Interactive Solution Inspection

An interactive solution inspection tool based on WGLMakie and Bonito is provided through the helper package NetworkDynamicsInspector.

First, we need to define the system we want to inspect.

Define some network, simulate it and get a solution object

using NetworkDynamics
using NetworkDynamicsInspector
using OrdinaryDiffEqTsit5
using Graphs

include(joinpath(pkgdir(NetworkDynamics), "test", "ComponentLibrary.jl"))
function get_sol(;limit=1.0)
    g = SimpleGraph([0 1 1 0 1;
                    1 0 1 1 0;
                    1 1 0 1 0;
                    0 1 1 0 1;
                    1 0 0 1 0])
    vs = [Lib.swing_mtk() for _ in 1:5];
    set_default!(vs[1], :Pmech, -1)
    set_default!(vs[2], :Pmech, 1.5)
    set_default!(vs[3], :Pmech, -1)
    set_default!(vs[4], :Pmech, -1)
    set_default!(vs[5], :Pmech, 1.5)
    ls = [Lib.line_mtk() for _ in 1:7];
    nw = Network(g, vs, ls)
    sinit = NWState(nw)
    s0 = find_fixpoint(nw)
    set_defaults!(nw, s0)

    # set_position!(vs[1], (0.0, 0.0))
    set_marker!(vs[1], :dtriangle)
    set_marker!(vs[2], :utriangle)
    set_marker!(vs[3], :dtriangle)
    set_marker!(vs[4], :dtriangle)
    set_marker!(vs[5], :utriangle)

    cond = ComponentCondition([:P, :₋P, :srcθ], [:limit, :K]) do u, p, t
        abs(u[:P]) - p[:limit]
    end
    affect = ComponentAffect([],[:active]) do u, p, ctx
        @info "Trip line $(ctx.eidx) between $(ctx.src) and $(ctx.dst) at t=$(ctx.t)"
        p[:active] = 0
    end
    cb = ContinousComponentCallback(cond, affect)
    set_callback!.(ls, Ref(cb))

    tripfirst = PresetTimeComponentCallback(1.0, affect) # reuse the same affect
    add_callback!(nw[EIndex(5)], tripfirst)

    nwcb = NetworkDynamics.get_callbacks(nw);
    s0 = NWState(nw)
    s0.p.e[:, :limit] .= limit

    prob = ODEProblem(nw, uflat(s0), (0,6), copy(pflat(s0)), callback=nwcb)
    sol = solve(prob, Tsit5())
end

sol = get_sol()

[ Info: Trip line 5 between 2 and 4 at t=1.0
[ Info: Trip line 7 between 4 and 5 at t=2.2476763961108466
[ Info: Trip line 4 between 2 and 3 at t=2.5025231919919295
[ Info: Trip line 1 between 1 and 2 at t=3.1947647110612936
[ Info: Trip line 3 between 1 and 5 at t=3.338052868169857
[ Info: Trip line 2 between 1 and 3 at t=3.4042696460061506

Now that we have an ODESolution sol, we can call inspect to open the inspector GUI. The docstring provides several options to customize how the app is displayed.

inspect(sol; reset=true)

[ Info: New GUI Session started

screenshot

Programmatic Access and GUI State Manipulation

Internally, the NetworkDynamicsInspector maintains a global reference to an AppState object. This AppState reflects changes made to the GUI by the user and can also be modified programmatically.

See the NetworkDynamicsInspector API for a complete list of available functions. A good starting point is the dump_app_state function, which helps you recreate a GUI state that was previously configured manually.

Let's say we've adjusted the AppState to include an additional time series plot for the node states.

We can dump the code which helps us to recreate the app state:

dump_app_state()

# To recreate the current state, run the following commands:

set_sol!(sol) # optional if after inspect(sol)
set_state!(; t=1.75, tmin=0.0, tmax=6.0)
set_graphplot!(; nstate=[:θ], estate=[:₋P], nstate_rel=false, estate_rel=false, ncolorrange=(-8.428582f0, 8.428582f0), ecolorrange=(-0.98994404f0, 0.98994404f0))
define_timeseries!([
    (; selcomp=[VIndex(1), VIndex(2), VIndex(3), VIndex(4), VIndex(5)], states=[:θ, :ω], rel=false),
    (; selcomp=[EIndex(1), EIndex(2), EIndex(3), EIndex(4), EIndex(5), EIndex(6), EIndex(7)], states=[:P], rel=false),
])

Now we can use this code to recreate the app state even though we've reseted it.

inspect(sol; reset=true)

"copy-paste and execute code returned by `dump_app_state` here"

screenshot