arvidsson / mix Goto Github PK

To something less generic than "entity"

Thus we don't really need to expose World to the System

execution 1:
auto e = world.CreateEntity();
e.AddComponent(100, 100); - ID = 0
e.AddComponent(10, 10); - ID = 1

execution 2:
auto e = world.CreateEntity();
e.AddComponent(10, 10); - ID = 0 // different from the last time
e.AddComponent(100, 100); - ID = 1 // different from the last time

This may cause problems if something like serialization of the entities and pools takes place. Though it can be worked around.

You can map the type_index to the runtime generated ones, but the type_index hash function returns a size_t type which is not consistent on 32 bit and 64 bit architectures.
My suggestion is some kind of consistent uint32_t hash on the component's type name but that would require some dependency on a rtti/reflection library. Your call :). Though a really nice library :)

I'm thinking of making the library single-header for ease of use in projects, similar to how https://github.com/vurtun/nuklear does it.

Any thoughts?

i.e. no assertions but rather return invalid entities

Currently an entity is added/removed to/from a system only when it is created/destroyed. It should also be added/removed when a component matching the requirements is added/removed to the entity(maybe at a later point of the entity's existance).

With inspiration from https://github.com/libgdx/ashley

struct PhysicsBodyComponent
{
// some general data
};
struct RigidBodyComponent : public PhysicsBodyComponent
{
// some specific data
};
struct SoftBodyComponent : public PhysicsBodyComponent
{
// some specific data
};
auto e = world.CreateEntity();
e.AddComponent();

// somewhere else maybe a system's requirements
RequireComponent();
auto& component = e.GetComponent();