Refactor network data

[FHell]

This is a big refactor in preparation for work on adding more Type awareness to the system. Some of the old API that I believe we never used is deprecated in this version. The internal structure now relies on the Vertex|EdgeFunction types directly, and the network dynamics retains the full information in its internal fields.

All the logic for calculating the indices, etc, that was duplicated all over the place now has a clear place, the indices and internal variables were combined into a struct that lives in NetworkStructures

The tests work, but they are rather limited. Could you please check if your branch of PowerDynamics works against this? Also happy to hear about what you think about the design as such.

[janlisse]

When i run the branch against our PowerDynamics example i get this error:

ERROR: LoadError: UndefVarError: sparse not defined
Stacktrace:
[1] construct_mass_matrix(::Array{Union{Nothing, Array{Int64,1}},1}, ::Int64, ::GraphStructure) at /Users/jan/.julia/dev/NetworkDynamics/src/NetworkStructures.jl:77
 [2] network_dynamics(::Array{ODEVertex,1}, ::Array{StaticEdge,1}, ::LightGraphs.SimpleGraphs.SimpleGraph{Int64}) at /Users/jan/.julia/dev/NetworkDynamics/src/NetworkDynamics.jl:48
 [3] read_network_from_csv(::String, ::String) at /Users/jan/projects/julia/PowerDynBase.jl/src/parsers/csv_parser.jl:11
 [4] top-level scope at none:0
in expression starting at /Users/jan/projects/julia/PowerDynBase.jl/ieee14-minimal-example.jl:10

[janlisse]

Ok, i fixed the above error myself by adding a "using SparseArrays" to NetworkStructures.jl
Now i get another error:

ERROR: LoadError: MethodError: Cannot `convert` an object of type Nothing to an object of type Float64
Closest candidates are:
  convert(::Type{T<:Number}, ::T<:Number) where T<:Number at number.jl:6
  convert(::Type{T<:Number}, ::Number) where T<:Number at number.jl:7
  convert(::Type{T<:Number}, ::Base.TwicePrecision) where T<:Number at twiceprecision.jl:250
  ...
Stacktrace:
 [1] fill!(::SubArray{Float64,2,SparseMatrixCSC{Float64,Int64},Tuple{Array{Int32,1},Array{Int32,1}},false}, ::Nothing) at ./multidimensional.jl:837
 [2] copyto! at ./broadcast.jl:805 [inlined]
 [3] materialize!(::SubArray{Float64,2,SparseMatrixCSC{Float64,Int64},Tuple{Array{Int32,1},Array{Int32,1}},false}, ::Base.Broadcast.Broadcasted{Base.Broadcast.DefaultArrayStyle{0},Nothing,typeof(identity),Tuple{Base.RefValue{Nothing}}}) at ./broadcast.jl:756
 [4] construct_mass_matrix(::Array{Union{Nothing, Array{Int64,1}},1}, ::Int64, ::GraphStructure) at /Users/jan/.julia/dev/NetworkDynamics/src/NetworkStructures.jl:81

Before the refactoring the code handled the case of the mass_matrix being Nothing. This
special treatment has been removed and that is the cause for the error.

[janlisse]

I think the refactoring makes much sense and looks good to me in general!
We need to fix the above error though.
I also was wondering why you changed:
dim_nd = sum(dim_e) + sum(dim_v)
to:
dim_nd = sum(dim_v)
in
function network_dynamics(vertices!::Array{ODEVertex}, edges!::Array{ODEEdge}, graph)

Shouldn't the total dimension include the dynamic edge variables?

And one additional note: The DDE parts still use the old code path not the refactored version.

[FHell]

First one is a breaking API change. Before we had the incorrect default of mass_matrix=nothing, the correct default is mass_matrix=I, if you were passing mass_matrix=nothing manually this needs to be adjusted.

Second one is a copy paste bug. I will add a test for this.

[janlisse]

After adding a few more fixes our PowerDynamic examples are running again using this branch.
From my side it is now good to go for a merge. @FHell Shall we merge then?