Updaters and closures

scene.always

scene.always(m, |state, t| ...) makes shape m follow a function from then on. The closure is called every frame and returns m’s whole shape at time t, in seconds since the always call.

// A dot orbits the origin once every four seconds.
let at = |t| point((TAU * t / 4.0).cos(), (TAU * t / 4.0).sin()) * 2.0;
let d = scene.add(dot(at.call(0.0)).fill(YELLOW));
scene.always(d, |s, t| dot(at.call(t)).fill(YELLOW));
scene.wait(4.0);

Index state with a mobject to read another shape’s current state, so one shape can follow another:

let c = scene.add(circle(0.5));
let label = scene.add(dot(ORIGIN));
scene.always(label, |s, t| dot(ORIGIN).next_to(s[c], UP));
scene.play(shift(c, RIGHT * 3.0));

Pure in t

Every frame must be a function of t and the scene state alone. The timeline can be scrubbed, rendered from a marker or sampled out of order, so a closure cannot carry state from one frame to the next.

Manim updaters often accumulate (dot.shift(dt * v) every frame). Rewrite them in terms of t:

// Instead of moving the dot a little each frame, say where it is at time t.
scene.always(d, |s, t| dot(point(-4.0 + t, 0.0)));

growing_wave.rhai in fastanim-script/scenes plots f(t) = sin(2πt)·t this way: the dot and its trace are both functions of t.

Closures

Closures are values. Call them with .call:

let f = |x| x * 2.0;
print(f.call(1.5));   // 3.0

They capture variables from where they are written, so helpers like at above can be shared between updaters. Functions declared with fn cannot see outer variables; pass scene and anything else in as arguments.

fn fade_all(scene, mobs) {
    scene.play(mobs.map(|m| fade_out(m)));
    for m in mobs {
        scene.remove(m);
    }
}

scene itself cannot be used inside a closure that runs during playback (updaters, apply_function, update, graph functions). Use state instead.