MonoGame.Extended
MonoGame.Extended copied to clipboard
craftworkgames
[Core] Duplicate `struct`s for things already available in `Microsoft.Xna.Framework`
For example :
Why have a Size and a Size2 if they solve the same purpose?
Why have a Point2 if Vector2 is much the same?
It's kinda weird to have different structs to do the exact same thing as something that exists already in the API, but with different names. I haven't checked yet, but I'm scared of seeing the dependencies those objects have.
Hopefully, this will be adressed with 2.0.
Honestly, I'm still on the fence about this. I'll certainly consider it for 2.0.
Similar to #650
Point2 will be removed in favour of Vector2. I made previous comments about why Point2 was created here. With System.Numerics, the responsibility of Vector2 is taken over by the .NET team.
Actually I think keeping Point2 would be good with the new direction of v4. The issue of why Point2 exists when Vector2 exists could be better explained via documentation.
It's kinda weird to have different structs to do the exact same thing as something that exists already in the API, but with different names. I haven't checked yet, but I'm scared of seeing the dependencies those objects have.
This is the exact problem being addressed with v4 by forking FNA. Vector2 however will be picked up from System.Numerics while other structs will be removed from the Microsoft.Xna.Framework namespace such as Rectangle.
I have actually been looking into the .NET 5 code behind the System.Numerics namespace. There are loads of JIT optimizations and intrinsics running behind the scenes which greatly improve performance. I did a few benchmarks about a month ago and a duplicate of any of the System.Numerics structs, identical down to the IL code, runs about 75-50% slower than the legitimate version.
I know I haven't worked actively on the project in a few years, but hopefully this helps! 😁
@stefanrbk That's one major driving force for sure. I'm creating a game (from a fork of FNA) and re-doing some of the code in Extended. I think I would rather drop Point2 after looking at all the code I would have to maintain. I think that the problem is just an educational problem. I could tackle the "Point vs Vector" problem with proper documentation.
To recap: From a mathematical perspective, a "point" is not a "vector," but they are similar!
A mathematical vector describes a length and a direction. An arrow can usefully represent a vector. Two vectors are equal if they have the same length and direction, even if we think of them as being located in different places. Vectors are often used to store a displacement or offset in space. Points, on the other hand, are locations in space. The difference is that points are anchored to some origin while vectors don't have any meaning of an origin. Game developers commonly think of a vector as a point by saying that zero is the origin, but vectors could be starting at any origin, not just zero. For the mathematically curious, see https://math.stackexchange.com/questions/645672/what-is-the-difference-between-a-point-and-a-vector and http://immersivemath.com/ila/ch02_vectors/ch02.html.
The pragmatic approach here is that if you come across a vector that is masking as a point, you should think of it as a displacement from zero, the result of which is a location. However, it would be best to take care that a vector in any other context does not have to start from the origin of zero.
So the solution here is that System.Numerics with Vector2 will be used instead of Point2 or Size or Size2 or anything like that. A Vector2 is a displacement which could encode any of Point2, Size or Size2, but the context is very important; that is where proper documentation will be required.
Something for that education piece there (although this does contain a bit of speculation/assumption from me, but it makes more sense to me):
Vector2 and all the "Vector*" data structures in System.Numberics are referring to a "vector" in terms of computer science, not mathematics. Either it is referring to C++ where a Vector is similar to a List<T> in C#, or it refers to the act of doing 1 CPU operation on multiple pieces of data at the same time called either "SIMD" or "vector operations".
I'm thinking any kind of explanation when it comes to documentation could be as simple as "There are multiple conflicting definition of vectors in computer science and programming. Here it means a specified number of float values passed around together."
A vector does not have to be Euclidean or spatial. The components of a vector could be any tuple such that the components respect addition and multiplication operations so that they form a vector space. The std::vector in C++ is just a dynamic array with amortized constant time; the equivalent of List<T> in C#. The name vector was IMO a poor choice for a dynamic array.
I completely agree. I have been working on SIMD with LOADS of VectorXXX<T> use, and every time I try searching for "vectors" I get std::vector more often than not =/