Nested containers in the system function

[dextorious]

Currently, using a nested container such as a Vector{SVector{2,Float64}} produces the following error when using a method with an adaptive timestep:

ERROR: DimensionMismatch("Cannot left-divide transposed vector by matrix")
Stacktrace:
 [1] /(::Array{Float64,1}, ::Array{Float64,1}) at .\linalg\generic.jl:823
 [2] broadcast_t(::Function, ::Type{Any}, ::Tuple{Base.OneTo{Int64}}, ::CartesianRange{CartesianIndex{1}}, ::Array{Array{Float64,1},1}, ::Array{Array{Float64,1},1}) at .\broadcast.jl:256
 [3] broadcast_c at .\broadcast.jl:319 [inlined]
 [4] broadcast at .\broadcast.jl:434 [inlined]
 [5] ode_determine_initdt(::Array{Array{Float64,1},1}, ::Float64, ::Float64, ::Float64, ::Float64, ::Float64, ::DiffEqBase.#ODE_DEFAULT_NORM, ::DiffEqBase.ODEProblem{Array{Array{Float64,1},1},Float64,false,DiffEqBase.ParameterizedFunction{false,BacterialEnsembles.#ensemble_dynamics#2,Tuple{Int64,Float64}},Void,UniformScaling{Int64}}, ::Int64) at C:\Users\admin\.julia\v0.6\OrdinaryDiffEq\src\initdt.jl:57
 [6] #init#134(::Int64, ::Array{Float64,1}, ::Array{Float64,1}, ::Array{Float64,1}, ::Void, ::Bool, ::Void, ::Bool, ::Bool, ::Bool, ::Float64, ::Bool, ::Rational{Int64}, ::Void, ::Void, ::Int64, ::Rational{Int64}, ::Rational{Int64}, ::Bool, ::Rational{Int64}, ::Rational{Int64}, ::Int64, ::Float64, ::Float64, ::DiffEqBase.#ODE_DEFAULT_NORM, ::DiffEqBase.#ODE_DEFAULT_ISOUTOFDOMAIN, ::DiffEqBase.#ODE_DEFAULT_UNSTABLE_CHECK, ::Bool, ::Bool, ::Bool, ::Bool, ::Bool, ::Bool, ::Bool, ::Int64, ::String, ::DiffEqBase.#ODE_DEFAULT_PROG_MESSAGE, ::Void, ::Void, ::Bool, ::Bool, ::Array{Any,1}, ::DiffEqBase.#init, ::DiffEqBase.ODEProblem{Array{Array{Float64,1},1},Float64,false,DiffEqBase.ParameterizedFunction{false,BacterialEnsembles.#ensemble_dynamics#2,Tuple{Int64,Float64}},Void,UniformScaling{Int64}}, ::OrdinaryDiffEq.Tsit5, ::Array{Any,1}, ::Array{Any,1}, ::Array{Any,1}, ::Type{Val{true}}) at C:\Users\admin\.julia\v0.6\OrdinaryDiffEq\src\solve.jl:137

The problem does not affect methods with a fixed timestep.

[ChrisRackauckas]

Can't reproduce your exact error. Are you using out of place?

[dextorious]

Yes, that was out of place.

[ChrisRackauckas]

./ is a hack in the adaptivity that makes out-of-place work with array equations even though it's written for Number. To make that work for "infinite nests", there needs to be a "recursive broadcast".

In-place is easier here: a one-line change makes it work. I need to see what exactly that effects in terms of AbstractArray types though.

[ChrisRackauckas]

JuliaLang/julia#22622

We'll see who comes just to say that the application is just too bizarre to care.

[ChrisRackauckas]

We're going to take that the proper way to handle this is by using a VectorOfArray or an ArrayPartition. Vector{SVector} won't be possible to support in general because it will not broadcast correctly, while those containers will do it well.