Add Val::Zero

[jim-ec]

What problem does this solve or what need does it fill?

When requesting a UI length of zero, there is no clear choice of unit.

In CSS e.g., the zero literal does not have to be followed by any unit:

p {
  right: 0;
}

What solution would you like?

Add the Val::Zero case to the ui_node::Val enum.

What alternative(s) have you considered?

Just continue using Val::Px(0.0) / Val::Percent(0.0) / etc.

Additional context

https://stackoverflow.com/questions/7923042/when-specifying-a-0-value-in-css-should-i-explicitly-mark-the-units-or-omit

[ickshonpe]

Having different representations of zero is annoying but I'm not sure another variant would help. Like now we not only have Val::Px(0.) or Val::Percent(0.) or Val::VMin(0.) etc to choose from but also Val::Zero, is that progress?

I've suggested before that we change all the UI functions to take Into<Val> and implement:

impl From<f32> for Val {
    fn from(value: f32) -> Self {
        Val::Px(value)
    }
}

to enable:

margin: UiRect::all(0.0),

It wasn't a popular idea though and I withdrew the PR.

[jim-ec]

Link to PR mentioned above: #7500

[jim-ec]

As far as I see it, the problem with #7500 is the ambiguity between units. Val::from(2.0) should mean pixels, but that was not clear from context. For zero, however, the unit does not matter, so the problem with unit ambiguity vanishes.

Of course, it is somewhat paradoxical to complain about too many options and trying to fix the problem by adding another option. However, I personally find it more elegant to use Val::Zero instead of deciding on a specific unit, but I guess that is a purely personal preference. That's why I am raising this issue, to see how others think about it.

[mockersf]

Val::ZERO could be a const value that means Val::Px(0.0)

[ickshonpe]

Val::ZERO could be a const value that means Val::Px(0.0)

Yeah, that's not so bad.

[ickshonpe]

This got me thinking, and I realised that the Val arithmetic functions are wrong. We should be able to add numeric Vals of zero to other numeric Vals, regardless of variant: #9534 (needs reviews)

[alice-i-cecile]

I think we should go with the const approach instead of another variant, because of the multiple representations problem.