Merge pull request #92 from JuliaOpt/bl/moiv0.9

Updates to MOI v0.9

Author: blegat

Diff for: .gitignore

@@ -0,0 +1 @@
+Manifest.toml

Diff for: .travis.yml

@@ -3,11 +3,9 @@ os:
   - linux
   - osx
 julia:
-  - 0.6
-  - 0.7
   - 1.0
+  - 1.1
 notifications:
   email: false
 after_success:
-  - julia -e 'cd(Pkg.dir("ECOS")); Pkg.add("Coverage"); using Coverage; Coveralls.submit(Coveralls.process_folder())'
-  - julia -e 'cd(Pkg.dir("ECOS")); Pkg.add("Coverage"); using Coverage; Codecov.submit(Codecov.process_folder())'
+  - julia -e 'import Pkg; Pkg.add("Coverage"); using Coverage; Coveralls.submit(process_folder()); Codecov.submit(process_folder())'

Diff for: Project.toml

@@ -0,0 +1,25 @@
+name = "ECOS"
+uuid = "e2685f51-7e38-5353-a97d-a921fd2c8199"
+repo = "https://github.com/JuliaOpt/ECOS.jl.git"
+version = "0.10.0"
+
+[deps]
+BinaryProvider = "b99e7846-7c00-51b0-8f62-c81ae34c0232"
+Compat = "34da2185-b29b-5c13-b0c7-acf172513d20"
+Libdl = "8f399da3-3557-5675-b5ff-fb832c97cbdb"
+LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
+MathOptInterface = "b8f27783-ece8-5eb3-8dc8-9495eed66fee"
+MathProgBase = "fdba3010-5040-5b88-9595-932c9decdf73"
+SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
+
+[compat]
+BinaryProvider = "≥ 0.3.0"
+MathOptInterface = "0.9"
+MathProgBase = "~0.5.0, ~0.6, ~0.7"
+julia = "1"
+
+[extras]
+Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
+
+[targets]
+test = ["Test"]

Diff for: REQUIRE

@@ -1,8 +1,7 @@
 environment:
   matrix:
-  - julia_version: 0.6
-  - julia_version: 0.7
   - julia_version: 1.0
+  - julia_version: 1.1
 
 platform:
   - x86 # 32-bit

Diff for: appveyor.yml

@@ -7,12 +7,10 @@
 # Contains the wrapper itself
 #############################################################################
 
-__precompile__()
 module ECOS
 
-using Compat
-using Compat.SparseArrays
-using Compat.LinearAlgebra
+using SparseArrays
+using LinearAlgebra
 
 # Try to load the binary dependency
 if isfile(joinpath(dirname(@__FILE__),"..","deps","deps.jl"))

Diff for: src/ECOS.jl

@@ -11,12 +11,14 @@ struct Solution
     dual_eq::Vector{Float64}
     dual_ineq::Vector{Float64}
     slack::Vector{Float64}
-    objval::Float64
-    objbnd::Float64
+    objective_value::Float64
+    dual_objective_value::Float64
+    objective_constant::Float64
+    solve_time::Float64
 end
 const OPTIMIZE_NOT_CALLED = -1
 Solution() = Solution(OPTIMIZE_NOT_CALLED, Float64[], Float64[], Float64[],
-                      Float64[], NaN, NaN)
+                      Float64[], NaN, NaN, NaN, NaN)
 
 # Used to build the data with allocate-load during `copy_to`.
 # When `optimize!` is called, a the data is used to build `ECOSMatrix`
@@ -32,7 +34,7 @@ mutable struct ModelData
     JG::Vector{Int} # List of equality cols
     VG::Vector{Float64} # List of equality coefficients
     h::Vector{Float64} # List of equality coefficients
-    objconstant::Float64 # The objective is min c'x + objconstant
+    objective_constant::Float64 # The objective is min c'x + objective_constant
     c::Vector{Float64}
 end
 
