Julia 0.7

deps/build.jl

@@ -2,13 +2,16 @@ using BinDeps
 
 @BinDeps.setup
 
-blasvendor = Base.BLAS.vendor()
+using Libdl
+using LinearAlgebra.BLAS
+
+blasvendor = BLAS.vendor()
 
 direct = library_dependency("libscsdir", aliases=["libscsdir64"])
 indirect = library_dependency("libscsindir", aliases=["libscsindir64"])
 
 # TODO: Provide both libs in the "scs" Homebrew package.
-# if is_apple()
+# if Sys.isapple()
 #     using Homebrew
 #     provides(Homebrew.HB, "scs", [direct, indirect], os = :Darwin)
 # end
@@ -35,7 +38,7 @@ prefix = joinpath(BinDeps.depsdir(direct), "usr")
 srcdir = joinpath(BinDeps.depsdir(direct), "src", "scs-$version/")
 
 ldflags = ""
-if is_apple()
+if Sys.isapple()
     ldflags = "$ldflags -undefined suppress -flat_namespace"
 end
 cflags = "-DCOPYAMATRIX -DDLONG -DUSE_LAPACK -DCTRLC=1"

src/MOIWrapper.jl

@@ -60,7 +60,7 @@ end
 mutable struct SCSOptimizer <: MOI.AbstractOptimizer
     cone::ConeData
     maxsense::Bool
-    data::Union{Void, ModelData} # only non-Void between MOI.copy! and MOI.optimize!
+    data::Union{Nothing, ModelData} # only non-Void between MOI.copy! and MOI.optimize!
     sol::MOISolution
     function SCSOptimizer()
         new(ConeData(), false, nothing, MOISolution())
@@ -169,7 +169,7 @@ function _scalecoef(rows, coef, minus, ::Type{MOI.PositiveSemidefiniteConeTriang
     scaling = minus ? -1 : 1
     scaling2 = rev ? scaling / √2 : scaling * √2
     output = copy(coef)
-    diagidx = IntSet()
+    diagidx = BitSet()
     for i in 1:d
         push!(diagidx, trimap(i, i))
     end
@@ -232,11 +232,11 @@ function MOIU.loadconstraint!(optimizer::SCSOptimizer, ci, f::MOI.VectorAffineFu
     offset = constroffset(optimizer, ci)
     rows = constrrows(s)
     optimizer.cone.nrows[offset] = length(rows)
-    i = offset + rows
+    i = offset .+ rows
     # The SCS format is b - Ax ∈ cone
     # so minus=false for b and minus=true for A
     optimizer.data.b[i] = scalecoef(rows, orderval(f.constant, s), false, s)
-    append!(optimizer.data.I, offset + orderidx(I, s))
+    append!(optimizer.data.I, offset .+ orderidx(I, s))
     append!(optimizer.data.J, J)
     append!(optimizer.data.V, scalecoef(I, V, true, s))
 end
@@ -268,7 +268,7 @@ MOIU.canload(::SCSOptimizer, ::MOI.ObjectiveSense) = true
 function MOIU.load!(::SCSOptimizer, ::MOI.ObjectiveSense, ::MOI.OptimizationSense) end
 MOIU.canload(::SCSOptimizer, ::MOI.ObjectiveFunction{MOI.ScalarAffineFunction{Float64}}) = true
 function MOIU.load!(optimizer::SCSOptimizer, ::MOI.ObjectiveFunction, f::MOI.ScalarAffineFunction)
-    c0 = full(sparsevec(_varmap(f), f.coefficients, optimizer.data.n))
+    c0 = Vector(sparsevec(_varmap(f), f.coefficients, optimizer.data.n))
     optimizer.data.objconstant = f.constant
     optimizer.data.c = optimizer.maxsense ? -c0 : c0
 end
@@ -352,7 +352,7 @@ end
 function MOI.get(optimizer::SCSOptimizer, ::MOI.ConstraintPrimal, ci::CI{<:MOI.AbstractFunction, S}) where S <: MOI.AbstractSet
     offset = constroffset(optimizer, ci)
     rows = constrrows(optimizer, ci)
-    _unshift(optimizer, offset, unscalecoef(rows, reorderval(optimizer.sol.slack[offset + rows], S), S, length(rows)), S)
+    _unshift(optimizer, offset, unscalecoef(rows, reorderval(optimizer.sol.slack[offset .+ rows], S), S, length(rows)), S)
 end
 
 MOI.canget(optimizer::SCSOptimizer, ::MOI.DualStatus) = true
@@ -372,7 +372,7 @@ end
 function MOI.get(optimizer::SCSOptimizer, ::MOI.ConstraintDual, ci::CI{<:MOI.AbstractFunction, S}) where S <: MOI.AbstractSet
     offset = constroffset(optimizer, ci)
     rows = constrrows(optimizer, ci)
-    unscalecoef(rows, reorderval(optimizer.sol.dual[offset + rows], S), S, length(rows))
+    unscalecoef(rows, reorderval(optimizer.sol.dual[offset .+ rows], S), S, length(rows))
 end
 
