Merge pull request #46 from slimgroup/pythoncall

switch to pythoncall

.github/workflows/ci-joli.yaml

@@ -19,7 +19,7 @@ jobs:
       fail-fast: false
 
       matrix:
-        version: ['1.6', '1.7', '1']
+        version: ['lts', '1.10', '1']
         os:
           - ubuntu-latest
           - macos-latest
@@ -31,29 +31,14 @@ jobs:
         uses: actions/checkout@v2
 
       - name: Setup julia
-        uses: julia-actions/setup-julia@v1
+        uses: julia-actions/setup-julia@v2
         with:
           version: ${{ matrix.version }}
           arch: ${{ matrix.arch }}
-
-      - name: Set up Python 3.7
-        uses: actions/setup-python@v1
-        with:
-          python-version: 3.7
-
-      - name: Install wavelets
-        run: |
-          python -m pip install --upgrade pip
-          pip install PyWavelets
 
       - name: Build JOLI
         uses: julia-actions/julia-buildpkg@latest
 
-      - name: Rebuild PyCall with system python
-        run: |
-          export PYTHON=$(which python)
-          julia --color=yes --check-bounds=yes --project -e 'using Pkg; Pkg.build("PyCall")'
-
       - name: Run tests
         uses: julia-actions/julia-runtest@latest
 

CondaPkg.toml

@@ -0,0 +1,3 @@
+[deps]
+python = ""
+pywavelets = ""

Project.toml

@@ -1,7 +1,7 @@
 name = "JOLI"
 uuid = "bb331ad6-a1cf-11e9-23da-9bcb53c69f6f"
 authors = ["Henryk Modzelewski <[email protected]>"]
-version = "0.8.5"
+version = "0.9.0"
 
 [deps]
 ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
@@ -16,7 +16,7 @@ LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 NFFT = "efe261a4-0d2b-5849-be55-fc731d526b0d"
 Nullables = "4d1e1d77-625e-5b40-9113-a560ec7a8ecd"
 Printf = "de0858da-6303-5e67-8744-51eddeeeb8d7"
-PyCall = "438e738f-606a-5dbb-bf0a-cddfbfd45ab0"
+PythonCall = "6099a3de-0909-46bc-b1f4-468b9a2dfc0d"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 SharedArrays = "1a1011a3-84de-559e-8e89-a11a2f7dc383"
 SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
@@ -27,12 +27,12 @@ Wavelets = "29a6e085-ba6d-5f35-a997-948ac2efa89a"
 ChainRulesCore = "1"
 DistributedArrays = "0.5, 0.6"
 FFTW = "1"
-Flux = "0.12, 0.13"
+Flux = "0.12, 0.13, 0.14"
 InplaceOps = "0.3.0"
 IterativeSolvers = "0.8, 0.9"
 NFFT = "0.6 - 0.13"
 Nullables = "1"
-PyCall = "1.18, 1.90, 1.91, 1.62"
+PythonCall = "0.9"
 SpecialFunctions = "1.2, 2"
 Wavelets = "0.8, 0.9"
 julia = "1"

README.md

@@ -53,7 +53,7 @@ JOLI is registered and can be added like any standard julia package with the com
 
 - **CurveLab**: In order to use `joCurvelet2D` or `joCurvelet2DnoFFT` operators, you need to obtain *CurveLab-2.1.2-SLIM*, a SLIM extension to *CurveLab-2.1.2*. The tarball of this extension is available from [curvelet.org](http://www.curvelet.org) under [Software](http://www.curvelet.org/software.html) tab. The installation instructions are included in the tarball of *CurveLab-2.1.2-SLIM*. Note, that CurveLab is free only for academic use and requires registration.
 
