Merge pull request #586 from SciML/sparspak

Throw proper error for Sparspak missing

Author: ChrisRackauckas

.github/workflows/Tests.yml

@@ -29,6 +29,7 @@ jobs:
           - "pre"
         group:
           - "Core"
+          - "DefaultsLoading"
           - "LinearSolveHYPRE"
           - "LinearSolvePardiso"
           - "LinearSolveBandedMatrices"

ext/LinearSolveSparseArraysExt.jl

@@ -3,6 +3,7 @@ module LinearSolveSparseArraysExt
 using LinearSolve, LinearAlgebra
 using SparseArrays
 using SparseArrays: AbstractSparseMatrixCSC, nonzeros, rowvals, getcolptr
+using LinearSolve: BLASELTYPES
 
 # Can't `using KLU` because cannot have a dependency in there without
 # requiring the user does `using KLU`
@@ -39,7 +40,7 @@ function LinearSolve.handle_sparsematrixcsc_lu(A::AbstractSparseMatrixCSC)
 end
 
 function LinearSolve.defaultalg(
-        A::Symmetric{<:Number, <:SparseMatrixCSC}, b, ::OperatorAssumptions{Bool})
+        A::Symmetric{<:BLASELTYPES, <:SparseMatrixCSC}, b, ::OperatorAssumptions{Bool})
     LinearSolve.DefaultLinearSolver(LinearSolve.DefaultAlgorithmChoice.CHOLMODFactorization)
 end
 
@@ -78,7 +79,7 @@ function LinearSolve.init_cacheval(
 end
 
 function LinearSolve.init_cacheval(
-        alg::UMFPACKFactorization, A::AbstractSparseArray, b, u, Pl, Pr,
+        alg::UMFPACKFactorization, A::AbstractSparseArray{Float64}, b, u, Pl, Pr,
         maxiters::Int, abstol,
         reltol,
         verbose::Bool, assumptions::OperatorAssumptions)
@@ -136,7 +137,7 @@ function LinearSolve.init_cacheval(
 end
 
 function LinearSolve.init_cacheval(
-        alg::KLUFactorization, A::AbstractSparseArray, b, u, Pl, Pr,
+        alg::KLUFactorization, A::AbstractSparseArray{Float64}, b, u, Pl, Pr,
         maxiters::Int, abstol,
         reltol,
         verbose::Bool, assumptions::OperatorAssumptions)
@@ -186,15 +187,15 @@ function LinearSolve.init_cacheval(alg::CHOLMODFactorization,
         Pl, Pr,
         maxiters::Int, abstol, reltol,
         verbose::Bool, assumptions::OperatorAssumptions) where {T <:
-                                                                Union{Float32, Float64}}
+                                                                BLASELTYPES}
     PREALLOCATED_CHOLMOD
 end
 
 function LinearSolve.init_cacheval(alg::NormalCholeskyFactorization,
-        A::Union{AbstractSparseArray, LinearSolve.GPUArraysCore.AnyGPUArray,
-            Symmetric{<:Number, <:AbstractSparseArray}}, b, u, Pl, Pr,
+        A::Union{AbstractSparseArray{T}, LinearSolve.GPUArraysCore.AnyGPUArray,
+            Symmetric{T, <:AbstractSparseArray{T}}}, b, u, Pl, Pr,
         maxiters::Int, abstol, reltol, verbose::Bool,
-        assumptions::OperatorAssumptions)
+        assumptions::OperatorAssumptions) where {T <: BLASELTYPES}
     LinearSolve.ArrayInterface.cholesky_instance(convert(AbstractMatrix, A))
 end
 

src/default.jl

@@ -229,7 +229,7 @@ function algchoice_to_alg(alg::Symbol)
     elseif alg === :DirectLdiv!
         DirectLdiv!()
     elseif alg === :SparspakFactorization
-        SparspakFactorization()
+        SparspakFactorization(throwerror = false)
     elseif alg === :KLUFactorization
         KLUFactorization()
     elseif alg === :UMFPACKFactorization

src/factorization.jl

@@ -319,7 +319,8 @@ function init_cacheval(alg::CholeskyFactorization, A::GPUArraysCore.AnyGPUArray,
     cholesky(A; check = false)
 end
 
