Move SparseArrays dependency to extension

Project.toml

@@ -1,24 +1,25 @@
 name = "LazyArrays"
 uuid = "5078a376-72f3-5289-bfd5-ec5146d43c02"
-version = "2.6.2"
+version = "2.7.0"
 
 [deps]
 ArrayLayouts = "4c555306-a7a7-4459-81d9-ec55ddd5c99a"
 FillArrays = "1a297f60-69ca-5386-bcde-b61e274b549b"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 MacroTools = "1914dd2f-81c6-5fcd-8719-6d5c9610ff09"
-SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
 
 [weakdeps]
 BandedMatrices = "aae01518-5342-5314-be14-df237901396f"
 BlockArrays = "8e7c35d0-a365-5155-bbbb-fb81a777f24e"
 BlockBandedMatrices = "ffab5731-97b5-5995-9138-79e8c1846df0"
 StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
+SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
 
 [extensions]
 LazyArraysBandedMatricesExt = "BandedMatrices"
 LazyArraysBlockArraysExt = "BlockArrays"
 LazyArraysBlockBandedMatricesExt = "BlockBandedMatrices"
+LazyArraysSparseArraysExt = "SparseArrays"
 LazyArraysStaticArraysExt = "StaticArrays"
 
 [compat]
@@ -48,8 +49,9 @@ BlockBandedMatrices = "ffab5731-97b5-5995-9138-79e8c1846df0"
 Infinities = "e1ba4f0e-776d-440f-acd9-e1d2e9742647"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
+SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 Tracker = "9f7883ad-71c0-57eb-9f7f-b5c9e6d3789c"
 
 [targets]
-test = ["Aqua", "BandedMatrices", "Base64", "BlockArrays", "BlockBandedMatrices", "StaticArrays", "Tracker", "Test", "Infinities", "Random"]
+test = ["Aqua", "BandedMatrices", "Base64", "BlockArrays", "BlockBandedMatrices", "StaticArrays", "SparseArrays", "Tracker", "Test", "Infinities", "Random"]

ext/LazyArraysSparseArraysExt.jl

@@ -0,0 +1,13 @@
+module LazyArraysSparseArraysExt
+
+using SparseArrays: issparse, nnz, AbstractSparseArray
+import LazyArrays: local_issparse, local_nnz
+
+#
+# Add methods to local_nnz and local_issparse calling back to the corresponding
+# methods defined in SparseArrays in case SparseArrays is loaded.
+# 
+local_nnz(A::AbstractSparseArray) = nnz(A)
+local_issparse(A::AbstractArray) = issparse(A)
+
+end

src/LazyArrays.jl

@@ -3,7 +3,7 @@ module LazyArrays
 # Use README as the docstring of the module:
 @doc read(joinpath(dirname(@__DIR__), "README.md"), String) LazyArrays
 
-using Base.Broadcast, LinearAlgebra, FillArrays, ArrayLayouts, SparseArrays
+using Base.Broadcast, LinearAlgebra, FillArrays, ArrayLayouts
 import LinearAlgebra.BLAS
 
 import Base: *, +, -, /, <, ==, >, \, ≤, ≥, (:), @_gc_preserve_begin, @_gc_preserve_end, @propagate_inbounds,

src/lazyoperations.jl

@@ -161,6 +161,16 @@ function copy(M::Mul{ApplyLayout{typeof(kron)}})
     return shuffle_algorithm(algo_type, M.A, M.B, eltype(M))
 end
 
+#
+# The following functions introduce a local version of `issparse` and `nnz` which allows to use
+# them without 'using SparseArrays'. If SparseArrays is loaded, the
+# LazyArraysSparseArraysExt extension adds methods to these functions which
+# call back to `issparse` and `nnz` from SparseArrays for the appropriate types.
+# This design allows to use e.g. LinearSolve.jl together with LazyArrays.jl without
+# triggering the SparseArrays extension of LinearSolve.
+#
+local_issparse(A) = false
+local_nnz(A) = length(A)
 
 function shuffle_algorithm(
     ::ModifiedShuffle, K::Kron{T,2} where T, p::AbstractVecOrMat, OT::Type{<:Number}
@@ -184,11 +194,11 @@ function shuffle_algorithm(
     R_H::Vector{Vector{Int}} = []
     C_H::Vector{Vector{Int}} = []
 
-    is_dense = !any(issparse, K.args)
+    is_dense = !any(local_issparse, K.args)
 
     # note: the following computation costs are for multiplication against
     #       a single vector.
-    nnz_ = [issparse(X_h) ? nnz(X_h) : prod(size(X_h)) for X_h in K.args]
+    nnz_ = [local_issparse(X_h) ? local_nnz(X_h) : prod(size(X_h)) for X_h in K.args]
     trad_cost = 2*sum([
         prod(size.(K.args[1:h-1], 2)) * nnz_[h] * prod(size.(K.args[h+1:end], 1))
         for (h, X_h) in enumerate(K.args)
@@ -226,7 +236,7 @@ function shuffle_algorithm(
         return shuffle_algorithm(Shuffle(), K, p, OT)
     end
 
-    nnz_m = [issparse(X_h) ? nnz(X_h) : prod(size(X_h)) for X_h in K_shrunk_factors]
+    nnz_m = [local_issparse(X_h) ? local_nnz(X_h) : prod(size(X_h)) for X_h in K_shrunk_factors]
     modified_cost = 2*sum([
         prod(length.(C_H[1:h-1])) * nnz_m[h] * prod(length.(R_H[h+1:end]))
         for (h, X_h) in enumerate(K.args)