Utility functions

arg_n(n, args...)

returns a modified version of args, such that scalars remain scalar but in vectors the n'th position is picked. This is useful for calling separable functions with their scalar arguments differing for each vector entry.

Example

jdoctest julia> args = (1,(4,5)) (1, (4, 5)) julia> collect(SeparableFunctions.arg_n(2, args)) 2-element Vector{Int64}: 1 5

kwarg_n(n, args...)

returns a modified version of keyword-args kwargs, such that scalar values remain scalar but in vectors the n'th position is picked. This is useful for calling separable functions with their scalar arguments differing for each vector entry.

# Example

julia> kw = (a=1,b=(4,5))
(a = 1, b = (4, 5))
julia> SeparableFunctions.kwarg_n(2, kw)
(a = 1, b = 5)

pick_n(n, v)

picks the nth value of the vector v or return the scalar v.

get_corner_ranges(sz::NTuple{N}; shifted_dims = zeros(Bool, N), inv_dims = zeros(Bool, N), full_dims=zeros(Bool, N)) where {N}

returns a tuple of ranges that can be used for indexing various corners of an N-dimensional dataset.

Arguments

• sz: total size of the data
• shifted_dims: an iterable of Boolean defining which dimension to shift (by half the datasize)
• inv_dims: an iterable of Boolean defining which which range dimension
• full_dims: an iterable of Boolean defining which dimension to include fully as a colon (:)

copy_last_dim(arr::AbstractArray{T,N}) where{T,N}

mirrors the last dimension exploiting the 1D-nature of the trailing dimension using 1D indexing of the ND array. However even-size dimensions need special treatment.