Why do chains don't use B2ShapeId?

[reynarzz]

At the moment, when you create a chain using B2Shapes.b2CreateChain, it returns a B2ChainId instead of a B2ShapeId. This not only breaks the api's convention but also makes it difficult to create proper abstractions around shape creation.
Since both structs share the same structure, using B2ShapeId instead of B2ChainId for chains will "just work."

These two are essentially the same:

public readonly struct B2ChainId
{
    public readonly int index1;
    public readonly ushort world0;
    public readonly ushort generation;

    public B2ChainId(int index1, ushort world0, ushort generation)
    {
        this.index1 = index1;
        this.world0 = world0;
        this.generation = generation;
    }
}

And

public readonly struct B2ShapeId
{
    public readonly int index1;
    public readonly ushort world0;
    public readonly ushort generation;

    public B2ShapeId(int index1, ushort world0, ushort generation)
    {
        this.index1 = index1;
        this.world0 = world0;
        this.generation = generation;
    }


    public override string ToString()
    {
        return $"(i:{index1}, w:{world0}, g:{generation})";
    }
}

[ikpil]

I think it follows a different ID scheme because it is different from the existing Shapes.
Personally, I also believe this approach is safer, as it helps ensure type safety.

Since this is just our interpretation, could you ask the original author, Eric, about it?
I’d also like to understand from Eric’s perspective why it was designed this way. 🙂

[erincatto]

Chain are an aggregate of shapes and shapes are atomic. You could say it is a leaky abstraction but I think this is justified for a physics engine which has performance as the priority.