DecimalNumber with ValueTracker and updaters don't work

[VladislavKorecky]

I have the following code:

from manim import *


class NeuronFormula(Scene):
    @staticmethod
    def make_formula_group(first_term, second_term, third_term):
        return VGroup(first_term, second_term, third_term).arrange(RIGHT, buff=0.1).move_to([0, 0, 0])

    def construct(self):
        # create the original formula: wx+b
        formula = NeuronFormula.make_formula_group(MathTex("w"), MathTex("x+"), MathTex("b"))

        # set the target to be: ?x+?
        formula.target = NeuronFormula.make_formula_group(MathTex("?"), MathTex("x+"), MathTex("?"))

        # run the animations
        self.add(formula)
        self.wait(1)
        self.play(MoveToTarget(formula), run_time=0.5)
        self.wait(1)

        a = ValueTracker(0)
        b = ValueTracker(0)

        a_targets = [3, -3] * 3
        b_targets = [3, -3] * 3

        # create and transform the formula to 0.00x+0.00
        a_display = DecimalNumber(a.get_value())
        b_display = DecimalNumber(b.get_value())
        formula.target = NeuronFormula.make_formula_group(a_display, MathTex("x+"), b_display)
        self.play(MoveToTarget(formula))

        a_display.add_updater(
            lambda mob: mob.set_value(a.get_value())
        )
        b_display.add_updater(
            lambda mob: mob.set_value(b.get_value())
        )

        # I have to use ApplyMethod instead of .animate syntax because I'm animating the same mobject
        # https://stackoverflow.com/a/77002707/16343968
        a_value_animations = Succession(*[ApplyMethod(a.set_value, target) for target in a_targets], lag_ratio=1)
        b_value_animations = Succession(*[ApplyMethod(b.set_value, target) for target in b_targets], lag_ratio=1)

        self.play(a_value_animations, b_value_animations)

The code produces the following video:
https://github.com/ManimCommunity/manim/assets/55646812/7f5e603f-c905-49fc-b470-6982ec50d5db

I expected the numbers (0.00) to change based on my list of targets, but it stays still for several seconds, why?

[uwezi]

I think your problem is in your use of MoveToTarget which I have never seen or used before. From the description of this animation class:

(class) MoveToTarget

Transforms a mobject to the mobject stored in its target attribute.

After calling the ~.Mobject.generate_target method, the target attribute of the mobject is populated with a copy of it. After modifying the attribute,
playing the .MoveToTarget animation transforms the original mobject
into the modified one stored in the target attribute.

The emphasis is on is populated with a copy of it which means that your original a_display which gets the updater is never actually added to the scene, but only a copy of the object from the time before it got an updater attached to it.

If you use the ReplacementTransform instead, it works.... (for the correct alignment you will have to work a bit on your decimal numbers)

class NeuronFormula(Scene):


    @staticmethod
    def make_formula_group(first_term, second_term, third_term):
        return VGroup(first_term, second_term, third_term).arrange(RIGHT, buff=0.1).move_to([0, 0, 0])

    def construct(self):
        formula = NeuronFormula.make_formula_group(MathTex("w"), MathTex("x+"), MathTex("b"))

        formula.target = NeuronFormula.make_formula_group(MathTex("?"), MathTex("x+"), MathTex("?"))

        self.add(formula)
        self.wait(1)
        self.play(MoveToTarget(formula), run_time=0.5)
        self.wait(1)

        a = ValueTracker(0)
        b = ValueTracker(0)

        a_targets = [3, -3] * 3
        b_targets = [3, -3] * 3

        a_display = DecimalNumber(a.get_value())
        b_display = DecimalNumber(b.get_value())

        self.play(
            ReplacementTransform(
                formula,
                NeuronFormula.make_formula_group(a_display, MathTex("x+"), b_display)
            )
        )

        a_display.add_updater(
            lambda mob: mob.set_value(a.get_value())
        )
        b_display.add_updater(
            lambda mob: mob.set_value(b.get_value())
        )

        a_value_animations = Succession(*[ApplyMethod(a.set_value, target) for target in a_targets], lag_ratio=1)
        b_value_animations = Succession(*[ApplyMethod(b.set_value, target) for target in b_targets], lag_ratio=1)

        self.play(a_value_animations, b_value_animations)

NeuronFormula.mp4

[VladislavKorecky]

Wow, that was tricky. The weird thing is that I'm not even using generate_target(), so there should be no copies. But it works, and that's what matters. Thanks!

[uwezi]

it is the .MoveToTarget() animation which places the copy of what was previously in the .target property into the object itself.