3.0 - Simple IHandle interface for XYZHandle structs

[aquagoose]

Summary of the PR

Adds a simple IHandle struct for XYZHandle structs, and implements an IsNull property.

Related issues, Discord discussions, or proposals

https://discord.com/channels/521092042781229087/587346162802229298/1443561471245291592

Further Comments

Only tested locally with SDL.

[Perksey]

Missing docs comments are resulting in compilation failure.

[Perksey]

Could we do this == default to make it agnostic to non-pointer-sized handles?

[Perksey]

Hmmm on second thought, maybe we could do IHandle<TSelf> : IEquatable<TSelf>, IEqualityOperators<TSelf, TSelf, bool>? At this point we totally negate the need for IsNull as users generalising over T: IHandle<T>>? Could maybe keep this non-generic interface and DIM it in the generic one if we want (is that a thing?) If we go that route no changes will be required to TransformHandles other than the base list.

[Exanite]

I recently worked on some unofficial HarfBuzz bindings using Silk 3 and HarfBuzz uses a null object pattern, which can be exposed as an IsNullOrEmpty(IHarfBuzz) or IsEmpty(IHarfBuzz) method on the HarfBuzz handles. This means having the IsNull property would be nice for consistency.

To be clear, not asking for the methods to be added to the IHandle interface. There likely will need to be a HarfBuzz-specific handle interface.

https://harfbuzz.github.io/object-model-lifecycle.html

Finally, object constructors (and, indeed, as much of the shaping API as possible) will never return NULL. Instead, if there is an allocation error, each constructor will return an “empty” object singleton.

HarfBuzz exposes these "empty" singletons as *GetEmpty() functions:

This matters more if we ever add HarfBuzz bindings officially to Silk, but does represent a potential edge case.

Also, open to having both the equality interfaces and the property/methods, but I have to give it some more thought.

[Perksey]

Is this still being worked on?