Re:Are we game yet? – A guide to the Rust game development e...
by meheleventyone on Wednesday February 10, 2021 @04:17PM
(#26094940)
Yeah the confusion comes because entities are no longer a concrete element of the program but a concept tying components together. It’s just as valid to look at a subset of the components that are referenced by an entity id as a view as it is to look at them all. Like a relational database.
And by breaking things into components the granularity of ownership is increased compared with the equivalent concrete representation of the same data. So whilst you can’t reason about ownership of an entity as it primarily exists conceptually the ownership of its constituent parts is well defined.
I don't see how that's the case [1]. The entities are owned by the ECS (or the game state, or whatever), not by each other. Entities can conceptually reference each other, but not own each other. The only time this is problematic, is when one entity is destroyed while another one is still holding a reference to it. At the beginning of the video, he discussed basically all approaches to solve that problem:
1) Raw pointers. Bad for obvious reasons.
2) Smart pointers which keep the referenced object alive. No good, he says, because one entity should not keep another one alive, if the game logic says it should be removed.
3) Weak pointers which are safely invalidated when the referenced entity is removed. No good, he says, because keeping track of back-references is inefficient.
4) Weak pointers which check whether the referenced entity is still alive before accessing it. Which is exactly what the ECS does. Rust's ownership semantics still help here, because entities are unambiguously owned by the ECS, which returns a type-safe None value when you try to access a deleted entity.
[1] I'm not arguing against you, indy; I understand that you paraphrased Blow's argument.