-- **PyWavelets**: In order to use `joSWT` that implements 1D stationary/shift invariant wavelet transform, the [PyWavelets](https://github.com/PyWavelets/pywt) package needs to be installed within the python environement used by [PyCall.jl](https://github.com/JuliaPy/PyCall.jl). By default, it will use [Conda.jl](https://github.com/JuliaPy/Conda.jl) and you should follow its directiv to install a package within it. If you have configured `PyCall` with your own python environment you can install `PyWavelets` via `pip install --upgrade PyWavelets`.
+- **PyWavelets**: In order to use `joSWT` that implements 1D stationary/shift invariant wavelet transform, the [PyWavelets](https://github.com/PyWavelets/pywt) package needs to be installed within the python environment used by [PythonCall.jl](https://github.com/JuliaPy/PythonCall.jl). By default, it will use [CondaPkg.jl](https://github.com/JuliaPy/CondaPkg.jl) and you should follow its directiv to install a package within it. If you have configured `PythonCall` with your own python environment you can install `PyWavelets` via `pip install --upgrade PyWavelets`.
 
 ## Documentation (more to come)
 

src/JOLI.jl

@@ -51,10 +51,10 @@ using DistributedArrays.SPMD
 using LinearAlgebra
 using InplaceOps
 using IterativeSolvers
+using PythonCall
 using FFTW
 using NFFT
 using Wavelets
-using PyCall
 using SpecialFunctions
 using ChainRulesCore
 

src/joLinearFunctionConstructors/joSWT.jl

@@ -3,25 +3,32 @@
 ## helper module
 module joSWT_etc
     using JOLI: jo_convert
-    using PyCall
+    using PythonCall
+
+    const pywt = PythonCall.pynew()
+
+    function __init__()
+        PythonCall.pycopy!(pywt, pyimport("pywt"))
+    end
+
     # 1D
     function apply_swt(v::Vector{vdt}, wt::String, L::Integer,rdt::DataType,pad::Integer) where vdt<:Union{AbstractFloat,Complex}
-        pywt = pyimport("pywt")
         v = [v;zeros(vdt, pad)]
         rv = pywt.swt(v, wt, level=L, start_level=0, norm=true, trim_approx=true)
-        rv = vcat(rv...)
+        rv = vcat([PyArray(r) for r in rv]...)
         rv = jo_convert(rdt, rv, false)
         return rv
     end
+
     function apply_iswt(v::Vector{vdt},wt::String,L::Integer,rdt::DataType,pad::Integer) where vdt<:Union{AbstractFloat,Complex}
-        pywt = pyimport("pywt")
         v = reshape(v, :, L+1)
         v = [v[:, i] for i=1:L+1]
-        rv = pywt.iswt(v, wt, norm=true)
+        rv = PyArray(pywt.iswt(v, wt, norm=true))
         # Convert and remove pad
         rv = jo_convert(rdt, rv[1:end-pad], false)
         return rv
     end
+
 end
 using .joSWT_etc
 
@@ -68,7 +75,7 @@ examples with DDT/RDT
     joDWT(m,"sym5"; DDT=Float32,RDT=Float64)
 
 """
-function joSWT(m::Integer,wt::String="db20";
+function joSWT(m::Integer, wt::String="db20";
     L::Integer=maxtransformlevels(m + m%2),
     DDT::DataType=joFloat,RDT::DataType=DDT,
     name::String="joSWT")
@@ -84,3 +91,5 @@ function joSWT(m::Integer,wt::String="db20";
         DDT,RDT;
         name=name)
 end
+
+joSWT(m::Integer, wt::Py; kwargs...) = joSWT(m, string(wt); kwargs...)