@@ -58,14 +60,33 @@ mutable struct Optimizer <: MOI.AbstractOptimizer
     maxsense::Bool
     data::Union{Nothing, ModelData} # only non-Nothing between MOI.copy_to and MOI.optimize!
     sol::Solution
-    options
+    silent::Bool
+    options::Dict{Symbol, Any}
     function Optimizer(; kwargs...)
-        new(ConeData(), false, nothing, Solution(), kwargs)
+        optimizer = new(ConeData(), false, nothing, Solution(), false, Dict{Symbol, Any}())
+        for (key, value) in kwargs
+            MOI.set(optimizer, MOI.RawParameter(key), value)
+        end
+        return optimizer
     end
 end
 
 MOI.get(::Optimizer, ::MOI.SolverName) = "ECOS"
 
+function MOI.set(optimizer::Optimizer, param::MOI.RawParameter, value)
+    optimizer.options[param.name] = value
+end
+function MOI.get(optimizer::Optimizer, param::MOI.RawParameter)
+    # TODO: This gives a poor error message if the name of the parameter is invalid.
+    return optimizer.options[param.name]
+end
+
+MOI.supports(::Optimizer, ::MOI.Silent) = true
+function MOI.set(optimizer::Optimizer, ::MOI.Silent, value::Bool)
+    optimizer.silent = value
+end
+MOI.get(optimizer::Optimizer, ::MOI.Silent) = optimizer.silent
+
 function MOI.is_empty(instance::Optimizer)
     !instance.maxsense && instance.data === nothing
 end
@@ -97,7 +118,7 @@ function MOI.copy_to(dest::Optimizer, src::MOI.ModelLike; kws...)
     return MOIU.automatic_copy_to(dest, src; kws...)
 end
 
