support sparse array conversion

.github/workflows/aot.yml

@@ -36,7 +36,7 @@ jobs:
           python-version: ${{ matrix.python-version }}
           architecture: ${{ matrix.architecture }}
       - run: python --version
-      - run: python -m pip install --user numpy
+      - run: python -m pip install --user numpy scipy
       - name: Install libpython2.7 for `find_libpython` test
         run: sudo apt-get install python2.7-dev
         if: ${{ matrix.python-version != '2.7' && matrix.os == 'ubuntu-latest' }}

.github/workflows/system.yml

@@ -64,7 +64,7 @@ jobs:
         with:
           python-version: ${{ matrix.python-version }}
           architecture: ${{ matrix.architecture }}
-      - run: python -m pip install --user numpy
+      - run: python -m pip install --user numpy scipy
       - run: python -m pip install virtualenv
         # virtualenv test with Python 2.7 is failing for some reason.
         # Skipping it for now.

Project.toml

@@ -10,6 +10,7 @@ Libdl = "8f399da3-3557-5675-b5ff-fb832c97cbdb"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 MacroTools = "1914dd2f-81c6-5fcd-8719-6d5c9610ff09"
 Serialization = "9e88b42a-f829-5b0c-bbe9-9e923198166b"
+SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
 VersionParsing = "81def892-9a0e-5fdd-b105-ffc91e053289"
 
 [compat]

src/conversions.jl

@@ -426,6 +426,9 @@ PyObject(a::BitArray) = PyObject(Array(a))
 # NumPy conversions (multidimensional arrays)
 include("numpy.jl")
 
+# SciPy conversions
+include("scipy.jl")
+
 #########################################################################
 # PyDict: no-copy wrapping of a Julia object around a Python dictionary
 

src/scipy.jl

@@ -0,0 +1,19 @@
+using SparseArrays
+
+const scipysparse_ = PyNULL()
+
+function scipysparse()
+    if ispynull(scipysparse_)
+        copy!(scipysparse_, pyimport("scipy.sparse"))
+    end
+    return scipysparse_
+end
+
+function PyObject(S::SparseMatrixCSC)
+    scipysparse().csc_matrix((S.nzval, S.rowval .- 1, S.colptr .- 1), shape = size(S))
+end
+
+function convert(::Type{SparseMatrixCSC}, o::PyObject)
+    I, J, V = scipysparse().find(o)
+    return sparse(I .+ 1, J .+ 1, V, o.shape...)
+end

test/runtests.jl

@@ -1,6 +1,7 @@
 using PyCall
 using PyCall: hasproperty
 using Test, Dates, Serialization
+using SparseArrays: SparseMatrixCSC, sparse, sprand
 
 filter(f, itr) = collect(Iterators.filter(f, itr))
 filter(f, d::AbstractDict) = Base.filter(f, d)
@@ -13,6 +14,11 @@ PYTHONEXECUTABLE=get(ENV,"PYTHONEXECUTABLE","")
 @testset "CI setup" begin
     if lowercase(get(ENV, "CI", "false")) == "true"
         @test !ispynull(pyimport_e("numpy"))
+        try
+            pyimport_conda("scipy.sparse", "scipy")
+        catch
+        end
+        @test !ispynull(pyimport_e("scipy.sparse"))
     end
 end
 
@@ -615,6 +621,11 @@ const PyInt = pyversion < v"3" ? Int : Clonglong
     @test float(PyObject(1+2im)) === 1.0 + 2.0im
     @test float(PyObject([1,2,3]))[2] === 2.0
     @test_throws ArgumentError float(pybuiltin("type"))
+
+    # sparse array conversion
+    S = sprand(10, 10, 0.2)
+    PyS = PyObject(S)
+    @test convert(SparseMatrixCSC, PyS) == S
 end
 
 ######################################################################