Are there any performance implications to the spacing of entity IDs?

[Net5F]

Hello, I'm trying to decide whether I should persist my entities with their entt::entity ID directly, or if I should give each entity a separate, persistent ID. The difference with a separate ID would be that, on startup, I could just use create(). If I persisted the entt::entity directly, I would need to use create(hint) and their IDs would be more sparse.

Are there any situations where the closeness of IDs has significant performance implications?

[skypjack]

There is a subtle difference actually to make it possible to recycle identifiers without having to keep track of all holes in your set.
When generating identifier N, it triggers the generation of all non-existing identifiers with an entity id lower than its own, if any exist.
It's not a big deal since you're going to generate them all anyway sooner or later. Therefore, I don't think it will cause you any problems.

[Net5F]

Perfect, thanks for the explanation 😄

[skypjack]

You're welcome.

[Net5F]

Funny timing, I just ran into a related question. If you create an entity as such: registry.create(static_cast<entt::entity>(1)), is registry.valid(static_cast<entt::entity>(0)) afterwards meant to return true? That's what I appear to be seeing, and I realized this is probably what you were explaining.

My game is networked, so I have a server creating entities locally and telling clients to replicate them. If one of my clients is told to create entity 20 followed by entity 15, it will do a check for if 15 already exists, then fail since it returns true. I supposed I could skip the registry.valid() check and just registry.create() and see if the returned ID matches the desired ID. Does that sound reasonable?

[skypjack]

In fact, it depends. If entity 0 exists already and got destroyed before you create entity 1, then the call to valid isn't guaranteed to return true.
If you do that on an empty registry instead, then yes, it returns true. In order to create entity 1, the registry also creates entity 0 and puts it in the list of available entities (the ones to return if you invoke create() without arguments). Therefore, technically speaking the identifier exists.
What you want to check is probably that is exists and that its index as in index(entity)is lower thanin_use(). What I want probably is a new in_use(entity)` function instead in order to avoid this ambiguity. 🙂

[Net5F]

Ah, that makes sense. Thanks again for the explanation 😄