Fix deprecations and clean tests & CI (#19)

* Fix deprecations

* Use separate test environment and test GLPK as well

* Use eachrow and eachcol

* Update CI

* Fix format

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

Author: devmotion

.github/workflows/CI.yml

@@ -38,8 +38,6 @@ jobs:
         with:
           python-version: '3.9'
           architecture: ${{ matrix.arch }}
-      # Limitation of pip: https://pythonot.github.io/index.html#pip-installation
-      - run: python -m pip install cython numpy
       - run: python -m pip install pot
       - uses: julia-actions/setup-julia@v1
         with:
@@ -63,7 +61,7 @@ jobs:
           PYTHON: python
       - uses: julia-actions/julia-processcoverage@v1
         if: matrix.version == '1' && matrix.os == 'ubuntu-latest'
-      - uses: codecov/codecov-action@v1
+      - uses: codecov/codecov-action@v2
         if: matrix.version == '1' && matrix.os == 'ubuntu-latest'
         with:
           file: lcov.info

.github/workflows/CompatHelper.yml

@@ -25,7 +25,7 @@ jobs:
       - name: "Run CompatHelper"
         run: |
           import CompatHelper
-          CompatHelper.main(; subdirs=["", "docs"], bump_version=true)
+          CompatHelper.main(; subdirs=["", "docs", "test"], bump_version=true)
         shell: julia --color=yes {0}
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/Documentation.yml

@@ -23,8 +23,6 @@ jobs:
         with:
           python-version: '3.9'
           architecture: ${{ matrix.arch }}
-      # Limitation of pip: https://pythonot.github.io/index.html#pip-installation
-      - run: python -m pip install cython numpy
       - run: python -m pip install pot
       - uses: julia-actions/setup-julia@latest
         with:

Project.toml

@@ -1,7 +1,7 @@
 name = "ExactOptimalTransport"
 uuid = "24df6009-d856-477c-ac5c-91f668376b31"
 authors = ["JuliaOptimalTransport"]
-version = "0.2.0"
+version = "0.2.1"
 
 [deps]
 Distances = "b4f34e82-e78d-54a5-968a-f98e89d6e8f7"
@@ -18,19 +18,9 @@ StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
 Distances = "0.9.0, 0.10"
 Distributions = "0.24, 0.25"
 FillArrays = "0.12"
-MathOptInterface = "0.9, 0.10"
+MathOptInterface = "0.10"
 PDMats = "0.10, 0.11"
 QuadGK = "2"
 StatsBase = "0.33.8"
 julia = "1.6"
 
-[extras]
-HCubature = "19dc6840-f33b-545b-b366-655c7e3ffd49"
-PythonOT = "3c485715-4278-42b2-9b5f-8f00e43c12ef"
-Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
-SafeTestsets = "1bc83da4-3b8d-516f-aca4-4fe02f6d838f"
-Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
-Tulip = "6dd1b50a-3aae-11e9-10b5-ef983d2400fa"
-
-[targets]
-test = ["PythonOT", "Random", "SafeTestsets", "Test", "Tulip", "HCubature"]

src/exact.jl

@@ -71,21 +71,21 @@ function emd(μ, ν, C, model::MOI.ModelLike)
 
     # add non-negativity constraints
     for xi in x
-        MOI.add_constraint(model, MOI.SingleVariable(xi), MOI.GreaterThan(zero_T))
+        MOI.add_constraint(model, xi, MOI.GreaterThan(zero_T))
     end
 
     # add constraints for source
-    for (i, μi) in zip(axes(xmat, 1), μ) # eachrow(xmat) is not available on Julia 1.0
+    for (xrow, μi) in zip(eachrow(xmat), μ)
         f = MOI.ScalarAffineFunction(
-            [MOI.ScalarAffineTerm(one(μi), xi) for xi in view(xmat, i, :)], zero(μi)
+            [MOI.ScalarAffineTerm(one(μi), xi) for xi in xrow], zero(μi)
         )
         MOI.add_constraint(model, f, MOI.EqualTo(μi))
     end
 
     # add constraints for target
-    for (i, νi) in zip(axes(xmat, 2), ν) # eachcol(xmat) is not available on Julia 1.0
+    for (xcol, νi) in zip(eachcol(xmat), ν)
         f = MOI.ScalarAffineFunction(
-            [MOI.ScalarAffineTerm(one(νi), xi) for xi in view(xmat, :, i)], zero(νi)
+            [MOI.ScalarAffineTerm(one(νi), xi) for xi in xcol], zero(νi)
         )
         MOI.add_constraint(model, f, MOI.EqualTo(νi))
     end

test/Project.toml

@@ -0,0 +1,28 @@
+[deps]
+Distances = "b4f34e82-e78d-54a5-968a-f98e89d6e8f7"
+Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
+FillArrays = "1a297f60-69ca-5386-bcde-b61e274b549b"
+GLPK = "60bf3e95-4087-53dc-ae20-288a0d20c6a6"
+HCubature = "19dc6840-f33b-545b-b366-655c7e3ffd49"
+LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
+MathOptInterface = "b8f27783-ece8-5eb3-8dc8-9495eed66fee"
+PDMats = "90014a1f-27ba-587c-ab20-58faa44d9150"
+PythonOT = "3c485715-4278-42b2-9b5f-8f00e43c12ef"
+Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
+SafeTestsets = "1bc83da4-3b8d-516f-aca4-4fe02f6d838f"
+SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
+Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
+Tulip = "6dd1b50a-3aae-11e9-10b5-ef983d2400fa"
+
+[compat]
+Distances = "0.9, 0.10"
+Distributions = "0.24, 0.25"
+FillArrays = "0.12"
+GLPK = "0.15"
+HCubature = "1.5"
+MathOptInterface = "0.10"
+PDMats = "0.10, 0.11"
+PythonOT = "0.1"
+SafeTestsets = "0.0.1"
+Tulip = "0.9"
+julia = "1.6"

