Something off with broadcast of ArrayPartition with NTuple about recursivearraytools.jl

[ChrisRackauckas]

Interesting. I can't say I know why this would occur.

[tomaklutfu]

Let me chase down a bit more then. Now I see there are some discrepancies for broadcast here and Base

julia> Base.Broadcast.combine_styles(((NTuple{5, Float64})))
Base.Broadcast.Style{Tuple}()

julia> RecursiveArrayTools.combine_styles(((NTuple{5, Float64}).parameters...,))
Base.Broadcast.DefaultArrayStyle{0}()

May be this is a peculiar and desired situation of ArrayPartition. I think, OrdinaryDiffEq should have just returned Array{NTuple{5,Float64}, 1} similar to normal array initial conditions (u0::Array{Float64, 1} gives sol.u::Array{Array{Float64, 1}, 1}). If that's the case I should open an issue to OrdinaryDiffEq.

[tomaklutfu]

I meant to ask first. Should I open issue to OrdinaryDiffEq.jl?

[ChrisRackauckas]

This is a DifferentialEquations.jl issue, that we shouldn't be changing from tuple in DiffEqBase. Please open that issue. I hope to fix it soon.