VertexArray#[](index) should not return a copy of the Vertex

[Alan Quinot]

It seems VertexArray#[](index) returns a copy of Vertex at index instead of the original instance.

Look:

vertex_array = SF::VertexArray.new(SF::Quads, 4)

puts vertex_array[0].position.inspect
# SF::Vector2(Float32)(@x=0, @y=0)

vertex_array[0].position = SF.vector2(64, 128)

puts vertex_array[0].position.inspect
# SF::Vector2(Float32)(@x=0, @y=0)

The workaround is creating a new vertex and overwriting it:

vertex_array = SF::VertexArray.new(SF::Quads, 4)

puts vertex_array[0].position.inspect
# SF::Vector2(Float32)(@x=0, @y=0)

vertex = SF::Vertex.new
vertex.position = SF.vector2(64, 128)
vertex_array[0] = vertex

puts vertex_array[0].position.inspect
# SF::Vector2(Float32)(@x=64, @y=128)

[Oleh Prypin]

That is true. This is how structs work in Crystal. And it's not the only place where this problem is present.
I gave this a lot of thought today, and there is nothing I can do to change this. Making a class-based wrapper would be a huge performance hit.
The only way to not copy a struct is to use pointers, and that's just not an option.

C++ code relies a lot on the mutability of these structs. The mindset here should be different to avoid the need to edit structs.

Compare: original (broken), remade examples