Is this a bug? A = sparse(Int32[], Int32[], Float64[], 2, 2) A + I # Loses Int32 A is of type SparseMatrixCSC{Float64,Int32} and A + I is of type SparseMatrixCSC{Float64,Int64}, losing the integer type. Thanks

julia> @which A+I +(A::SparseMatrixCSC, J::UniformScaling) in SparseArrays at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.0/SparseArrays/src/sparsematrix.jl:3500 This is the code: (+)(A::SparseMatrixCSC, J::UniformScaling) = A + sparse(J, size(A)...) Now julia> typeof(siz…

I don’t think that’s enough because sparse(I, Int32(3), Int32(3)) still has integer type Int64. Maybe it can be solved changing the definition of +, since SparseMatrixCSC{Float64,Int32}(I, 3, 3) works. Should I open an issue? I think I can fix it.

I don’t see how it is related to +. julia> @which sparse(I, Int32(3), Int32(3)) sparse(s::UniformScaling, m::Integer, n::Integer) in SparseArrays at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.0/SparseArrays/src/sparsematrix.jl:1558 This is the code: sparse(s::UniformScali…

I’m saying that instead of redefining SparseMatrixCSC{Tv,Ti} so that its sizes have type Ti, we simply change (+)(A::SparseMatrixCSC, J::UniformScaling) = A + sparse(J, size(A)...) to something like (+)(A::SparseMatrixCSC{Tv,Ti}, J::UniformScaling) where {Tv, Ti <: Integer} = A + SparseMatri…

[image] abelsiqueira: I’m saying that instead of redefining SparseMatrixCSC{Tv,Ti} so that its sizes have type Ti , we simply change But this would only work for +. What about -? The rest of the code should be made consistent. The first thing bothering me is that Dims is defined with Int64 b…

While it might solve your particular problem, it may not be a definitive solution. There are other methods that use Int64 (eg via Dims) and the number of rows and cols being stored as Int64 makes it unusable for BigInt indices (which are not practical for dense stored arrays but may be useful for sp…

UniformScaling with SparseMatrixCSC{T,Int32} loses integer type

Internals & Design

abelsiqueira January 10, 2019, 4:55pm 7

Yes, - as well. And I + A also. All three are together in the code.

Sparse vectors with Int128 indices

Topic		Replies	Views
`SparseArray` types Internals & Design type , staticarrays , internals , rfc , type-inference	5	358	July 16, 2026
Sparse matrix product causes unnecessary casting to "Any" thereby causing issues? General Usage sparse	11	694	July 15, 2021
Lufact SparseMatrixCSC Int32 Numerics bug , linearalgebra	9	1435	February 27, 2018
Index type of sparse matrix General Usage type , linearalgebra , sparse	7	1219	February 8, 2021
Dense .+ sparse -> sparse? Internals & Design	2	662	January 26, 2019

UniformScaling with SparseMatrixCSC{T,Int32} loses integer type

Related topics