Sparse matrix multiplication error; difference between sparse and dense matrix multiplication

[yurivish]

julia> a = ones(Bool, 3, 3);

julia> b = ones(Bool, 3, 3);

julia> a * b
3×3 Array{Int64,2}:
 3  3  3
 3  3  3
 3  3  3

julia> using SparseArrays

julia> sparse(a) * sparse(b)
ERROR: InexactError: Bool(Bool, 2)
Stacktrace:
 [1] Type at ./float.jl:73 [inlined]
 [2] convert at ./number.jl:7 [inlined]
 [3] setindex! at ./array.jl:745 [inlined]
 [4] #spmatmul#60(::Symbol, ::Function, ::SparseMatrixCSC{Bool,Int64}, ::SparseMatrixCSC{Bool,Int64}) at /Users/yurivish/Documents/julia/usr/share/julia/stdlib/v0.7/SparseArrays/src/linalg.jl:186
 [5] spmatmul at /Users/yurivish/Documents/julia/usr/share/julia/stdlib/v0.7/SparseArrays/src/linalg.jl:151 [inlined]
 [6] *(::SparseMatrixCSC{Bool,Int64}, ::SparseMatrixCSC{Bool,Int64}) at /Users/yurivish/Documents/julia/usr/share/julia/stdlib/v0.7/SparseArrays/src/linalg.jl:141
 [7] top-level scope at none:0

I would have expected sparse and dense matrix multiplication to work the same way, but instead the sparse version seems to require the resulting array to be of the same type as the input arrays.

I believe the behavior on dense matrices is correct; true + true + true == 3.

[stevengj]

matmul uses promote_op(matprod, eltype(A), eltype(B)), whereas spmatmul does not.

[stevengj]

On a related note, mixed-type sparse-matrix operations copy the whole input matrices to promoted types (using type-unstable assignments), which seems wasteful.

[KristofferC]

For mixed see JuliaLang/julia#24045

[ViralBShah]

Works now.