-using Compat.SparseArrays
+using SparseArrays
 
 # Computes cone dimensions
 constroffset(cone::ConeData, ci::CI{<:MOI.AbstractFunction, MOI.Zeros}) = ci.value
@@ -152,7 +173,7 @@ expmap(i) = (1, 3, 2)[i]
 function orderidx(idx, s::MOI.ExponentialCone)
     expmap.(idx)
 end
-function MOIU.load_constraint(instance::Optimizer, ci, f::MOI.VectorAffineFunction, s::MOI.AbstractVectorSet)
+function MOIU.load_constraint(instance::Optimizer, ci::MOI.ConstraintIndex, f::MOI.VectorAffineFunction, s::MOI.AbstractVectorSet)
     A = sparse(output_index.(f.terms), variable_index_value.(f.terms), coefficient.(f.terms))
     # sparse combines duplicates with + but does not remove zeros created so we call dropzeros!
     dropzeros!(A)
@@ -214,7 +235,7 @@ function MOIU.load(instance::Optimizer, ::MOI.ObjectiveFunction,
                    f::MOI.ScalarAffineFunction)
     c0 = Vector(sparsevec(variable_index_value.(f.terms), coefficient.(f.terms),
                           instance.data.n))
-    instance.data.objconstant = f.constant
+    instance.data.objective_constant = f.constant
     instance.data.c = instance.maxsense ? -c0 : c0
     return nothing
 end
@@ -231,28 +252,61 @@ function MOI.optimize!(instance::Optimizer)
     b = instance.data.b
     G = ECOS.ECOSMatrix(sparse(instance.data.IG, instance.data.JG, instance.data.VG, m, n))
     h = instance.data.h
-    objconstant = instance.data.objconstant
+    objective_constant = instance.data.objective_constant
     c = instance.data.c
     instance.data = nothing # Allows GC to free instance.data before A is loaded to ECOS
+    options = instance.options
+    if instance.silent
+        options = copy(options)
+        options[:verbose] = false
+    end
     ecos_prob_ptr = ECOS.setup(n, m, cone.f, cone.l, length(cone.qa), cone.qa,
-                               cone.ep, G, A, c, h, b; instance.options...)
+                               cone.ep, G, A, c, h, b; options...)
     ret_val = ECOS.solve(ecos_prob_ptr)
+    stat = unsafe_load(unsafe_load(ecos_prob_ptr).info)
+    solve_time = stat.tsetup + stat.tsolve
     ecos_prob = unsafe_wrap(Array, ecos_prob_ptr, 1)[1]
     primal    = unsafe_wrap(Array, ecos_prob.x, n)[:]
     dual_eq   = unsafe_wrap(Array, ecos_prob.y, cone.f)[:]
     dual_ineq = unsafe_wrap(Array, ecos_prob.z, m)[:]
     slack     = unsafe_wrap(Array, ecos_prob.s, m)[:]
     ECOS.cleanup(ecos_prob_ptr, 0)
-    objval = (instance.maxsense ? -1 : 1) * dot(c, primal)
-    if ret_val != ECOS.ECOS_DINF
-        objval += objconstant
-    end
-    objbnd = -(dot(b, dual_eq) + dot(h, dual_ineq))
-    if ret_val != ECOS.ECOS_PINF
-        objbnd += objconstant
+    objective_value = (instance.maxsense ? -1 : 1) * stat.pcost
+    dual_objective_value = (instance.maxsense ? -1 : 1) * stat.dcost
+    instance.sol = Solution(ret_val, primal, dual_eq, dual_ineq, slack, objective_value,
+                            dual_objective_value, objective_constant, solve_time)
+end
+
+MOI.get(optimizer::Optimizer, ::MOI.SolveTime) = optimizer.sol.solve_time
+function MOI.get(optimizer::Optimizer, ::MOI.RawStatusString)
+    # Strings from https://github.com/ifa-ethz/ecos/blob/master/include/ecos.h
+    flag = optimizer.sol.ret_val
+    if flag == OPTIMIZE_NOT_CALLED
+        return "Optimize not called"
+    elseif flag == ECOS_OPTIMAL
+        return "Problem solved to optimality"
+    elseif flag == ECOS_OPTIMAL + ECOS_INACC_OFFSET
+        return "Problem solved to inaccurate optimality"
+    elseif flag == ECOS_PINF
+        return "Found certificate of primal infeasibility"
+    elseif flag == ECOS_PINF + ECOS_INACC_OFFSET
+        return "Found inaccurate certificate of primal infeasibility"
+    elseif flag == ECOS_DINF
+        return "Found certificate of dual infeasibility"
+    elseif flag == ECOS_DINF + ECOS_INACC_OFFSET
+        return "Found inaccurate certificate of dual infeasibility"
+    elseif flag == ECOS_MAXIT
+        return "Maximum number of iterations reached"
+    elseif flag == ECOS_NUMERICS
+        return "Search direction unreliable"
+    elseif flag == ECOS_OUTCONE
+        return "s or z got outside the cone, numerics?"
+    elseif flag == ECOS_SIGINT
+        return "solver interrupted by a signal/ctrl-c"
+    else
+        @assert flag == ECOS_FATAL
+        return "Unknown problem in solver"
     end
-    instance.sol = Solution(ret_val, primal, dual_eq, dual_ineq, slack, objval,
-                            objbnd)
 end
 
 # Implements getter for result value and statuses
@@ -279,8 +333,20 @@ function MOI.get(instance::Optimizer, ::MOI.TerminationStatus)
     end
 end
 
-MOI.get(instance::Optimizer, ::MOI.ObjectiveValue) = instance.sol.objval
-MOI.get(instance::Optimizer, ::MOI.ObjectiveBound) = instance.sol.objbnd
+function MOI.get(optimizer::Optimizer, ::MOI.ObjectiveValue)
+    value = optimizer.sol.objective_value
+    if !MOIU.is_ray(MOI.get(optimizer, MOI.PrimalStatus()))
+        value += optimizer.sol.objective_constant
+    end
+    return value
+end
+function MOI.get(optimizer::Optimizer, ::MOI.DualObjectiveValue)
+    value = optimizer.sol.dual_objective_value
+    if !MOIU.is_ray(MOI.get(optimizer, MOI.DualStatus()))
+        value += optimizer.sol.objective_constant
+    end
+    return value
+end
 
 function MOI.get(instance::Optimizer, ::MOI.PrimalStatus)
     flag = instance.sol.ret_val

