Sorry for my wrong wording and the error in the code snippet.
By elementwise, I meant: A < B <=> for all i: a[i] < b[i].
why not? That is the standard partial order for functions. Array may be considered as a function that is mapping its index set to its values.
In a narrow set of contexts, this comparison could useful for arrays, but in the standard library I don’t think it is useful to define a < function in which most arguments are unordered.
1 Like
True. That is the same argument, which let me dislike to define isless for arrays in general (even vectors).
1 Like
There is another argument: If we define isless/cmp for AbstractArray that will be usable for all subtypes. Many of them would cry for specialized implementations for performance reasons. The current list:
julia> subtypes(AbstractArray)
29-element Array{Union{DataType, UnionAll},1}:
AbstractRange
AbstractSparseArray
Adjoint
Base.LinAlg.AbstractQ
Base.LinAlg.AbstractTriangular
Base.LinAlg.HessenbergQ
Base.LinAlg.LQPackedQ
Base.LogicalIndex
Base.ReinterpretArray
Base.ReshapedArray
Bidiagonal
CartesianIndices
ConjArray
Core.Inference.AbstractRange
DenseArray
Diagonal
Hermitian
IterativeEigensolvers.AtA_or_AAt
IterativeEigensolvers.SVDAugmented
LinearIndices
PermutedDimsArray
RowVector
SubArray
SuiteSparse.CHOLMOD.FactorComponent
SymTridiagonal
Symmetric
Test.GenericArray
Transpose
Tridiagonal
julia>