test/exact.jl

@@ -2,6 +2,7 @@ using ExactOptimalTransport
 
 using Distances
 using FillArrays
+using GLPK
 using PythonOT: PythonOT
 using Tulip
 using MathOptInterface
@@ -32,30 +33,34 @@ Random.seed!(100)
             pot_P = POT.emd(μ, ν, C)
             pot_cost = POT.emd2(μ, ν, C)
 
-            # compute optimal transport map and cost with Tulip
-            lp = Tulip.Optimizer()
-            P = emd(μ, ν, C, lp)
-            @test size(C) == size(P)
-            @test MOI.get(lp, MOI.TerminationStatus()) == MOI.OPTIMAL
-            @test maximum(abs, P .- pot_P) < 1e-2
-
-            lp = Tulip.Optimizer()
-            cost = emd2(μ, ν, C, lp)
-            @test dot(C, P) ≈ cost atol = 1e-5
-            @test MOI.get(lp, MOI.TerminationStatus()) == MOI.OPTIMAL
-            @test cost ≈ pot_cost atol = 1e-5
+            # compute optimal transport map and cost with Tulip and GLPK
+            for T in (Tulip.Optimizer, GLPK.Optimizer)
+                lp = T()
+                P = emd(μ, ν, C, lp)
+                @test size(C) == size(P)
+                @test MOI.get(lp, MOI.TerminationStatus()) == MOI.OPTIMAL
+                @test maximum(abs, P .- pot_P) < 1e-2
+
+                lp = T()
+                cost = emd2(μ, ν, C, lp)
+                @test dot(C, P) ≈ cost atol = 1e-5
+                @test MOI.get(lp, MOI.TerminationStatus()) == MOI.OPTIMAL
+                @test cost ≈ pot_cost atol = 1e-5
+            end
         end
 
         @testset "pre-computed plan" begin
             # create random cost matrix
             C = pairwise(SqEuclidean(), rand(1, M), rand(1, N); dims=2)
 
-            # compute optimal transport map
-            P = emd(μ, ν, C, Tulip.Optimizer())
+            # compute optimal transport map with Tulip and GLPK
+            for T in (Tulip.Optimizer, GLPK.Optimizer)
+                P = emd(μ, ν, C, T())
 
-            # do not use μ and ν to ensure that provided map is used
-            cost = emd2(similar(μ), similar(ν), C, Tulip.Optimizer(); plan=P)
-            @test cost ≈ emd2(μ, ν, C, Tulip.Optimizer())
+                # do not use μ and ν to ensure that provided map is used
+                cost = emd2(similar(μ), similar(ν), C, T(); plan=P)
+                @test cost ≈ emd2(μ, ν, C, T())
+            end
         end
 
         # https://github.com/JuliaOptimalTransport/OptimalTransport.jl/issues/71
@@ -106,8 +111,10 @@ Random.seed!(100)
             xs = rand(μ, m)
             μdiscrete = fill(1 / m, m)
             C = pairwise(Euclidean(), xs', (1:length(νprobs))'; dims=2)
-            c2 = emd2(μdiscrete, νprobs, C, Tulip.Optimizer())
-            @test c2 ≈ c rtol = 1e-1
+            for optimizer in (Tulip.Optimizer(), GLPK.Optimizer())
+                c2 = emd2(μdiscrete, νprobs, C, optimizer)
+                @test c2 ≈ c rtol = 1e-1
+            end
         end
 
         @testset "discrete case" begin
@@ -164,8 +171,10 @@ Random.seed!(100)
                 # DiscreteNonParametric sorts the support automatically, here we have to sort
                 # manually
                 C = pairwise(Euclidean(), μsupport', νsupport'; dims=2)
-                c2 = emd2(μprobs, νprobs, C, Tulip.Optimizer())
-                @test c2 ≈ c rtol = 1e-5
+                for optimizer in (Tulip.Optimizer(), GLPK.Optimizer())
+                    c2 = emd2(μprobs, νprobs, C, optimizer)
+                    @test c2 ≈ c rtol = 1e-5
+                end
 
                 # compare with POT
                 # disabled currently since https://github.com/PythonOT/POT/issues/169 causes bounds
@@ -222,8 +231,10 @@ Random.seed!(100)
             μprobs = normalize!(pdf(μ, μsupp'), 1)
             νprobs = normalize!(pdf(ν, νsupp'), 1)
             C = pairwise(SqEuclidean(), μsupp', νsupp'; dims=2)
-            @test emd2(μprobs, νprobs, C, Tulip.Optimizer()) ≈ ot_cost(SqEuclidean(), μ, ν) rtol =
-                1e-3
+            for optimizer in (Tulip.Optimizer(), GLPK.Optimizer())
+                @test emd2(μprobs, νprobs, C, optimizer) ≈ ot_cost(SqEuclidean(), μ, ν) rtol =
+                    1e-3
+            end
 
             # Use hcubature integration to perform ``\\int c(x,T(x)) d\\mu``
             T = ot_plan(SqEuclidean(), μ, ν)