-function init_cacheval(alg::CholeskyFactorization, A, b, u, Pl, Pr,
+function init_cacheval(
+        alg::CholeskyFactorization, A::AbstractArray{<:BLASELTYPES}, b, u, Pl, Pr,
         maxiters::Int, abstol, reltol, verbose::Bool, assumptions::OperatorAssumptions)
     ArrayInterface.cholesky_instance(convert(AbstractMatrix, A), alg.pivot)
 end
@@ -333,7 +334,7 @@ function init_cacheval(alg::CholeskyFactorization, A::Matrix{Float64}, b, u, Pl,
 end
 
 function init_cacheval(alg::CholeskyFactorization,
-        A::Union{Diagonal, AbstractSciMLOperator}, b, u, Pl, Pr,
+        A::Union{Diagonal, AbstractSciMLOperator, AbstractArray}, b, u, Pl, Pr,
         maxiters::Int, abstol, reltol, verbose::Bool,
         assumptions::OperatorAssumptions)
     nothing
@@ -1044,8 +1045,17 @@ dispatch to route around standard BLAS routines in the case e.g. of arbitrary-pr
 floating point numbers or ForwardDiff.Dual.
 This e.g. allows for Automatic Differentiation (AD) of a sparse-matrix solve.
 """
-Base.@kwdef struct SparspakFactorization <: AbstractSparseFactorization
-    reuse_symbolic::Bool = true
+struct SparspakFactorization <: AbstractSparseFactorization
+    reuse_symbolic::Bool
+
+    function SparspakFactorization(; reuse_symbolic = true, throwerror = true)
+        ext = Base.get_extension(@__MODULE__, :LinearSolveSparspakExt)
+        if throwerror && ext === nothing
+            error("SparspakFactorization requires that Sparspak is loaded, i.e. `using Sparspak`")
+        else
+            new(reuse_symbolic)
+        end
+    end
 end
 
 function init_cacheval(alg::SparspakFactorization,

test/defaults_loading.jl

@@ -0,0 +1,34 @@
+using SparseArrays
+using LinearSolve
+using Test
+
+n = 10
+dx = 1 / n
+dx2 = dx^-2
+vals = Vector{BigFloat}(undef, 0)
+cols = Vector{Int}(undef, 0)
+rows = Vector{Int}(undef, 0)
+for i in 1:n
+    if i != 1
+        push!(vals, dx2)
+        push!(cols, i - 1)
+        push!(rows, i)
+    end
+    push!(vals, -2dx2)
+    push!(cols, i)
+    push!(rows, i)
+    if i != n
+        push!(vals, dx2)
+        push!(cols, i + 1)
+        push!(rows, i)
+    end
+end
+mat = sparse(rows, cols, vals, n, n)
+rhs = big.(zeros(n))
+rhs[begin] = rhs[end] = -2
+prob = LinearProblem(mat, rhs)
+@test_throws ["SparspakFactorization required", "using Sparspak"] sol=solve(prob).u
+
+using Sparspak
+sol = solve(prob).u
+@test sol isa Vector{BigFloat}

test/runtests.jl

@@ -25,6 +25,10 @@ if GROUP == "All" || GROUP == "Enzyme"
     @time @safetestset "Enzyme Derivative Rules" include("enzyme.jl")
 end
 
+if GROUP == "All" || GROUP == "DefaultsLoading"
+    @time @safetestset "Enzyme Derivative Rules" include("defaults_loading.jl")
+end
+
 if GROUP == "LinearSolveCUDA"
     Pkg.activate("gpu")
     Pkg.develop(PackageSpec(path = dirname(@__DIR__)))