 MOI.canget(optimizer::SCSOptimizer, ::MOI.ResultCount) = true

src/MPBWrapper.jl

@@ -6,7 +6,14 @@
 # MathProgBase.jl interface for the SCS.jl solver wrapper
 #############################################################################
 
-importall MathProgBase.SolverInterface
+using LinearAlgebra: dot
+using MathProgBase.SolverInterface
+
+import MathProgBase.SolverInterface: ConicModel, LinearQuadraticModel,
+    getdual, getobjval, getsolution, getsolvetime, getvardual, loadproblem!,
+    numconstr, numvar, optimize!, setbvec!, setwarmstart!, status,
+    supportedcones
+
 import Base.convert
 
 # TODO: don't add to Base.convert!
@@ -145,14 +152,14 @@ function orderconesforscs(A_in, b_in, c_cones, v_cones)
     A_t = spzeros(n,0)
     b = zeros(0)
     row_map_ind = zeros(Int, length(b_in))
-    row_map_type = Array{Symbol}(length(b_in))
+    row_map_type = Array{Symbol}(undef, length(b_in))
     col_map_ind = zeros(Int, n)
-    col_map_type = Array{Symbol}(n)
+    col_map_type = Array{Symbol}(undef, n)
 
     # First, count the total number of variables
     num_vars = 0
     for (cone, idxs) in v_cones
-        col_map_type[idxs] = cone
+        col_map_type[idxs] .= cone
         num_vars += length(idxs)
     end
     @assert num_vars == n
@@ -432,7 +439,7 @@ end
 # warmstart
 # kwargs can be `primal_sol`, `dual_sol`, and `slack`
 function setwarmstart!(m::SCSMathProgModel, primal_sol; kwargs...)
-    push!(m.options, (:warm_start, true))
+    m.options = pairs(merge(m.options.data, (warm_start=true,)))
     m.primal_sol = primal_sol
     for (k,v) in kwargs
         setfield!(m, k, v)

src/SCS.jl

@@ -8,12 +8,12 @@ else
     error("SCS not properly installed. Please run Pkg.build(\"SCS\") and restart julia")
 end
 
-import Base.Libdl: RTLD_LAZY, RTLD_DEEPBIND, RTLD_GLOBAL, dlopen
+using Libdl
 
 function __init__()
     vnum = VersionNumber(SCS_version())
     # default binaries need access to Julia's lapack symbols
-    if is_unix()
+    if Sys.isunix()
         dlopen(Base.liblapack_name, RTLD_LAZY|RTLD_DEEPBIND|RTLD_GLOBAL)
     end
     depsdir = realpath(joinpath(dirname(@__FILE__),"..","deps"))

src/c_wrapper.jl

@@ -4,7 +4,7 @@ macro compat_gc_preserve(args...)
     vars = args[1:end-1]
     body = args[end]
     if VERSION > v"0.7.0-"
-        return esc(Expr(:macrocall, Expr(:., :Base, Base.Meta.quot(Symbol("@gc_preserve"))), __source__, args...))
+        return esc(Expr(:macrocall, Expr(:., :GC, Base.Meta.quot(Symbol("@preserve"))), __source__, args...))
     else
         return esc(body)
     end
@@ -52,7 +52,7 @@ function SCS_solve(T::Union{Type{Direct}, Type{Indirect}},
     cone = Ref(SCSCone(f, l, q, s, ep, ed, p))
     info = Ref(SCSInfo())
 
-    if (:warm_start, true) in options && length(primal_sol) == n && length(dual_sol) == m && length(slack) == m
+    if (:warm_start=>true) in options && length(primal_sol) == n && length(dual_sol) == m && length(slack) == m
         x = primal_sol
         y = dual_sol
         s = slack
@@ -82,26 +82,26 @@ for (T, lib) in zip([SCS.Direct, SCS.Indirect], [SCS.direct, SCS.indirect])
     @eval begin
         function SCS_init(::Type{$T}, data::Ref{SCSData}, cone::Ref{SCSCone}, info::Ref{SCSInfo})
 
