Right way to transform keyword arguments

General Usage
1

If f(xf...;of...) is going to call g(xg...; og...) after applying some transformations xf->xg and of->og, what is the most efficient way to do this? of happens to be a pairs iterator of NamedTuples, so direct map on it doesn’t work. Here is what I managed but I don’t think it is portable:

function f(xf...; of...)
    xg = map(transform,xf)
    og = pairs(map(transform,of.data))
    g(xg...; og...)
end
2

Not sure if this is useful but I think you can also splat the NamedTuple as keyword arguments:

julia> f(; a=1, b=2) = a-b
f (generic function with 1 method)

julia> f(; (b=10, a=20)...)
10

Is of.data is the recommended way to convert the iterator to NamedTuple? Maybe there should be some exported accessor function.

3

How would it be with

function f(xf...; of...)
    xg = map(transform,xf)
    og = (transform(o) for in pairs(o))
    g(xg...; og...)
end

Here (transform(o) for in pairs(o)) is a chain of two iterators which then gets splatted again, so this should theoretically be not bad.

4

values(of) appears to give the underlying named tuple (like of.data), so you could use that to make it portable.

function f(xf...; of...)
    xg = map(transform,xf)
    og = pairs(map(transform,values(of)))
    g(xg...; og...)
end

In terms of efficiency, I tried with generator expressions but your map version was better by a long way.

(For reference -

function f2(xf...; of...)
    xg = map(transform,xf)
    og = (key => transform(value) for (key, value) in of)
    g(xg...; og...)
end

it wasn’t very efficient.)

5

Thanks for all the responses. Here is a benchmark comparing different suggestions:

julia> @benchmark og = (x[1]=>sum(x[2]) for x in of)
  minimum time:     71.074 ns (0.00% GC)
julia> @benchmark collect(x[1]=>sum(x[2]) for x in of) # note that this gives an array
  minimum time:     260.378 ns (0.00% GC)
julia> @benchmark og = pairs(map(sum, values(of)))
  minimum time:     118.039 ns (0.00% GC)
julia> @benchmark og = collect(map(sum, values(of)))
  minimum time:     130.588 ns (0.00% GC)