Add Aqua.jl to CI

.github/workflows/ci.yml

@@ -40,6 +40,24 @@ jobs:
           files: lcov.info
           token: ${{ secrets.CODECOV_TOKEN }}
           fail_ci_if_error: false
+  aqua:
+    name: Aqua
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+      - uses: julia-actions/setup-julia@v3
+        with:
+          version: '1'
+      - uses: julia-actions/cache@v3
+      - run: |
+          julia --color=yes -e '
+            using Pkg
+            Pkg.activate(temp = true)
+            Pkg.develop(PackageSpec(path = pwd()))
+            Pkg.add(name = "Aqua", version = "0.8")
+            using Aqua, SuiteSparseGraphBLAS
+            # ambiguities: ~190 pre-existing.
+            Aqua.test_all(SuiteSparseGraphBLAS; ambiguities = false)'
   docs:
     name: Documentation
     runs-on: ubuntu-latest

Project.toml

@@ -23,9 +23,15 @@ SuiteSparse = "4607b0f0-06f3-5cda-b6b1-a6196a1729e9"
 Blobs = "1"
 ChainRulesCore = "1"
 KLU = "0.6"
+Libdl = "1"
+LinearAlgebra = "1"
 MacroTools = "0.5"
 Preferences = "1"
-SSGraphBLAS_jll = "10.5"
+Random = "1"
+Serialization = "1"
+SparseArrays = "1"
 SpecialFunctions = "2"
+SSGraphBLAS_jll = "10.5"
 StorageOrders = "0.3.2"
+SuiteSparse = "1"
 julia = "1.10"

src/SuiteSparseGraphBLAS.jl

@@ -109,7 +109,6 @@ include("solvers/klu.jl")
 include("solvers/umfpack.jl")
 include("solvers/cholmod.jl")
 
-export SparseArrayCompat
 export LibGraphBLAS
 # export UnaryOps, BinaryOps, Monoids, Semirings #Submodules
 export unaryop, binaryop, Monoid, semiring #UDFs
@@ -131,7 +130,7 @@ export firsti, firstj, secondi, secondj
 #SelectOps not found in Julia/stdlibs
 export offdiag
 
-export extract, extract!, subassign!, assign!, hvcat! #array functions
+export extract, extract!, subassign!, assign! #array functions
 
 #operations
 export select, select!, eadd, eadd!, emul, emul!, gbtranspose, gbtranspose!,

src/operators/binaryops.jl

@@ -4,9 +4,9 @@ using ..SuiteSparseGraphBLAS: isGxB, isGrB, TypedBinaryOperator, GBType,
     valid_vec, juliaop, gbtype, symtotype, Itypes, Ftypes, Ztypes, FZtypes, Rtypes, optype,
     Ntypes, Ttypes, suffix, valid_union, GBArrayOrTranspose, binaryop
 using ..LibGraphBLAS
-export BinaryOp, binaryop
+export binaryop
 
-export second, rminus, iseq, isne, isgt, islt, isge, isle, ∨, ∧, lxor, xnor, fmod, 
+export second, rminus, iseq, isne, isgt, islt, isge, isle, ∨, ∧, lxor, xnor,
 bxnor, bget, bset, bclr, firsti0, firsti, firstj0, firstj, secondi0, secondi, secondj0, 
 secondj, pair
 

src/types.jl

@@ -991,13 +991,15 @@ StorageOrders.comptime_storageorder(::AbstractGBArray{<:Any, <:Any, O}) where O
 defaultfill(::Type{T}) where {Te, T <: GBVecOrMat{Te}} = defaultfill(Te)
 defaultfill(::Type{T}) where {Te, Tf, T <: GBVecOrMat{Te, Tf}} = defaultfill(Tf)
 
-defaultfill(::Type{T}) where T = zero(T)
-defaultfill(::Type{Nothing}) = nothing
-defaultfill(::Type{Missing}) = missing
+# Union{T, Nothing} and Union{T, Missing} default to zero(T) rather than nothing/missing.
 # This is bold, I'm not sure if I like it...
 # It boils down to whether we want numeric sparse arrays to be default
 # or for graph sparse arrays to be default.
 # I don't think it is onerous for graph algorithm writers to say `GBMatrix{Int64, Nothing}`,
 # and provides better defaults for other users.
-defaultfill(::Type{Union{T, Nothing}}) where T = T
-defaultfill(::Type{Union{T, Missing}}) where T = T
+function defaultfill(::Type{T}) where T
+    S = nonmissingtype(Base.nonnothingtype(T))
+    return S === T ? zero(T) : defaultfill(S)
+end
+defaultfill(::Type{Nothing}) = nothing
+defaultfill(::Type{Missing}) = missing