-            p_work = ccall((:scs_init, $lib), Ptr{Void},
+            p_work = ccall((:scs_init, $lib), Ptr{Nothing},
                 (Ptr{SCSData}, Ptr{SCSCone}, Ptr{SCSInfo}),
                 data, cone, info)
 
             return p_work
         end
 
         # solution struct is simple enough, we know it won't be modified by SCS so take by value
-        function SCS_solve(::Type{$T}, p_work::Ptr{Void}, data::Ref{SCSData}, cone::Ref{SCSCone}, solution::SCSSolution, info::Ref{SCSInfo})
+        function SCS_solve(::Type{$T}, p_work::Ptr{Nothing}, data::Ref{SCSData}, cone::Ref{SCSCone}, solution::SCSSolution, info::Ref{SCSInfo})
 
             status = ccall((:scs_solve, $lib), Int,
-                (Ptr{Void}, Ptr{SCSData}, Ptr{SCSCone}, Ref{SCSSolution}, Ptr{SCSInfo}),
+                (Ptr{Nothing}, Ptr{SCSData}, Ptr{SCSCone}, Ref{SCSSolution}, Ptr{SCSInfo}),
                 p_work, data, cone, solution, info)
 
             return status
         end
 
-        function SCS_finish(::Type{$T}, p_work::Ptr{Void})
-            ccall((:scs_finish, $lib), Void,
-                (Ptr{Void}, ),
+        function SCS_finish(::Type{$T}, p_work::Ptr{Nothing})
+            ccall((:scs_finish, $lib), Nothing,
+                (Ptr{Nothing}, ),
                 p_work)
         end
     end

src/types.jl

@@ -1,5 +1,6 @@
-export SCSMatrix, SCSData, SCSSettings, SCSSolution, SCSInfo, SCSCone, SCSVecOrMatOrSparse
+using SparseArrays
 
+export SCSMatrix, SCSData, SCSSettings, SCSSolution, SCSInfo, SCSCone, SCSVecOrMatOrSparse
 
 SCSVecOrMatOrSparse = Union{VecOrMat, SparseMatrixCSC{Float64,Int}}
 
@@ -68,9 +69,9 @@ struct SCSData
 end
 
 struct SCSSolution
-    x::Ptr{Void}
-    y::Ptr{Void}
-    s::Ptr{Void}
+    x::Ptr{Nothing}
+    y::Ptr{Nothing}
+    s::Ptr{Nothing}
 end
 
 

test/MPBWrapper.jl

@@ -2,9 +2,9 @@
     include("mpb_linear.jl")
 end
 
-import SCS
+using Pkg.dir
 @testset "MathProgBase" begin
-    include(joinpath(Pkg.dir("MathProgBase"),"test","conicinterface.jl"))
+    include(joinpath(dir("MathProgBase"),"test","conicinterface.jl"))
     coniclineartest(SCS.SCSSolver(), duals=true, tol=1e-2)
     conicSOCtest(SCS.SCSSolver(), duals=true, tol=1e-2)
     conicEXPtest(SCS.SCSSolver(), duals=true, tol=1e-2)

test/direct.jl

@@ -1,6 +1,3 @@
-using SCS
-using Base.Test
-
 # Solve a trivial problem
 A = reshape([1.0],(1,1))
 solution = SCS_solve(SCS.Direct, 1, 1, A, [1.0], [1.0], 1, 0, Int[], Int[], 0, 0, Float64[]);

test/indirect.jl

@@ -1,6 +1,3 @@
-using SCS
-using Base.Test
-
 # Solve a trivial problem
 A = reshape([1.0],(1,1))
 solution = SCS_solve(SCS.Indirect, 1, 1, A, [1.0], [1.0], 1, 0, Int[], Int[], 0, 0, Float64[]);

test/mpb_linear.jl

@@ -7,9 +7,9 @@
 # Test the MathProgBase.jl interface for the SCS.jl solver wrapper
 #############################################################################
 
-using Base.Test
 using MathProgBase
-using SCS
+using LinearAlgebra
+using SparseArrays
 
 solver = SCSSolver()
 objtol = 1e-4

test/options.jl

@@ -7,10 +7,7 @@
 # Tests the ability to pass options
 #############################################################################
 
-using Base.Test
-using MathProgBase.SolverInterface
-using SCS
-
+using MathProgBase
 
 # The normal test
 A = [1.0 1.0 0.0 0.0 0.0;
@@ -33,8 +30,7 @@ MathProgBase.optimize!(m)
 @test isapprox(MathProgBase.getobjval(m), -99.0, atol=1e-5)
 
 # With a warmstart from the eps = 1e-8 solution, solution should be extremely accurate even after 1 iteration
-push!(m.options, (:warm_start, true))
-push!(m.options, (:max_iters, 1))
+m.options = pairs(merge(m.options.data, (warm_start=true, max_iters=1)))
 MathProgBase.optimize!(m)
 @test isapprox(MathProgBase.getobjval(m), -99.0, atol=1e-5)
 

test/runtests.jl

@@ -1,4 +1,5 @@
-using Base.Test
+using Test
+using SCS
 
 tests = ["direct.jl",
          "indirect.jl",