Diff for: src/MOI_wrapper.jl

@@ -1,5 +1,4 @@
-using Compat
-using Compat.Test
+using Test
 
 using MathOptInterface
 const MOI = MathOptInterface
@@ -8,7 +7,8 @@ const MOIU = MOI.Utilities
 const MOIB = MOI.Bridges
 
 import ECOS
-const optimizer = ECOS.Optimizer(verbose=false)
+const optimizer = ECOS.Optimizer()
+MOI.set(optimizer, MOI.Silent(), true)
 
 @testset "SolverName" begin
     @test MOI.get(optimizer, MOI.SolverName()) == "ECOS"
@@ -19,18 +19,10 @@ end
     @test !MOIU.supports_allocate_load(optimizer, true)
 end
 
-MOIU.@model(ECOSModelData,
-            (), (), # No scalar functions
-            (MOI.Zeros, MOI.Nonnegatives, MOI.Nonpositives, MOI.SecondOrderCone,
-             MOI.ExponentialCone),
-            (),
-            (), (), # No scalar sets
-            (), (MOI.VectorAffineFunction,))
 # UniversalFallback is needed for starting values, even if they are ignored by ECOS
-const cache = MOIU.UniversalFallback(ECOSModelData{Float64}())
+const cache = MOIU.UniversalFallback(MOIU.Model{Float64}())
 const cached = MOIU.CachingOptimizer(cache, optimizer)
 
-# Essential bridges that are needed for all tests
 const bridged = MOIB.full_bridge_optimizer(cached, Float64)
 
 # SOC2 requires 1e-4
@@ -39,10 +31,16 @@ const config = MOIT.TestConfig(atol=1e-4, rtol=1e-4)
 @testset "Unit" begin
     MOIT.unittest(bridged,
                   config,
-                  [# Need https://github.com/JuliaOpt/MathOptInterface.jl/issues/529
-                   "solve_qp_edge_cases",
-                   # Integer and ZeroOne sets are not supported
-                   "solve_integer_edge_cases", "solve_objbound_edge_cases"])
+                  [
+        # `TimeLimitSec` not supported.
+        "time_limit_sec",
+        # Need https://github.com/JuliaOpt/MathOptInterface.jl/issues/529
+        "solve_qp_edge_cases",
+        # Integer and ZeroOne sets are not supported
+        "solve_integer_edge_cases", "solve_objbound_edge_cases",
+        "solve_zero_one_with_bounds_1",
+        "solve_zero_one_with_bounds_2",
+        "solve_zero_one_with_bounds_3"])
 end
 
 @testset "Continuous linear problems" begin
@@ -54,6 +52,6 @@ end
 end
 
 @testset "Continuous conic problems" begin
-    exclude = ["sdp", "rootdet", "logdet"]
+    exclude = ["pow", "sdp", "rootdet", "logdet"]
     MOIT.contconictest(bridged, config, exclude)
 end

Diff for: test/MOI_wrapper.jl

@@ -7,8 +7,8 @@
 # Test the MathProgBase.jl interface for the ECOS.jl solver wrapper
 #############################################################################
 
-using Compat.Test
-using Compat.LinearAlgebra
+using Test
+using LinearAlgebra
 using MathProgBase
 using ECOS
 

Diff for: test/MPB_linear.jl

@@ -7,9 +7,9 @@
 # Test the direct interface for the ECOS.jl solver wrapper
 #############################################################################
 
-using Compat.Test
-using Compat.LinearAlgebra
-using Compat.SparseArrays
+using Test
+using LinearAlgebra
+using SparseArrays
 
 # The values below are copied from data.h in ECOS source code
 import ECOS

Diff for: test/direct.jl

@@ -8,7 +8,7 @@
 #############################################################################
 
 import ECOS
-using Compat.Test
+using Test
 
 println(ECOS.ver())
 

Diff for: test/options.jl

@@ -1,5 +1,5 @@
 # Generates a random cone problem to test the MPB interface perf
-using Compat.SparseArrays
+using SparseArrays
 
 rows = 10000
 cols = 10000