Simplify handling of Δt in trajectory GUI: just use `stop_at_dt = tru…

…e` in `step!` (#249)

* Some fixes for the notebook output

* simplify Dt handling in GUI by not remaking the system

* documentation improvement

* even more clarification

* give instructions on how to make smooth animations

CHANGELOG.md

@@ -3,9 +3,16 @@
 The package DynamicalSystems.jl has little actual code.
 It mainly re-exports other packages. The source code contained here is only regarding graphical interactive applications for dynamical systems.
 
-The changelog here therefore lists either major changes to the overarching DynamicalSystems.jl ecosystem, or major changes to plotting infastructure.
+The changelog here therefore lists either major changes to the overarching DynamicalSystems.jl ecosystem, or major changes to plotting infrastructure.
 
-The changelogs of individual subpackages are self-contained for each package.
+The changelogs of individual sub-packages are self-contained for each package.
+
+# v3.4
+
+Some important fixes on the `interactive_trajectory` GUI:
+
+- Changed the internal handling for continuous time systems. Now they are stepped for exactly `Δt` by giving `true` as third input to `step!`. This increases consistency with discrete systems by not altering the integration protocol.
+- The documentation around the "run" button of the GUI has been clarified.
 
 # v3.3
 

Project.toml

@@ -1,7 +1,7 @@
 name = "DynamicalSystems"
 uuid = "61744808-ddfa-5f27-97ff-6e42cc95d634"
 repo = "https://github.com/JuliaDynamics/DynamicalSystems.jl.git"
-version = "3.3.26"
+version = "3.4.0"
 
 [deps]
 Attractors = "f3fd9213-ca85-4dba-9dfd-7fc91308fec7"

ext/src/dynamicalsystemobservable.jl

@@ -23,7 +23,7 @@ function DynamicalSystems.step!(dso::DynamicalSystemObservable, n::Int = 1)
     N = length(dso.tail_observables)
     # Always store values, but only update observables after loop
     for _ in 1:n
-        step!(dso.pds, Δt)
+        step!(dso.pds, Δt, true)
         for i in 1:N
             ob = dso.tail_observables[i]
             last_state = current_state(dso.pds, i)

ext/src/interactive_trajectory.jl

@@ -32,11 +32,6 @@ function DynamicalSystems.interactive_trajectory(
         throw(ArgumentError("State space plot can be up to 3 dimensional! Change `idxs`."))
     end
 
-    if ds isa CoupledODEs # force time evolution into non-adaptive
-        newdiffeq = (ds.diffeq..., adaptive = false, dt = Δt)
-        ds = CoupledODEs(ds, newdiffeq)
-    end
-
     p0 = initial_parameters(ds)
     pds = DynamicalSystems.ParallelDynamicalSystem(ds, u0s)
     u00s = deepcopy(current_states(pds))

src/visualizations.jl

@@ -32,34 +32,41 @@ See also [`interactive_trajectory`](@ref).
 
 ## Interactivity and time stepping keywords
 
-_GUI functionality is possible when the plotting backend is `GLMakie`.
-Do `using GLMakie; GLMakie.activate!()` to ensure this is the chosen backend._
+!!! note "GLMakie"
+    GUI functionality is possible when the plotting backend is `GLMakie` or `WGLMakie`.
+    Do `using GLMakie; GLMakie.activate!()` to ensure this is the chosen backend.
 
 - `add_controls = true`: If `true`, below the state space axis
   some buttons for animating the trajectories live are added:
-  - `reset`: results the parallel trajectories to their initial conditions
-  - `run`: when clicked it evolves the trajectories forwards in time indefinitely.
-    click again to stop the evolution.
-  - `step`: when clicked it evolves the trajectories forwards in time for the
-    amount of steps chosen by the slider to its right.
+  - "reset": results the parallel trajectories to their initial conditions.
+  - "run": when clicked it evolves the trajectories forwards in time indefinitely.
+    click again to stop the evolution. Note that "run" simply presses the "step"
+    button indefinitely, and hence the visual progress you see on the screen depends
+    on the value of the "steps" slider. Thus, while the animation "runs" you can
+    increase/decrease the "steps" slider to increase/decrease the animation speed.
+    For smooth animations for continuous time systems you should have small `Δt` and large `tail`.
+  - "step": when clicked it evolves the trajectories forwards in time for the
+    amount of steps chosen by the "steps" slider to its right. Each step is an
+    evolution of `Δt` unit of time long (and clicking the "step" button may do
+    more than one steps according to the slider).
   The plotted trajectories can always be evolved
-  manually using the custom animations etup that we describe below; `add_controls` only
+  manually using the custom animations setup that we describe below; `add_controls` only
   concerns the buttons and interactivity added to the created figure.
 - `parameter_sliders = nothing`: If given, it must be a dictionary, mapping
   parameter indices (any valid index that can be given to [`set_parameter!`](@ref))
   to ranges of parameter values. Each combination of index and range becomes a slider
   that can be interactively controlled to alter a system parameter on the fly during time
   evolution. Below the parameter sliders, three buttons are added for GUI usage:
-  - `update`: when clicked the chosen parameter values are propagated into the system
-  - `u.r.s.`: when clicked it is equivalent with clicking in order: "update", "reset", "step".
-  - `reset p`: when clicked it resets
+  - "update": when clicked the chosen parameter values are propagated into the system.
+  - "u.r.s.": when clicked it is equivalent with clicking in order: "update", "reset", "step".
+  - "reset p": when clicked it resets parameters to their initial values.
   Parameters can also be altered using the custom animation setup that we describe below;
   `parameter_sliders` only conserns the buttons and interactivity added to the created figure.
 - `parameter_names = Dict(keys(ps) .=> string.(keys(ps)))`: Dictionary mapping parameter
   keys to labels. Only used if `parameter_sliders` is given.
 - `Δt`: Time step of time evolution. Defaults to 1 for discrete time,
-  0.01 for continuous time systems. For internal simplicity, continuous time dynamical
-  systems are evolved non-adaptively with constant step size equal to `Δt`.
+  0.01 for continuous time systems. Continuous time dynamical
+  systems are stepped for exactly `Δt` time (third argument to `step!` is `true`).
 - `pause = nothing`: If given, it must be a real number. This number is given to the `sleep`
   function, which is called between each plot update. Useful when time integration is
   computationally inexpensive and animation proceeds too fast.