@@ -303,7 +303,7 @@ function _trigonometricproblem(N::Int;
                         trigonometric_hessian!,
                         nothing, # Constraints
                         initial_x,
-                        zeros(initial_x),
+                        fill(T(0), length(initial_x)),
                         zero(T),
                         true,
                         false,
@@ -333,7 +333,7 @@ examples["Trigonometric"] = _trigonometricproblem(100)
 sumsq_obj(f, x) = sum(f(x).^2)
 
 function sumsq_gradient!(g::AbstractVector, f, J, x::AbstractVector)
-    copy!(g, sum((2.0 .* f(x)) .* J(x), 1))
+    copyto!(g, sum((2.0 .* f(x)) .* J(x), dims = 1))
 end
 
 function sumsq_hessian!(h::AbstractMatrix, f, J, H, x::AbstractVector)
@@ -343,7 +343,7 @@ function sumsq_hessian!(h::AbstractMatrix, f, J, H, x::AbstractVector)
     for i = 1:length(fx)
         htmp += (2.0 * fx[i]) * H(x, i)
     end
-    copy!(h, htmp)
+    copyto!(h, htmp)
 end
 
 const beale_y = [1.5, 2.25, 2.625]

src/optim_tests/multivariate/quad_transforms.jl

@@ -6,7 +6,7 @@
 function quad(x::Vector, param)
     mat = param.mat
     xt = x-param.vec
-    return 0.5*vecdot(xt, mat*xt)
+    return 0.5*dot(xt, mat*xt)
 end
 
 function quad_gradient!(storage::Vector, x::Vector, param)
@@ -19,7 +19,7 @@ function quad_fun_gradient!(storage::Vector, x::Vector, param)
     mat = param.mat
     xt = x-param.vec
     storage .= mat*xt
-    return 0.5*vecdot(xt, mat*xt)
+    return 0.5*dot(xt, mat*xt)
 end
 
 function quad_hessian!(storage::Matrix, x::Vector, param)
@@ -32,7 +32,7 @@ struct MatVecHolder{Tv <: AbstractVector,
     vec::Tv
 end
 
-function _quadraticproblem(N::Int; mat::AbstractArray{T,2} = spdiagm(float(1:N)),
+function _quadraticproblem(N::Int; mat::AbstractArray{T,2} = sparse(Diagonal(float(1:N))),
                            x0::AbstractVector{T} = ones(N),
                            initial_x::AbstractVector{T} = zeros(N),
                            name::AbstractString = "Quadratic Diagonal ($N)") where T <: Number
@@ -72,7 +72,7 @@ function paraboloid(x::AbstractArray, param::ParaboloidStruct)
     @. xt = x - param.vec
     xt[2:end] .-= param.alpha*xt[1]^2
 
-    return 0.5*vecdot(xt, mat*xt)
+    return 0.5*dot(xt, mat*xt)
 end
 
 function paraboloid_gradient!(storage::AbstractArray, x::AbstractArray, param::ParaboloidStruct)
@@ -97,14 +97,14 @@ function paraboloid_fun_gradient!(storage::AbstractArray, x::AbstractArray, para
     storage .= mat*xt
     storage[1] -= 2.0*param.alpha*xt[1]*sum(storage[2:end])
 
-    return 0.5*vecdot(xt, mat*xt)
+    return 0.5*dot(xt, mat*xt)
 end
 
 function paraboloid_hessian!(storage,x,param)
     error("Hessian not implemented for Paraboloid")
 end
 
-function _paraboloidproblem(N::Int; mat::AbstractArray{T,2} = spdiagm(float(1:N)),
+function _paraboloidproblem(N::Int; mat::AbstractArray{T,2} = sparse(Diagonal(float(1:N))),
                             x0::AbstractVector{T} = ones(N),
                             initial_x::AbstractVector{T} = zeros(N),
                             alpha::T = 10.0,
@@ -126,21 +126,17 @@ end
 examples["Paraboloid Diagonal"] = _paraboloidproblem(100)
 
 function _randommatrix(N::Int, scaling::Bool=true)
-    F = qrfact(randn(N,N))
+    F = qr(randn(N,N))
     if scaling
-        retval = F[:Q]'*spdiagm(float(1:N))*F[:Q]
+        retval = F.Q'*sparse(Diagonal(float(1:N)))*F.Q
     else
-        retval = F[:Q]'*F[:Q]
+        retval = F.Q'*F.Q
     end
     retval
 end
 
