Method to normalize zero vectors

[Yu266426]

Description
When normalizing a vector, it is necessary to first check it is not of length zero, otherwise the program will crash.

The game engine bevy has a method for its Vecs that is normalize_or_zero() that allows you to normalize a vector that is potentially of length zero, where it returns a zero vector if that is the case, otherwise normalizing it.

I think this would be useful to have in pygame, as it saves having to type an extra line pretty much every time you want to normalize a vector.

[bilhox]

Mathematically speaking, it's impossible to normalize a zero vector, so returning a zero vector would be weird. That's why, I don't agree with you.

[Yu266426]

Well, I don't agree in the sense that just because something doesn't "make sense mathematically" means that a more convenient method shouldn't exist.

But also, that's why the method is called normalize_OR_ZERO, which I think helps illustrate that it isn't just normalizing it.

[bigwhoopgames]

I would be in support of having this method added. In nearly ever instance of me using normalize() I use a zero vector if I can't normalize.

[Matiiss]

I like the idea of having a way to conveniently get the zero vector back when normalizing, the proposed method's name is quite clear in what it would do and thus I don't see how it wouldn't make sense mathematically either, you either can normalize or you return a zero vector.

An alternative idea would be to add an optional argument to normalize, like zero that would achieve the same thing: vec.normalize(zero=True) (perhaps, it could be a positional-only argument for performance: vec.normalize(True) albeit that makes it less readable, but maybe using keyword args doesn't make it all that slow, so that should be tested)

[DaNubCoding]

Perhaps xnormalize is also an option, meaning "exclusive normalize", where it will normalize everything excluding zero vector in which case it is handled differently. This is easier to type and doesn't clutter code with long method names. Stuff like FRect uses a similar method of naming.

[Mega-JC]

I'm not sure if pygame needs this to be honest, especially if there is already a super simple solution to this problem of getting a potential ValueError: Simply checking the vector's thuthiness using an if statement, or short-circuiting with and by writing the expression vector and vector.normalize(). In the latter case, if vector is a zero vector, the expression will simply return vector. If vector is not a zero vector, vector.normalize() will be returned instead. This seems to be the exact behavior desired here.

Another example:

>>> from pygame import Vector2
>>> 
>>> vec_a = Vector2(5, 0)
>>> vec_a.normalize()
Vector2(1, 0)
>>> 
>>> vec_b = Vector2(0, 0)
>>> vec_b.normalize()
Traceback (most recent call last):
  File "<pyshell#21>", line 1, in <module>
    vec_b.normalize()
ValueError: Can't normalize Vector of length Zero
>>>
>>> vec_a_norm_or_zero = vec_a and vec_a.normalize()
>>> vec_a_norm_or_zero
Vector2(1, 0)
>>>
>>> vec_b_norm_or_zero = vec_b and vec_b.normalize()
>>> vec_b_norm_or_zero
Vector2(0, 0)

[bilhox]

I think however we should have prebuilt Vectors (Like in Unity I think), like Vector2.zero (0 , 0) for example. So it would more easy to compare and check if a vector in null. (or a method that returns true if it's zero ?)

[itzpr3d4t0r]

I think however we should have prebuilt Vectors (Like in Unity I think), like Vector2.zero (0 , 0) for example. So it would more easy to compare and check if a vector in null. (or a method that returns true if it's zero ?)

There's no need for such method, you can already do if not vector:, vectors support bool checks.

[Matiiss]

I think however we should have prebuilt Vectors (Like in Unity I think), like Vector2.zero (0 , 0) for example. So it would more easy to compare and check if a vector in null. (or a method that returns true if it's zero ?)

Aside from vectors supporting boolean checks, you can get a zero vector by simply instantiating a vector: pygame.Vector2()

[DaNubCoding]

Simply checking the vector's thuthiness using an if statement, or short-circuiting with and by writing the expression vector and vector.normalize().

These methods work, however they are either cumbersome to write when normalize is used frequently (if statement solution) or simply unclear to the reader as to what they are doing (and solution).

[Mega-JC]

or simply unclear to the reader as to what they are doing (and solution).

To clarify, vector and vector.normalize() (short-circuting with and) is the same as vector.normalize() if vector else vector (ternary expression), for those who prefer the longer, more explicit syntax. The former just seemed to be the shortest and the easiest to substitute a bare .normalize() method call with in a long sequence of vector operations (Yu266426 implied wanting a concise solution).

[DaNubCoding]

To clarify, vector and vector.normalize() (short-circuting with and) is the same as vector.normalize() if vector else vector (ternary expression), for those who prefer the longer, more explicit syntax. The former just seemed to be the shortest and the easiest to substitute a bare .normalize() method call with in a long sequence of vector operations (Yu266426 implied wanting a concise solution).

Yes the short-circuiting solution is a good solution as of now, but I think it can be made clearer with the implementation of a normalize_or_zero(). It is indeed a more concise version of the ternary solution, but it loses some clarity in comparison.

[JiffyRob]

Just want to point out, if we did implement this, would we also want to do something similar with scale_to_length()? If we don't it's a simple normalize(zero=True) * length but the whole point of this issue is conciseness. I would be in favor of the normalize method just because I use it so often and end up erring out and having to write boilerplate for it constantly. You're right to say that it's pretty simple just to short circuit, but when it is needed so dang often the trade-off makes less sense. Also many pygame users are just learning python and don't know how short circuiting works. At the very least I would put a note in the docs.