float(S) and complex(S) should not copy data unnecessarily (#340)

Author: stevengj

src/sparsematrix.jl

@@ -970,8 +970,8 @@ convert(T::Type{<:LowerTriangular}, m::AbstractSparseMatrixCSC) = m isa T ? m :
 convert(T::Type{<:UpperTriangular}, m::AbstractSparseMatrixCSC) = m isa T ? m :
     istriu(m) ? T(m) : throw(ArgumentError("matrix cannot be represented as UpperTriangular"))
 
-float(S::SparseMatrixCSC) = SparseMatrixCSC(size(S, 1), size(S, 2), copy(getcolptr(S)), copy(rowvals(S)), float.(nonzeros(S)))
-complex(S::SparseMatrixCSC) = SparseMatrixCSC(size(S, 1), size(S, 2), copy(getcolptr(S)), copy(rowvals(S)), complex(copy(nonzeros(S))))
+float(S::SparseMatrixCSC) = SparseMatrixCSC(size(S, 1), size(S, 2), getcolptr(S), rowvals(S), float(nonzeros(S)))
+complex(S::SparseMatrixCSC) = SparseMatrixCSC(size(S, 1), size(S, 2), getcolptr(S), rowvals(S), complex(nonzeros(S)))
 
 """
     sparse(A)

src/sparsevector.jl

@@ -1086,11 +1086,11 @@ copy(x::AbstractSparseVector) = if _is_fixed(x)
 
 float(x::AbstractSparseVector{<:AbstractFloat}) = x
 float(x::AbstractSparseVector) =
-    SparseVector(length(x), copy(nonzeroinds(x)), float(nonzeros(x)))
+    SparseVector(length(x), nonzeroinds(x), float(nonzeros(x)))
 
 complex(x::AbstractSparseVector{<:Complex}) = x
 complex(x::AbstractSparseVector) =
-    SparseVector(length(x), copy(nonzeroinds(x)), complex(nonzeros(x)))
+    SparseVector(length(x), nonzeroinds(x), complex(nonzeros(x)))
 
 
 ### Concatenation