src/joMixedConstructors/joGaussian.jl

@@ -5,7 +5,10 @@ module joGaussian_etc
     using JOLI: jo_convert, LocalVector
     using Random
     using LinearAlgebra
-    function fI(v::LocalVector{vdt},m::Integer,n::Integer,rng::AbstractRNG,rns::Array{UInt32,1},rdt::DataType) where vdt<:Number
+
+    const rngT = Union{Vector{UInt32}, UInt128}
+
+    function fI(v::LocalVector{vdt},m::Integer,n::Integer,rng::AbstractRNG,rns::rngT,rdt::DataType) where vdt<:Number
         rv = zeros(rdt,m)
         lrng=Random.seed!(rng,rns)
         for i=1:n
@@ -15,7 +18,7 @@ module joGaussian_etc
         rv = jo_convert(rdt,rv,false)
         return rv
     end
-    function aI(v::LocalVector{vdt},m::Integer,n::Integer,rng::AbstractRNG,rns::Array{UInt32,1},rdt::DataType) where vdt<:Number
+    function aI(v::LocalVector{vdt},m::Integer,n::Integer,rng::AbstractRNG,rns::rngT,rdt::DataType) where vdt<:Number
         rv = Vector{rdt}(undef,n)
         lrng=Random.seed!(rng,rns)
         for i=1:n
@@ -25,7 +28,7 @@ module joGaussian_etc
         rv = jo_convert(rdt,rv,false)
         return rv
     end
-    function scale(m::Integer,n::Integer,rng::AbstractRNG,rns::Array{UInt32,1},rdt::DataType)
+    function scale(m::Integer,n::Integer,rng::AbstractRNG,rns::rngT,rdt::DataType)
         rv = Vector{rdt}(undef,n)
         lrng=Random.seed!(rng,rns)
         for i=1:n
@@ -37,7 +40,7 @@ module joGaussian_etc
         rv = jo_convert(rdt,rv,false)
         return rv
     end
-    function fIN(v::LocalVector{vdt},m::Integer,n::Integer,scale::LocalVector{<:Number},rng::AbstractRNG,rns::Array{UInt32,1},rdt::DataType) where vdt<:Number
+    function fIN(v::LocalVector{vdt},m::Integer,n::Integer,scale::LocalVector{<:Number},rng::AbstractRNG,rns::rngT,rdt::DataType) where vdt<:Number
         rv = zeros(rdt,m)
         lrng=Random.seed!(rng,rns)
         for i=1:n
@@ -47,7 +50,7 @@ module joGaussian_etc
         rv = jo_convert(rdt,rv,false)
         return rv
     end
-    function aIN(v::LocalVector{vdt},m::Integer,n::Integer,scale::LocalVector{<:Number},rng::AbstractRNG,rns::Array{UInt32,1},rdt::DataType) where vdt<:Number
+    function aIN(v::LocalVector{vdt},m::Integer,n::Integer,scale::LocalVector{<:Number},rng::AbstractRNG,rns::rngT,rdt::DataType) where vdt<:Number
         rv = Vector{rdt}(undef,n)
         lrng=Random.seed!(rng,rns)
         for i=1:n
@@ -134,7 +137,7 @@ function joGaussian(M::Integer,N::Integer=M;
         return joMatrix(a;DDT=DDT,RDT=RDT,name=name*"_o")
     else
         if implicit
-            rngs=copy(RNG.seed)
+            rngs = copy(RNG.seed)
             if normalized
                 fscale = joGaussian_etc.scale(M,N,RNG,rngs,DDT)
                 ascale = joGaussian_etc.scale(M,N,RNG,rngs,RDT)

test/runtests.jl

@@ -5,7 +5,7 @@ using SparseArrays
 using InplaceOps
 using FFTW
 using Wavelets
-using PyCall
+using PythonCall
 using JOLI
 
 jo_type_mismatch_error_set(false)

test/test_joSWT.jl

@@ -4,11 +4,11 @@ try
     global swt = pywt.swt
     global pywavelet = true
 catch
-    @warn "Skipping joSWTtests - PyCall clouldn't import PyWavelets"
+    @warn "Skipping joSWTtests - PythonCall clouldn't import PyWavelets"
 end
 
 families = ("haar", "db", "sym", "coif")
-pywavelet ? wavelets = vcat([pywt.wavelist(name) for name in families]...) : wavelets = []
+pywavelet ? wavelets = vcat([pyconvert(Vector{String}, pywt.wavelist(name)) for name in families]...) : wavelets = []
 
 tsname="joSWT"
 @testset "$tsname" begin