-# TODO: From Julia 0.7 onwards, we can use Base.Test.guardsrand() to restore the existing seed
-oldseed = copy(Base.GLOBAL_RNG) # Store current seed
-
-srand(0)
+guardsrand(0) do
 examples["Paraboloid Random Matrix"] = _paraboloidproblem(100;
                                                           name = "Paraboloid Random Matrix (100)",
                                                           mat = _randommatrix(100))
-
-copy!(Base.GLOBAL_RNG, oldseed) # Restore current seed
+end

src/optim_tests/multivariate/unconstrained.jl

@@ -1,9 +1,13 @@
 module UnconstrainedProblems
 
-using ..OptimizationProblem
+import ..OptimizationProblem
 
 export OptimizationProblem
 
+import LinearAlgebra: Diagonal, qr, dot
+import SparseArrays: sparse
+import Test: guardsrand
+
 import ..objective, ..gradient, ..hessian
 
 examples = Dict{AbstractString, OptimizationProblem}()

src/optim_tests/univariate/bounded.jl

@@ -37,8 +37,8 @@ module UnivariateProblems
     examples["Problem13"] = UnivariateProblem("Problem13",
                                                 p13,
                                                 [0.001, 0.99],
-                                                [1./sqrt(2.0),],
-                                                [p13(1./sqrt(2.0)),])
+                                                [1 ./ sqrt(2.0),],
+                                                [p13(1 ./ sqrt(2.0)),])
 
     # Problem 18 from [1]
     p18(x) = x <= 3.0 ? (x-2.0)^2 : 2.0*log(x-2.0)+1.0

test/runtests.jl

@@ -1,6 +1,8 @@
 using OptimTestProblems
 using OptimTestProblems.MultivariateProblems
 
+import LinearAlgebra: norm
+
 using Test
 
 verbose = false
@@ -35,14 +37,14 @@ end
         gs = similar(p.initial_x)
         g! = gradient(p)
         g!(gs, p.solutions)
-        soltest && @test vecnorm(gs, Inf) < tol
+        soltest && @test norm(gs, Inf) < tol
 
         fg! = objective_gradient(p)
         fgs = similar(gs)
         g!(gs, p.initial_x)
 
         @test fg!(fgs, p.initial_x) ≈ f(p.initial_x)
-        @test vecnorm(fgs.-gs, Inf)  < eps(eltype(gs))
+        @test norm(fgs.-gs, Inf)  < eps(eltype(gs))
     end
 end
 
@@ -62,8 +64,9 @@ end
             @test all(p.solutions .< p.constraintdata.ux)
         end
 
-        if !isempty(p.constraintdata.lc)
-            c = zeros(p.constraintdata.lc)
+        lc = p.constraintdata.lc
+        if !isempty(lc)
+            c = fill(eltype(lc)(0.0), size(lc))
             p.constraintdata.c!(c, p.solutions)
             @test all(c .>= p.constraintdata.lc)
             @test all(c .<= p.constraintdata.uc)
@@ -73,13 +76,13 @@ end
         gs = similar(p.initial_x)
         g! = gradient(p)
         g!(gs, p.solutions)
-        #soltest && @test vecnorm(gs, Inf) < tol
+        #soltest && @test norm(gs, Inf) < tol
 
         fg! = objective_gradient(p)
         fgs = similar(gs)
         g!(gs, p.initial_x)
 
         @test fg!(fgs, p.initial_x) ≈ f(p.initial_x)
-        @test vecnorm(fgs.-gs, Inf)  < eps(eltype(gs))
+        @test norm(fgs.-gs, Inf)  < eps(eltype(gs))
     end
 end