Vector2::ZERO.normalized() returns Vector2(NaN, NaN)

[Shou]

In Godot, Vector2.ZERO.normalized() returns Vector2.ZERO but the Rust equivalent uses glam which seems to be the cause for returning NaN (but I haven't verified). If that's the case, we should probably use normalize_or and return zero as the fallback instead.

This affects any function that uses normalized() as well, like direction_to().

[Bromeon]

gdnative is currently in maintenance mode (if at all), meaning we're quite hesitant to apply any changes that can be potentially breaking. This behavior has existed for a long time, and it's possible that some code relies on it. It's also documented as such (even if not explicitly, but that expression is NaN as well):

May I ask in which context you encountered this?

[Bromeon]

Note that in gdext (Godot 4 bindings), we indeed have multiple overloads with very explicit docs:

[Shou]

May I ask in which context you encountered this?

I have methods for pulling/pushing bodies towards/away from a specified position, and when that position is equal direction_to() ends up making the vector NaN. I wouldn't mind making a PR, but if it's unlikely to be accepted I'm fine either way. I mostly wanted to document this behavior as I tried searching for it here first.

[Bromeon]

We could maybe add a normalized_or_zero method just like in gdext, but I'd rather not change the semantics of the existing one right now 🙂

Documenting would be a good idea, too.

[Bromeon]

@Shou are you interested in a PR that adds normalized_or_zero?