nthperm!(a, k) overwrites a with the kth permutation. This is difficult to reason about if what one wants is to iterate over k and get the permutations of a without minimal allocations.
For instance, doing something like this is not a good idea:
julia> v = [1, 2, 3]
julia> for k in 1:6
println(nthperm!(v, k))
end
[1, 2, 3]
[1, 3, 2]
[3, 1, 2] # <---
[1, 2, 3]
[3, 1, 2] # <--- woops, seen this already
[2, 1, 3]since it returns the same permutations multiple times (because v is being permuted).
I think a three-argument version of nthperm! would be nice. This could just piggy-back off the current 2-argument implementation:
function nthperm!(dst::Vector{T}, a::AbstractVector{T}, k::Integer) where T
nthperm!(copyto!(dst, a), k)
endI suppose one could argue it's a trivial function, but it took me a while to realize this would be the right way to integrate with the 2-argument method.
nthperm!(a, k)overwritesawith thekth permutation. This is difficult to reason about if what one wants is to iterate overkand get the permutations ofawithout minimal allocations.For instance, doing something like this is not a good idea:
since it returns the same permutations multiple times (because
vis being permuted).I think a three-argument version of
nthperm!would be nice. This could just piggy-back off the current 2-argument implementation:I suppose one could argue it's a trivial function, but it took me a while to realize this would be the right way to integrate with the 2-argument method.